xref: /openbmc/qemu/target/i386/ops_sse.h (revision b0dd0a3d)
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 SIZE 8
26 #define XMM_ONLY(...)
27 #define B(n) MMX_B(n)
28 #define W(n) MMX_W(n)
29 #define L(n) MMX_L(n)
30 #define Q(n) MMX_Q(n)
31 #define SUFFIX _mmx
32 #else
33 #define Reg ZMMReg
34 #define SIZE 16
35 #define XMM_ONLY(...) __VA_ARGS__
36 #define B(n) ZMM_B(n)
37 #define W(n) ZMM_W(n)
38 #define L(n) ZMM_L(n)
39 #define Q(n) ZMM_Q(n)
40 #define SUFFIX _xmm
41 #endif
42 
43 /*
44  * Copy the relevant parts of a Reg value around. In the case where
45  * sizeof(Reg) > SIZE, these helpers operate only on the lower bytes of
46  * a 64 byte ZMMReg, so we must copy only those and keep the top bytes
47  * untouched in the guest-visible destination destination register.
48  * Note that the "lower bytes" are placed last in memory on big-endian
49  * hosts, which store the vector backwards in memory.  In that case the
50  * copy *starts* at B(SIZE - 1) and ends at B(0), the opposite of
51  * the little-endian case.
52  */
53 #if HOST_BIG_ENDIAN
54 #define MOVE(d, r) memcpy(&((d).B(SIZE - 1)), &(r).B(SIZE - 1), SIZE)
55 #else
56 #define MOVE(d, r) memcpy(&(d).B(0), &(r).B(0), SIZE)
57 #endif
58 
59 void glue(helper_psrlw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
60 {
61     int shift;
62 
63     if (s->Q(0) > 15) {
64         d->Q(0) = 0;
65 #if SHIFT == 1
66         d->Q(1) = 0;
67 #endif
68     } else {
69         shift = s->B(0);
70         d->W(0) >>= shift;
71         d->W(1) >>= shift;
72         d->W(2) >>= shift;
73         d->W(3) >>= shift;
74 #if SHIFT == 1
75         d->W(4) >>= shift;
76         d->W(5) >>= shift;
77         d->W(6) >>= shift;
78         d->W(7) >>= shift;
79 #endif
80     }
81 }
82 
83 void glue(helper_psraw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
84 {
85     int shift;
86 
87     if (s->Q(0) > 15) {
88         shift = 15;
89     } else {
90         shift = s->B(0);
91     }
92     d->W(0) = (int16_t)d->W(0) >> shift;
93     d->W(1) = (int16_t)d->W(1) >> shift;
94     d->W(2) = (int16_t)d->W(2) >> shift;
95     d->W(3) = (int16_t)d->W(3) >> shift;
96 #if SHIFT == 1
97     d->W(4) = (int16_t)d->W(4) >> shift;
98     d->W(5) = (int16_t)d->W(5) >> shift;
99     d->W(6) = (int16_t)d->W(6) >> shift;
100     d->W(7) = (int16_t)d->W(7) >> shift;
101 #endif
102 }
103 
104 void glue(helper_psllw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
105 {
106     int shift;
107 
108     if (s->Q(0) > 15) {
109         d->Q(0) = 0;
110 #if SHIFT == 1
111         d->Q(1) = 0;
112 #endif
113     } else {
114         shift = s->B(0);
115         d->W(0) <<= shift;
116         d->W(1) <<= shift;
117         d->W(2) <<= shift;
118         d->W(3) <<= shift;
119 #if SHIFT == 1
120         d->W(4) <<= shift;
121         d->W(5) <<= shift;
122         d->W(6) <<= shift;
123         d->W(7) <<= shift;
124 #endif
125     }
126 }
127 
128 void glue(helper_psrld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
129 {
130     int shift;
131 
132     if (s->Q(0) > 31) {
133         d->Q(0) = 0;
134 #if SHIFT == 1
135         d->Q(1) = 0;
136 #endif
137     } else {
138         shift = s->B(0);
139         d->L(0) >>= shift;
140         d->L(1) >>= shift;
141 #if SHIFT == 1
142         d->L(2) >>= shift;
143         d->L(3) >>= shift;
144 #endif
145     }
146 }
147 
148 void glue(helper_psrad, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
149 {
150     int shift;
151 
152     if (s->Q(0) > 31) {
153         shift = 31;
154     } else {
155         shift = s->B(0);
156     }
157     d->L(0) = (int32_t)d->L(0) >> shift;
158     d->L(1) = (int32_t)d->L(1) >> shift;
159 #if SHIFT == 1
160     d->L(2) = (int32_t)d->L(2) >> shift;
161     d->L(3) = (int32_t)d->L(3) >> shift;
162 #endif
163 }
164 
165 void glue(helper_pslld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
166 {
167     int shift;
168 
169     if (s->Q(0) > 31) {
170         d->Q(0) = 0;
171 #if SHIFT == 1
172         d->Q(1) = 0;
173 #endif
174     } else {
175         shift = s->B(0);
176         d->L(0) <<= shift;
177         d->L(1) <<= shift;
178 #if SHIFT == 1
179         d->L(2) <<= shift;
180         d->L(3) <<= shift;
181 #endif
182     }
183 }
184 
185 void glue(helper_psrlq, 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 void glue(helper_psllq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
204 {
205     int shift;
206 
207     if (s->Q(0) > 63) {
208         d->Q(0) = 0;
209 #if SHIFT == 1
210         d->Q(1) = 0;
211 #endif
212     } else {
213         shift = s->B(0);
214         d->Q(0) <<= shift;
215 #if SHIFT == 1
216         d->Q(1) <<= shift;
217 #endif
218     }
219 }
220 
221 #if SHIFT == 1
222 void glue(helper_psrldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
223 {
224     int shift, i;
225 
226     shift = s->L(0);
227     if (shift > 16) {
228         shift = 16;
229     }
230     for (i = 0; i < 16 - shift; i++) {
231         d->B(i) = d->B(i + shift);
232     }
233     for (i = 16 - shift; i < 16; i++) {
234         d->B(i) = 0;
235     }
236 }
237 
238 void glue(helper_pslldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
239 {
240     int shift, i;
241 
242     shift = s->L(0);
243     if (shift > 16) {
244         shift = 16;
245     }
246     for (i = 15; i >= shift; i--) {
247         d->B(i) = d->B(i - shift);
248     }
249     for (i = 0; i < shift; i++) {
250         d->B(i) = 0;
251     }
252 }
253 #endif
254 
255 #define SSE_HELPER_B(name, F)                                   \
256     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
257     {                                                           \
258         d->B(0) = F(d->B(0), s->B(0));                          \
259         d->B(1) = F(d->B(1), s->B(1));                          \
260         d->B(2) = F(d->B(2), s->B(2));                          \
261         d->B(3) = F(d->B(3), s->B(3));                          \
262         d->B(4) = F(d->B(4), s->B(4));                          \
263         d->B(5) = F(d->B(5), s->B(5));                          \
264         d->B(6) = F(d->B(6), s->B(6));                          \
265         d->B(7) = F(d->B(7), s->B(7));                          \
266         XMM_ONLY(                                               \
267                  d->B(8) = F(d->B(8), s->B(8));                 \
268                  d->B(9) = F(d->B(9), s->B(9));                 \
269                  d->B(10) = F(d->B(10), s->B(10));              \
270                  d->B(11) = F(d->B(11), s->B(11));              \
271                  d->B(12) = F(d->B(12), s->B(12));              \
272                  d->B(13) = F(d->B(13), s->B(13));              \
273                  d->B(14) = F(d->B(14), s->B(14));              \
274                  d->B(15) = F(d->B(15), s->B(15));              \
275                                                         )       \
276             }
277 
278 #define SSE_HELPER_W(name, F)                                   \
279     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
280     {                                                           \
281         d->W(0) = F(d->W(0), s->W(0));                          \
282         d->W(1) = F(d->W(1), s->W(1));                          \
283         d->W(2) = F(d->W(2), s->W(2));                          \
284         d->W(3) = F(d->W(3), s->W(3));                          \
285         XMM_ONLY(                                               \
286                  d->W(4) = F(d->W(4), s->W(4));                 \
287                  d->W(5) = F(d->W(5), s->W(5));                 \
288                  d->W(6) = F(d->W(6), s->W(6));                 \
289                  d->W(7) = F(d->W(7), s->W(7));                 \
290                                                         )       \
291             }
292 
293 #define SSE_HELPER_L(name, F)                                   \
294     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
295     {                                                           \
296         d->L(0) = F(d->L(0), s->L(0));                          \
297         d->L(1) = F(d->L(1), s->L(1));                          \
298         XMM_ONLY(                                               \
299                  d->L(2) = F(d->L(2), s->L(2));                 \
300                  d->L(3) = F(d->L(3), s->L(3));                 \
301                                                         )       \
302             }
303 
304 #define SSE_HELPER_Q(name, F)                                   \
305     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)   \
306     {                                                           \
307         d->Q(0) = F(d->Q(0), s->Q(0));                          \
308         XMM_ONLY(                                               \
309                  d->Q(1) = F(d->Q(1), s->Q(1));                 \
310                                                         )       \
311             }
312 
313 #if SHIFT == 0
314 static inline int satub(int x)
315 {
316     if (x < 0) {
317         return 0;
318     } else if (x > 255) {
319         return 255;
320     } else {
321         return x;
322     }
323 }
324 
325 static inline int satuw(int x)
326 {
327     if (x < 0) {
328         return 0;
329     } else if (x > 65535) {
330         return 65535;
331     } else {
332         return x;
333     }
334 }
335 
336 static inline int satsb(int x)
337 {
338     if (x < -128) {
339         return -128;
340     } else if (x > 127) {
341         return 127;
342     } else {
343         return x;
344     }
345 }
346 
347 static inline int satsw(int x)
348 {
349     if (x < -32768) {
350         return -32768;
351     } else if (x > 32767) {
352         return 32767;
353     } else {
354         return x;
355     }
356 }
357 
358 #define FADD(a, b) ((a) + (b))
359 #define FADDUB(a, b) satub((a) + (b))
360 #define FADDUW(a, b) satuw((a) + (b))
361 #define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
362 #define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
363 
364 #define FSUB(a, b) ((a) - (b))
365 #define FSUBUB(a, b) satub((a) - (b))
366 #define FSUBUW(a, b) satuw((a) - (b))
367 #define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
368 #define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
369 #define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
370 #define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
371 #define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
372 #define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
373 
374 #define FAND(a, b) ((a) & (b))
375 #define FANDN(a, b) ((~(a)) & (b))
376 #define FOR(a, b) ((a) | (b))
377 #define FXOR(a, b) ((a) ^ (b))
378 
379 #define FCMPGTB(a, b) ((int8_t)(a) > (int8_t)(b) ? -1 : 0)
380 #define FCMPGTW(a, b) ((int16_t)(a) > (int16_t)(b) ? -1 : 0)
381 #define FCMPGTL(a, b) ((int32_t)(a) > (int32_t)(b) ? -1 : 0)
382 #define FCMPEQ(a, b) ((a) == (b) ? -1 : 0)
383 
384 #define FMULLW(a, b) ((a) * (b))
385 #define FMULHRW(a, b) (((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16)
386 #define FMULHUW(a, b) ((a) * (b) >> 16)
387 #define FMULHW(a, b) ((int16_t)(a) * (int16_t)(b) >> 16)
388 
389 #define FAVG(a, b) (((a) + (b) + 1) >> 1)
390 #endif
391 
392 SSE_HELPER_B(helper_paddb, FADD)
393 SSE_HELPER_W(helper_paddw, FADD)
394 SSE_HELPER_L(helper_paddl, FADD)
395 SSE_HELPER_Q(helper_paddq, FADD)
396 
397 SSE_HELPER_B(helper_psubb, FSUB)
398 SSE_HELPER_W(helper_psubw, FSUB)
399 SSE_HELPER_L(helper_psubl, FSUB)
400 SSE_HELPER_Q(helper_psubq, FSUB)
401 
402 SSE_HELPER_B(helper_paddusb, FADDUB)
403 SSE_HELPER_B(helper_paddsb, FADDSB)
404 SSE_HELPER_B(helper_psubusb, FSUBUB)
405 SSE_HELPER_B(helper_psubsb, FSUBSB)
406 
407 SSE_HELPER_W(helper_paddusw, FADDUW)
408 SSE_HELPER_W(helper_paddsw, FADDSW)
409 SSE_HELPER_W(helper_psubusw, FSUBUW)
410 SSE_HELPER_W(helper_psubsw, FSUBSW)
411 
412 SSE_HELPER_B(helper_pminub, FMINUB)
413 SSE_HELPER_B(helper_pmaxub, FMAXUB)
414 
415 SSE_HELPER_W(helper_pminsw, FMINSW)
416 SSE_HELPER_W(helper_pmaxsw, FMAXSW)
417 
418 SSE_HELPER_Q(helper_pand, FAND)
419 SSE_HELPER_Q(helper_pandn, FANDN)
420 SSE_HELPER_Q(helper_por, FOR)
421 SSE_HELPER_Q(helper_pxor, FXOR)
422 
423 SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
424 SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
425 SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
426 
427 SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
428 SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
429 SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
430 
431 SSE_HELPER_W(helper_pmullw, FMULLW)
432 #if SHIFT == 0
433 SSE_HELPER_W(helper_pmulhrw, FMULHRW)
434 #endif
435 SSE_HELPER_W(helper_pmulhuw, FMULHUW)
436 SSE_HELPER_W(helper_pmulhw, FMULHW)
437 
438 SSE_HELPER_B(helper_pavgb, FAVG)
439 SSE_HELPER_W(helper_pavgw, FAVG)
440 
441 void glue(helper_pmuludq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
442 {
443     d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
444 #if SHIFT == 1
445     d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
446 #endif
447 }
448 
449 void glue(helper_pmaddwd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
450 {
451     int i;
452 
453     for (i = 0; i < (2 << SHIFT); i++) {
454         d->L(i) = (int16_t)s->W(2 * i) * (int16_t)d->W(2 * i) +
455             (int16_t)s->W(2 * i + 1) * (int16_t)d->W(2 * i + 1);
456     }
457 }
458 
459 #if SHIFT == 0
460 static inline int abs1(int a)
461 {
462     if (a < 0) {
463         return -a;
464     } else {
465         return a;
466     }
467 }
468 #endif
469 void glue(helper_psadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
470 {
471     unsigned int val;
472 
473     val = 0;
474     val += abs1(d->B(0) - s->B(0));
475     val += abs1(d->B(1) - s->B(1));
476     val += abs1(d->B(2) - s->B(2));
477     val += abs1(d->B(3) - s->B(3));
478     val += abs1(d->B(4) - s->B(4));
479     val += abs1(d->B(5) - s->B(5));
480     val += abs1(d->B(6) - s->B(6));
481     val += abs1(d->B(7) - s->B(7));
482     d->Q(0) = val;
483 #if SHIFT == 1
484     val = 0;
485     val += abs1(d->B(8) - s->B(8));
486     val += abs1(d->B(9) - s->B(9));
487     val += abs1(d->B(10) - s->B(10));
488     val += abs1(d->B(11) - s->B(11));
489     val += abs1(d->B(12) - s->B(12));
490     val += abs1(d->B(13) - s->B(13));
491     val += abs1(d->B(14) - s->B(14));
492     val += abs1(d->B(15) - s->B(15));
493     d->Q(1) = val;
494 #endif
495 }
496 
497 void glue(helper_maskmov, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
498                                   target_ulong a0)
499 {
500     int i;
501 
502     for (i = 0; i < (8 << SHIFT); i++) {
503         if (s->B(i) & 0x80) {
504             cpu_stb_data_ra(env, a0 + i, d->B(i), GETPC());
505         }
506     }
507 }
508 
509 void glue(helper_movl_mm_T0, SUFFIX)(Reg *d, uint32_t val)
510 {
511     d->L(0) = val;
512     d->L(1) = 0;
513 #if SHIFT == 1
514     d->Q(1) = 0;
515 #endif
516 }
517 
518 #ifdef TARGET_X86_64
519 void glue(helper_movq_mm_T0, SUFFIX)(Reg *d, uint64_t val)
520 {
521     d->Q(0) = val;
522 #if SHIFT == 1
523     d->Q(1) = 0;
524 #endif
525 }
526 #endif
527 
528 #if SHIFT == 0
529 void glue(helper_pshufw, SUFFIX)(Reg *d, Reg *s, int order)
530 {
531     Reg r;
532 
533     r.W(0) = s->W(order & 3);
534     r.W(1) = s->W((order >> 2) & 3);
535     r.W(2) = s->W((order >> 4) & 3);
536     r.W(3) = s->W((order >> 6) & 3);
537     MOVE(*d, r);
538 }
539 #else
540 void helper_shufps(Reg *d, Reg *s, int order)
541 {
542     Reg r;
543 
544     r.L(0) = d->L(order & 3);
545     r.L(1) = d->L((order >> 2) & 3);
546     r.L(2) = s->L((order >> 4) & 3);
547     r.L(3) = s->L((order >> 6) & 3);
548     MOVE(*d, r);
549 }
550 
551 void helper_shufpd(Reg *d, Reg *s, int order)
552 {
553     Reg r;
554 
555     r.Q(0) = d->Q(order & 1);
556     r.Q(1) = s->Q((order >> 1) & 1);
557     MOVE(*d, r);
558 }
559 
560 void glue(helper_pshufd, SUFFIX)(Reg *d, Reg *s, int order)
561 {
562     Reg r;
563 
564     r.L(0) = s->L(order & 3);
565     r.L(1) = s->L((order >> 2) & 3);
566     r.L(2) = s->L((order >> 4) & 3);
567     r.L(3) = s->L((order >> 6) & 3);
568     MOVE(*d, r);
569 }
570 
571 void glue(helper_pshuflw, SUFFIX)(Reg *d, Reg *s, int order)
572 {
573     Reg r;
574 
575     r.W(0) = s->W(order & 3);
576     r.W(1) = s->W((order >> 2) & 3);
577     r.W(2) = s->W((order >> 4) & 3);
578     r.W(3) = s->W((order >> 6) & 3);
579     r.Q(1) = s->Q(1);
580     MOVE(*d, r);
581 }
582 
583 void glue(helper_pshufhw, SUFFIX)(Reg *d, Reg *s, int order)
584 {
585     Reg r;
586 
587     r.Q(0) = s->Q(0);
588     r.W(4) = s->W(4 + (order & 3));
589     r.W(5) = s->W(4 + ((order >> 2) & 3));
590     r.W(6) = s->W(4 + ((order >> 4) & 3));
591     r.W(7) = s->W(4 + ((order >> 6) & 3));
592     MOVE(*d, r);
593 }
594 #endif
595 
596 #if SHIFT == 1
597 /* FPU ops */
598 /* XXX: not accurate */
599 
600 #define SSE_HELPER_S(name, F)                                           \
601     void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s)        \
602     {                                                                   \
603         d->ZMM_S(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
604         d->ZMM_S(1) = F(32, d->ZMM_S(1), s->ZMM_S(1));                  \
605         d->ZMM_S(2) = F(32, d->ZMM_S(2), s->ZMM_S(2));                  \
606         d->ZMM_S(3) = F(32, d->ZMM_S(3), s->ZMM_S(3));                  \
607     }                                                                   \
608                                                                         \
609     void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s)        \
610     {                                                                   \
611         d->ZMM_S(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
612     }                                                                   \
613                                                                         \
614     void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s)        \
615     {                                                                   \
616         d->ZMM_D(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
617         d->ZMM_D(1) = F(64, d->ZMM_D(1), s->ZMM_D(1));                  \
618     }                                                                   \
619                                                                         \
620     void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s)        \
621     {                                                                   \
622         d->ZMM_D(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
623     }
624 
625 #define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
626 #define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
627 #define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
628 #define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
629 #define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
630 
631 /* Note that the choice of comparison op here is important to get the
632  * special cases right: for min and max Intel specifies that (-0,0),
633  * (NaN, anything) and (anything, NaN) return the second argument.
634  */
635 #define FPU_MIN(size, a, b)                                     \
636     (float ## size ## _lt(a, b, &env->sse_status) ? (a) : (b))
637 #define FPU_MAX(size, a, b)                                     \
638     (float ## size ## _lt(b, a, &env->sse_status) ? (a) : (b))
639 
640 SSE_HELPER_S(add, FPU_ADD)
641 SSE_HELPER_S(sub, FPU_SUB)
642 SSE_HELPER_S(mul, FPU_MUL)
643 SSE_HELPER_S(div, FPU_DIV)
644 SSE_HELPER_S(min, FPU_MIN)
645 SSE_HELPER_S(max, FPU_MAX)
646 SSE_HELPER_S(sqrt, FPU_SQRT)
647 
648 
649 /* float to float conversions */
650 void helper_cvtps2pd(CPUX86State *env, Reg *d, Reg *s)
651 {
652     float32 s0, s1;
653 
654     s0 = s->ZMM_S(0);
655     s1 = s->ZMM_S(1);
656     d->ZMM_D(0) = float32_to_float64(s0, &env->sse_status);
657     d->ZMM_D(1) = float32_to_float64(s1, &env->sse_status);
658 }
659 
660 void helper_cvtpd2ps(CPUX86State *env, Reg *d, Reg *s)
661 {
662     d->ZMM_S(0) = float64_to_float32(s->ZMM_D(0), &env->sse_status);
663     d->ZMM_S(1) = float64_to_float32(s->ZMM_D(1), &env->sse_status);
664     d->Q(1) = 0;
665 }
666 
667 void helper_cvtss2sd(CPUX86State *env, Reg *d, Reg *s)
668 {
669     d->ZMM_D(0) = float32_to_float64(s->ZMM_S(0), &env->sse_status);
670 }
671 
672 void helper_cvtsd2ss(CPUX86State *env, Reg *d, Reg *s)
673 {
674     d->ZMM_S(0) = float64_to_float32(s->ZMM_D(0), &env->sse_status);
675 }
676 
677 /* integer to float */
678 void helper_cvtdq2ps(CPUX86State *env, Reg *d, Reg *s)
679 {
680     d->ZMM_S(0) = int32_to_float32(s->ZMM_L(0), &env->sse_status);
681     d->ZMM_S(1) = int32_to_float32(s->ZMM_L(1), &env->sse_status);
682     d->ZMM_S(2) = int32_to_float32(s->ZMM_L(2), &env->sse_status);
683     d->ZMM_S(3) = int32_to_float32(s->ZMM_L(3), &env->sse_status);
684 }
685 
686 void helper_cvtdq2pd(CPUX86State *env, Reg *d, Reg *s)
687 {
688     int32_t l0, l1;
689 
690     l0 = (int32_t)s->ZMM_L(0);
691     l1 = (int32_t)s->ZMM_L(1);
692     d->ZMM_D(0) = int32_to_float64(l0, &env->sse_status);
693     d->ZMM_D(1) = int32_to_float64(l1, &env->sse_status);
694 }
695 
696 void helper_cvtpi2ps(CPUX86State *env, ZMMReg *d, MMXReg *s)
697 {
698     d->ZMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
699     d->ZMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
700 }
701 
702 void helper_cvtpi2pd(CPUX86State *env, ZMMReg *d, MMXReg *s)
703 {
704     d->ZMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
705     d->ZMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
706 }
707 
708 void helper_cvtsi2ss(CPUX86State *env, ZMMReg *d, uint32_t val)
709 {
710     d->ZMM_S(0) = int32_to_float32(val, &env->sse_status);
711 }
712 
713 void helper_cvtsi2sd(CPUX86State *env, ZMMReg *d, uint32_t val)
714 {
715     d->ZMM_D(0) = int32_to_float64(val, &env->sse_status);
716 }
717 
718 #ifdef TARGET_X86_64
719 void helper_cvtsq2ss(CPUX86State *env, ZMMReg *d, uint64_t val)
720 {
721     d->ZMM_S(0) = int64_to_float32(val, &env->sse_status);
722 }
723 
724 void helper_cvtsq2sd(CPUX86State *env, ZMMReg *d, uint64_t val)
725 {
726     d->ZMM_D(0) = int64_to_float64(val, &env->sse_status);
727 }
728 #endif
729 
730 /* float to integer */
731 
732 /*
733  * x86 mandates that we return the indefinite integer value for the result
734  * of any float-to-integer conversion that raises the 'invalid' exception.
735  * Wrap the softfloat functions to get this behaviour.
736  */
737 #define WRAP_FLOATCONV(RETTYPE, FN, FLOATTYPE, INDEFVALUE)              \
738     static inline RETTYPE x86_##FN(FLOATTYPE a, float_status *s)        \
739     {                                                                   \
740         int oldflags, newflags;                                         \
741         RETTYPE r;                                                      \
742                                                                         \
743         oldflags = get_float_exception_flags(s);                        \
744         set_float_exception_flags(0, s);                                \
745         r = FN(a, s);                                                   \
746         newflags = get_float_exception_flags(s);                        \
747         if (newflags & float_flag_invalid) {                            \
748             r = INDEFVALUE;                                             \
749         }                                                               \
750         set_float_exception_flags(newflags | oldflags, s);              \
751         return r;                                                       \
752     }
753 
754 WRAP_FLOATCONV(int32_t, float32_to_int32, float32, INT32_MIN)
755 WRAP_FLOATCONV(int32_t, float32_to_int32_round_to_zero, float32, INT32_MIN)
756 WRAP_FLOATCONV(int32_t, float64_to_int32, float64, INT32_MIN)
757 WRAP_FLOATCONV(int32_t, float64_to_int32_round_to_zero, float64, INT32_MIN)
758 WRAP_FLOATCONV(int64_t, float32_to_int64, float32, INT64_MIN)
759 WRAP_FLOATCONV(int64_t, float32_to_int64_round_to_zero, float32, INT64_MIN)
760 WRAP_FLOATCONV(int64_t, float64_to_int64, float64, INT64_MIN)
761 WRAP_FLOATCONV(int64_t, float64_to_int64_round_to_zero, float64, INT64_MIN)
762 
763 void helper_cvtps2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
764 {
765     d->ZMM_L(0) = x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
766     d->ZMM_L(1) = x86_float32_to_int32(s->ZMM_S(1), &env->sse_status);
767     d->ZMM_L(2) = x86_float32_to_int32(s->ZMM_S(2), &env->sse_status);
768     d->ZMM_L(3) = x86_float32_to_int32(s->ZMM_S(3), &env->sse_status);
769 }
770 
771 void helper_cvtpd2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
772 {
773     d->ZMM_L(0) = x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
774     d->ZMM_L(1) = x86_float64_to_int32(s->ZMM_D(1), &env->sse_status);
775     d->ZMM_Q(1) = 0;
776 }
777 
778 void helper_cvtps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
779 {
780     d->MMX_L(0) = x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
781     d->MMX_L(1) = x86_float32_to_int32(s->ZMM_S(1), &env->sse_status);
782 }
783 
784 void helper_cvtpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
785 {
786     d->MMX_L(0) = x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
787     d->MMX_L(1) = x86_float64_to_int32(s->ZMM_D(1), &env->sse_status);
788 }
789 
790 int32_t helper_cvtss2si(CPUX86State *env, ZMMReg *s)
791 {
792     return x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
793 }
794 
795 int32_t helper_cvtsd2si(CPUX86State *env, ZMMReg *s)
796 {
797     return x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
798 }
799 
800 #ifdef TARGET_X86_64
801 int64_t helper_cvtss2sq(CPUX86State *env, ZMMReg *s)
802 {
803     return x86_float32_to_int64(s->ZMM_S(0), &env->sse_status);
804 }
805 
806 int64_t helper_cvtsd2sq(CPUX86State *env, ZMMReg *s)
807 {
808     return x86_float64_to_int64(s->ZMM_D(0), &env->sse_status);
809 }
810 #endif
811 
812 /* float to integer truncated */
813 void helper_cvttps2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
814 {
815     d->ZMM_L(0) = x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
816     d->ZMM_L(1) = x86_float32_to_int32_round_to_zero(s->ZMM_S(1), &env->sse_status);
817     d->ZMM_L(2) = x86_float32_to_int32_round_to_zero(s->ZMM_S(2), &env->sse_status);
818     d->ZMM_L(3) = x86_float32_to_int32_round_to_zero(s->ZMM_S(3), &env->sse_status);
819 }
820 
821 void helper_cvttpd2dq(CPUX86State *env, ZMMReg *d, ZMMReg *s)
822 {
823     d->ZMM_L(0) = x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
824     d->ZMM_L(1) = x86_float64_to_int32_round_to_zero(s->ZMM_D(1), &env->sse_status);
825     d->ZMM_Q(1) = 0;
826 }
827 
828 void helper_cvttps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
829 {
830     d->MMX_L(0) = x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
831     d->MMX_L(1) = x86_float32_to_int32_round_to_zero(s->ZMM_S(1), &env->sse_status);
832 }
833 
834 void helper_cvttpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
835 {
836     d->MMX_L(0) = x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
837     d->MMX_L(1) = x86_float64_to_int32_round_to_zero(s->ZMM_D(1), &env->sse_status);
838 }
839 
840 int32_t helper_cvttss2si(CPUX86State *env, ZMMReg *s)
841 {
842     return x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
843 }
844 
845 int32_t helper_cvttsd2si(CPUX86State *env, ZMMReg *s)
846 {
847     return x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
848 }
849 
850 #ifdef TARGET_X86_64
851 int64_t helper_cvttss2sq(CPUX86State *env, ZMMReg *s)
852 {
853     return x86_float32_to_int64_round_to_zero(s->ZMM_S(0), &env->sse_status);
854 }
855 
856 int64_t helper_cvttsd2sq(CPUX86State *env, ZMMReg *s)
857 {
858     return x86_float64_to_int64_round_to_zero(s->ZMM_D(0), &env->sse_status);
859 }
860 #endif
861 
862 void helper_rsqrtps(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     d->ZMM_S(1) = float32_div(float32_one,
869                               float32_sqrt(s->ZMM_S(1), &env->sse_status),
870                               &env->sse_status);
871     d->ZMM_S(2) = float32_div(float32_one,
872                               float32_sqrt(s->ZMM_S(2), &env->sse_status),
873                               &env->sse_status);
874     d->ZMM_S(3) = float32_div(float32_one,
875                               float32_sqrt(s->ZMM_S(3), &env->sse_status),
876                               &env->sse_status);
877     set_float_exception_flags(old_flags, &env->sse_status);
878 }
879 
880 void helper_rsqrtss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
881 {
882     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
883     d->ZMM_S(0) = float32_div(float32_one,
884                               float32_sqrt(s->ZMM_S(0), &env->sse_status),
885                               &env->sse_status);
886     set_float_exception_flags(old_flags, &env->sse_status);
887 }
888 
889 void helper_rcpps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
890 {
891     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
892     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
893     d->ZMM_S(1) = float32_div(float32_one, s->ZMM_S(1), &env->sse_status);
894     d->ZMM_S(2) = float32_div(float32_one, s->ZMM_S(2), &env->sse_status);
895     d->ZMM_S(3) = float32_div(float32_one, s->ZMM_S(3), &env->sse_status);
896     set_float_exception_flags(old_flags, &env->sse_status);
897 }
898 
899 void helper_rcpss(CPUX86State *env, ZMMReg *d, ZMMReg *s)
900 {
901     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
902     d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
903     set_float_exception_flags(old_flags, &env->sse_status);
904 }
905 
906 static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
907 {
908     uint64_t mask;
909 
910     if (len == 0) {
911         mask = ~0LL;
912     } else {
913         mask = (1ULL << len) - 1;
914     }
915     return (src >> shift) & mask;
916 }
917 
918 void helper_extrq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
919 {
920     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), s->ZMM_B(1), s->ZMM_B(0));
921 }
922 
923 void helper_extrq_i(CPUX86State *env, ZMMReg *d, int index, int length)
924 {
925     d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), index, length);
926 }
927 
928 static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
929 {
930     uint64_t mask;
931 
932     if (len == 0) {
933         mask = ~0ULL;
934     } else {
935         mask = (1ULL << len) - 1;
936     }
937     return (src & ~(mask << shift)) | ((src & mask) << shift);
938 }
939 
940 void helper_insertq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
941 {
942     d->ZMM_Q(0) = helper_insertq(s->ZMM_Q(0), s->ZMM_B(9), s->ZMM_B(8));
943 }
944 
945 void helper_insertq_i(CPUX86State *env, ZMMReg *d, int index, int length)
946 {
947     d->ZMM_Q(0) = helper_insertq(d->ZMM_Q(0), index, length);
948 }
949 
950 void helper_haddps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
951 {
952     ZMMReg r;
953 
954     r.ZMM_S(0) = float32_add(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
955     r.ZMM_S(1) = float32_add(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
956     r.ZMM_S(2) = float32_add(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
957     r.ZMM_S(3) = float32_add(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
958     MOVE(*d, r);
959 }
960 
961 void helper_haddpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
962 {
963     ZMMReg r;
964 
965     r.ZMM_D(0) = float64_add(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
966     r.ZMM_D(1) = float64_add(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
967     MOVE(*d, r);
968 }
969 
970 void helper_hsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
971 {
972     ZMMReg r;
973 
974     r.ZMM_S(0) = float32_sub(d->ZMM_S(0), d->ZMM_S(1), &env->sse_status);
975     r.ZMM_S(1) = float32_sub(d->ZMM_S(2), d->ZMM_S(3), &env->sse_status);
976     r.ZMM_S(2) = float32_sub(s->ZMM_S(0), s->ZMM_S(1), &env->sse_status);
977     r.ZMM_S(3) = float32_sub(s->ZMM_S(2), s->ZMM_S(3), &env->sse_status);
978     MOVE(*d, r);
979 }
980 
981 void helper_hsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
982 {
983     ZMMReg r;
984 
985     r.ZMM_D(0) = float64_sub(d->ZMM_D(0), d->ZMM_D(1), &env->sse_status);
986     r.ZMM_D(1) = float64_sub(s->ZMM_D(0), s->ZMM_D(1), &env->sse_status);
987     MOVE(*d, r);
988 }
989 
990 void helper_addsubps(CPUX86State *env, ZMMReg *d, ZMMReg *s)
991 {
992     d->ZMM_S(0) = float32_sub(d->ZMM_S(0), s->ZMM_S(0), &env->sse_status);
993     d->ZMM_S(1) = float32_add(d->ZMM_S(1), s->ZMM_S(1), &env->sse_status);
994     d->ZMM_S(2) = float32_sub(d->ZMM_S(2), s->ZMM_S(2), &env->sse_status);
995     d->ZMM_S(3) = float32_add(d->ZMM_S(3), s->ZMM_S(3), &env->sse_status);
996 }
997 
998 void helper_addsubpd(CPUX86State *env, ZMMReg *d, ZMMReg *s)
999 {
1000     d->ZMM_D(0) = float64_sub(d->ZMM_D(0), s->ZMM_D(0), &env->sse_status);
1001     d->ZMM_D(1) = float64_add(d->ZMM_D(1), s->ZMM_D(1), &env->sse_status);
1002 }
1003 
1004 /* XXX: unordered */
1005 #define SSE_HELPER_CMP(name, F)                                         \
1006     void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s)        \
1007     {                                                                   \
1008         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
1009         d->ZMM_L(1) = F(32, d->ZMM_S(1), s->ZMM_S(1));                  \
1010         d->ZMM_L(2) = F(32, d->ZMM_S(2), s->ZMM_S(2));                  \
1011         d->ZMM_L(3) = F(32, d->ZMM_S(3), s->ZMM_S(3));                  \
1012     }                                                                   \
1013                                                                         \
1014     void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s)        \
1015     {                                                                   \
1016         d->ZMM_L(0) = F(32, d->ZMM_S(0), s->ZMM_S(0));                  \
1017     }                                                                   \
1018                                                                         \
1019     void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s)        \
1020     {                                                                   \
1021         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
1022         d->ZMM_Q(1) = F(64, d->ZMM_D(1), s->ZMM_D(1));                  \
1023     }                                                                   \
1024                                                                         \
1025     void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s)        \
1026     {                                                                   \
1027         d->ZMM_Q(0) = F(64, d->ZMM_D(0), s->ZMM_D(0));                  \
1028     }
1029 
1030 #define FPU_CMPEQ(size, a, b)                                           \
1031     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? -1 : 0)
1032 #define FPU_CMPLT(size, a, b)                                           \
1033     (float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0)
1034 #define FPU_CMPLE(size, a, b)                                           \
1035     (float ## size ## _le(a, b, &env->sse_status) ? -1 : 0)
1036 #define FPU_CMPUNORD(size, a, b)                                        \
1037     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? -1 : 0)
1038 #define FPU_CMPNEQ(size, a, b)                                          \
1039     (float ## size ## _eq_quiet(a, b, &env->sse_status) ? 0 : -1)
1040 #define FPU_CMPNLT(size, a, b)                                          \
1041     (float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1)
1042 #define FPU_CMPNLE(size, a, b)                                          \
1043     (float ## size ## _le(a, b, &env->sse_status) ? 0 : -1)
1044 #define FPU_CMPORD(size, a, b)                                          \
1045     (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? 0 : -1)
1046 
1047 SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
1048 SSE_HELPER_CMP(cmplt, FPU_CMPLT)
1049 SSE_HELPER_CMP(cmple, FPU_CMPLE)
1050 SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
1051 SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
1052 SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
1053 SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
1054 SSE_HELPER_CMP(cmpord, FPU_CMPORD)
1055 
1056 static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
1057 
1058 void helper_ucomiss(CPUX86State *env, Reg *d, Reg *s)
1059 {
1060     FloatRelation ret;
1061     float32 s0, s1;
1062 
1063     s0 = d->ZMM_S(0);
1064     s1 = s->ZMM_S(0);
1065     ret = float32_compare_quiet(s0, s1, &env->sse_status);
1066     CC_SRC = comis_eflags[ret + 1];
1067 }
1068 
1069 void helper_comiss(CPUX86State *env, Reg *d, Reg *s)
1070 {
1071     FloatRelation ret;
1072     float32 s0, s1;
1073 
1074     s0 = d->ZMM_S(0);
1075     s1 = s->ZMM_S(0);
1076     ret = float32_compare(s0, s1, &env->sse_status);
1077     CC_SRC = comis_eflags[ret + 1];
1078 }
1079 
1080 void helper_ucomisd(CPUX86State *env, Reg *d, Reg *s)
1081 {
1082     FloatRelation ret;
1083     float64 d0, d1;
1084 
1085     d0 = d->ZMM_D(0);
1086     d1 = s->ZMM_D(0);
1087     ret = float64_compare_quiet(d0, d1, &env->sse_status);
1088     CC_SRC = comis_eflags[ret + 1];
1089 }
1090 
1091 void helper_comisd(CPUX86State *env, Reg *d, Reg *s)
1092 {
1093     FloatRelation ret;
1094     float64 d0, d1;
1095 
1096     d0 = d->ZMM_D(0);
1097     d1 = s->ZMM_D(0);
1098     ret = float64_compare(d0, d1, &env->sse_status);
1099     CC_SRC = comis_eflags[ret + 1];
1100 }
1101 
1102 uint32_t helper_movmskps(CPUX86State *env, Reg *s)
1103 {
1104     int b0, b1, b2, b3;
1105 
1106     b0 = s->ZMM_L(0) >> 31;
1107     b1 = s->ZMM_L(1) >> 31;
1108     b2 = s->ZMM_L(2) >> 31;
1109     b3 = s->ZMM_L(3) >> 31;
1110     return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
1111 }
1112 
1113 uint32_t helper_movmskpd(CPUX86State *env, Reg *s)
1114 {
1115     int b0, b1;
1116 
1117     b0 = s->ZMM_L(1) >> 31;
1118     b1 = s->ZMM_L(3) >> 31;
1119     return b0 | (b1 << 1);
1120 }
1121 
1122 #endif
1123 
1124 uint32_t glue(helper_pmovmskb, SUFFIX)(CPUX86State *env, Reg *s)
1125 {
1126     uint32_t val;
1127 
1128     val = 0;
1129     val |= (s->B(0) >> 7);
1130     val |= (s->B(1) >> 6) & 0x02;
1131     val |= (s->B(2) >> 5) & 0x04;
1132     val |= (s->B(3) >> 4) & 0x08;
1133     val |= (s->B(4) >> 3) & 0x10;
1134     val |= (s->B(5) >> 2) & 0x20;
1135     val |= (s->B(6) >> 1) & 0x40;
1136     val |= (s->B(7)) & 0x80;
1137 #if SHIFT == 1
1138     val |= (s->B(8) << 1) & 0x0100;
1139     val |= (s->B(9) << 2) & 0x0200;
1140     val |= (s->B(10) << 3) & 0x0400;
1141     val |= (s->B(11) << 4) & 0x0800;
1142     val |= (s->B(12) << 5) & 0x1000;
1143     val |= (s->B(13) << 6) & 0x2000;
1144     val |= (s->B(14) << 7) & 0x4000;
1145     val |= (s->B(15) << 8) & 0x8000;
1146 #endif
1147     return val;
1148 }
1149 
1150 void glue(helper_packsswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1151 {
1152     Reg r;
1153 
1154     r.B(0) = satsb((int16_t)d->W(0));
1155     r.B(1) = satsb((int16_t)d->W(1));
1156     r.B(2) = satsb((int16_t)d->W(2));
1157     r.B(3) = satsb((int16_t)d->W(3));
1158 #if SHIFT == 1
1159     r.B(4) = satsb((int16_t)d->W(4));
1160     r.B(5) = satsb((int16_t)d->W(5));
1161     r.B(6) = satsb((int16_t)d->W(6));
1162     r.B(7) = satsb((int16_t)d->W(7));
1163 #endif
1164     r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1165     r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1166     r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1167     r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1168 #if SHIFT == 1
1169     r.B(12) = satsb((int16_t)s->W(4));
1170     r.B(13) = satsb((int16_t)s->W(5));
1171     r.B(14) = satsb((int16_t)s->W(6));
1172     r.B(15) = satsb((int16_t)s->W(7));
1173 #endif
1174     MOVE(*d, r);
1175 }
1176 
1177 void glue(helper_packuswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1178 {
1179     Reg r;
1180 
1181     r.B(0) = satub((int16_t)d->W(0));
1182     r.B(1) = satub((int16_t)d->W(1));
1183     r.B(2) = satub((int16_t)d->W(2));
1184     r.B(3) = satub((int16_t)d->W(3));
1185 #if SHIFT == 1
1186     r.B(4) = satub((int16_t)d->W(4));
1187     r.B(5) = satub((int16_t)d->W(5));
1188     r.B(6) = satub((int16_t)d->W(6));
1189     r.B(7) = satub((int16_t)d->W(7));
1190 #endif
1191     r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1192     r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1193     r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1194     r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1195 #if SHIFT == 1
1196     r.B(12) = satub((int16_t)s->W(4));
1197     r.B(13) = satub((int16_t)s->W(5));
1198     r.B(14) = satub((int16_t)s->W(6));
1199     r.B(15) = satub((int16_t)s->W(7));
1200 #endif
1201     MOVE(*d, r);
1202 }
1203 
1204 void glue(helper_packssdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1205 {
1206     Reg r;
1207 
1208     r.W(0) = satsw(d->L(0));
1209     r.W(1) = satsw(d->L(1));
1210 #if SHIFT == 1
1211     r.W(2) = satsw(d->L(2));
1212     r.W(3) = satsw(d->L(3));
1213 #endif
1214     r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1215     r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1216 #if SHIFT == 1
1217     r.W(6) = satsw(s->L(2));
1218     r.W(7) = satsw(s->L(3));
1219 #endif
1220     MOVE(*d, r);
1221 }
1222 
1223 #define UNPCK_OP(base_name, base)                                       \
1224                                                                         \
1225     void glue(helper_punpck ## base_name ## bw, SUFFIX)(CPUX86State *env,\
1226                                                         Reg *d, Reg *s) \
1227     {                                                                   \
1228         Reg r;                                                          \
1229                                                                         \
1230         r.B(0) = d->B((base << (SHIFT + 2)) + 0);                       \
1231         r.B(1) = s->B((base << (SHIFT + 2)) + 0);                       \
1232         r.B(2) = d->B((base << (SHIFT + 2)) + 1);                       \
1233         r.B(3) = s->B((base << (SHIFT + 2)) + 1);                       \
1234         r.B(4) = d->B((base << (SHIFT + 2)) + 2);                       \
1235         r.B(5) = s->B((base << (SHIFT + 2)) + 2);                       \
1236         r.B(6) = d->B((base << (SHIFT + 2)) + 3);                       \
1237         r.B(7) = s->B((base << (SHIFT + 2)) + 3);                       \
1238         XMM_ONLY(                                                       \
1239                  r.B(8) = d->B((base << (SHIFT + 2)) + 4);              \
1240                  r.B(9) = s->B((base << (SHIFT + 2)) + 4);              \
1241                  r.B(10) = d->B((base << (SHIFT + 2)) + 5);             \
1242                  r.B(11) = s->B((base << (SHIFT + 2)) + 5);             \
1243                  r.B(12) = d->B((base << (SHIFT + 2)) + 6);             \
1244                  r.B(13) = s->B((base << (SHIFT + 2)) + 6);             \
1245                  r.B(14) = d->B((base << (SHIFT + 2)) + 7);             \
1246                  r.B(15) = s->B((base << (SHIFT + 2)) + 7);             \
1247                                                                       ) \
1248         MOVE(*d, r);                                                    \
1249     }                                                                   \
1250                                                                         \
1251     void glue(helper_punpck ## base_name ## wd, SUFFIX)(CPUX86State *env,\
1252                                                         Reg *d, Reg *s) \
1253     {                                                                   \
1254         Reg r;                                                          \
1255                                                                         \
1256         r.W(0) = d->W((base << (SHIFT + 1)) + 0);                       \
1257         r.W(1) = s->W((base << (SHIFT + 1)) + 0);                       \
1258         r.W(2) = d->W((base << (SHIFT + 1)) + 1);                       \
1259         r.W(3) = s->W((base << (SHIFT + 1)) + 1);                       \
1260         XMM_ONLY(                                                       \
1261                  r.W(4) = d->W((base << (SHIFT + 1)) + 2);              \
1262                  r.W(5) = s->W((base << (SHIFT + 1)) + 2);              \
1263                  r.W(6) = d->W((base << (SHIFT + 1)) + 3);              \
1264                  r.W(7) = s->W((base << (SHIFT + 1)) + 3);              \
1265                                                                       ) \
1266             MOVE(*d, r);                                                \
1267     }                                                                   \
1268                                                                         \
1269     void glue(helper_punpck ## base_name ## dq, SUFFIX)(CPUX86State *env,\
1270                                                         Reg *d, Reg *s) \
1271     {                                                                   \
1272         Reg r;                                                          \
1273                                                                         \
1274         r.L(0) = d->L((base << SHIFT) + 0);                             \
1275         r.L(1) = s->L((base << SHIFT) + 0);                             \
1276         XMM_ONLY(                                                       \
1277                  r.L(2) = d->L((base << SHIFT) + 1);                    \
1278                  r.L(3) = s->L((base << SHIFT) + 1);                    \
1279                                                                       ) \
1280             MOVE(*d, r);                                                \
1281     }                                                                   \
1282                                                                         \
1283     XMM_ONLY(                                                           \
1284              void glue(helper_punpck ## base_name ## qdq, SUFFIX)(CPUX86State \
1285                                                                   *env, \
1286                                                                   Reg *d, \
1287                                                                   Reg *s) \
1288              {                                                          \
1289                  Reg r;                                                 \
1290                                                                         \
1291                  r.Q(0) = d->Q(base);                                   \
1292                  r.Q(1) = s->Q(base);                                   \
1293                  MOVE(*d, r);                                           \
1294              }                                                          \
1295                                                                         )
1296 
1297 UNPCK_OP(l, 0)
1298 UNPCK_OP(h, 1)
1299 
1300 /* 3DNow! float ops */
1301 #if SHIFT == 0
1302 void helper_pi2fd(CPUX86State *env, MMXReg *d, MMXReg *s)
1303 {
1304     d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1305     d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1306 }
1307 
1308 void helper_pi2fw(CPUX86State *env, MMXReg *d, MMXReg *s)
1309 {
1310     d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1311     d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1312 }
1313 
1314 void helper_pf2id(CPUX86State *env, MMXReg *d, MMXReg *s)
1315 {
1316     d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1317     d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1318 }
1319 
1320 void helper_pf2iw(CPUX86State *env, MMXReg *d, MMXReg *s)
1321 {
1322     d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0),
1323                                                        &env->mmx_status));
1324     d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1),
1325                                                        &env->mmx_status));
1326 }
1327 
1328 void helper_pfacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1329 {
1330     MMXReg r;
1331 
1332     r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1333     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1334     MOVE(*d, r);
1335 }
1336 
1337 void helper_pfadd(CPUX86State *env, MMXReg *d, MMXReg *s)
1338 {
1339     d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1340     d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1341 }
1342 
1343 void helper_pfcmpeq(CPUX86State *env, MMXReg *d, MMXReg *s)
1344 {
1345     d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0),
1346                                    &env->mmx_status) ? -1 : 0;
1347     d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1),
1348                                    &env->mmx_status) ? -1 : 0;
1349 }
1350 
1351 void helper_pfcmpge(CPUX86State *env, MMXReg *d, MMXReg *s)
1352 {
1353     d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0),
1354                              &env->mmx_status) ? -1 : 0;
1355     d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1),
1356                              &env->mmx_status) ? -1 : 0;
1357 }
1358 
1359 void helper_pfcmpgt(CPUX86State *env, MMXReg *d, MMXReg *s)
1360 {
1361     d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0),
1362                              &env->mmx_status) ? -1 : 0;
1363     d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1),
1364                              &env->mmx_status) ? -1 : 0;
1365 }
1366 
1367 void helper_pfmax(CPUX86State *env, MMXReg *d, MMXReg *s)
1368 {
1369     if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status)) {
1370         d->MMX_S(0) = s->MMX_S(0);
1371     }
1372     if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status)) {
1373         d->MMX_S(1) = s->MMX_S(1);
1374     }
1375 }
1376 
1377 void helper_pfmin(CPUX86State *env, MMXReg *d, MMXReg *s)
1378 {
1379     if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status)) {
1380         d->MMX_S(0) = s->MMX_S(0);
1381     }
1382     if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status)) {
1383         d->MMX_S(1) = s->MMX_S(1);
1384     }
1385 }
1386 
1387 void helper_pfmul(CPUX86State *env, MMXReg *d, MMXReg *s)
1388 {
1389     d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1390     d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1391 }
1392 
1393 void helper_pfnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1394 {
1395     MMXReg r;
1396 
1397     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1398     r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1399     MOVE(*d, r);
1400 }
1401 
1402 void helper_pfpnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1403 {
1404     MMXReg r;
1405 
1406     r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1407     r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1408     MOVE(*d, r);
1409 }
1410 
1411 void helper_pfrcp(CPUX86State *env, MMXReg *d, MMXReg *s)
1412 {
1413     d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
1414     d->MMX_S(1) = d->MMX_S(0);
1415 }
1416 
1417 void helper_pfrsqrt(CPUX86State *env, MMXReg *d, MMXReg *s)
1418 {
1419     d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1420     d->MMX_S(1) = float32_div(float32_one,
1421                               float32_sqrt(d->MMX_S(1), &env->mmx_status),
1422                               &env->mmx_status);
1423     d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1424     d->MMX_L(0) = d->MMX_L(1);
1425 }
1426 
1427 void helper_pfsub(CPUX86State *env, MMXReg *d, MMXReg *s)
1428 {
1429     d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1430     d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1431 }
1432 
1433 void helper_pfsubr(CPUX86State *env, MMXReg *d, MMXReg *s)
1434 {
1435     d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1436     d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1437 }
1438 
1439 void helper_pswapd(CPUX86State *env, MMXReg *d, MMXReg *s)
1440 {
1441     MMXReg r;
1442 
1443     r.MMX_L(0) = s->MMX_L(1);
1444     r.MMX_L(1) = s->MMX_L(0);
1445     MOVE(*d, r);
1446 }
1447 #endif
1448 
1449 /* SSSE3 op helpers */
1450 void glue(helper_pshufb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1451 {
1452     int i;
1453     Reg r;
1454 
1455     for (i = 0; i < (8 << SHIFT); i++) {
1456         r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1457     }
1458 
1459     MOVE(*d, r);
1460 }
1461 
1462 void glue(helper_phaddw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1463 {
1464 
1465     Reg r;
1466 
1467     r.W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1468     r.W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1469     XMM_ONLY(r.W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1470     XMM_ONLY(r.W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1471     r.W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1472     r.W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1473     XMM_ONLY(r.W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1474     XMM_ONLY(r.W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1475 
1476     MOVE(*d, r);
1477 }
1478 
1479 void glue(helper_phaddd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1480 {
1481     Reg r;
1482 
1483     r.L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1484     XMM_ONLY(r.L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1485     r.L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1486     XMM_ONLY(r.L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1487 
1488     MOVE(*d, r);
1489 }
1490 
1491 void glue(helper_phaddsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1492 {
1493     Reg r;
1494 
1495     r.W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1496     r.W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1497     XMM_ONLY(r.W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1498     XMM_ONLY(r.W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1499     r.W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1500     r.W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1501     XMM_ONLY(r.W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1502     XMM_ONLY(r.W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1503 
1504     MOVE(*d, r);
1505 }
1506 
1507 void glue(helper_pmaddubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1508 {
1509     d->W(0) = satsw((int8_t)s->B(0) * (uint8_t)d->B(0) +
1510                     (int8_t)s->B(1) * (uint8_t)d->B(1));
1511     d->W(1) = satsw((int8_t)s->B(2) * (uint8_t)d->B(2) +
1512                     (int8_t)s->B(3) * (uint8_t)d->B(3));
1513     d->W(2) = satsw((int8_t)s->B(4) * (uint8_t)d->B(4) +
1514                     (int8_t)s->B(5) * (uint8_t)d->B(5));
1515     d->W(3) = satsw((int8_t)s->B(6) * (uint8_t)d->B(6) +
1516                     (int8_t)s->B(7) * (uint8_t)d->B(7));
1517 #if SHIFT == 1
1518     d->W(4) = satsw((int8_t)s->B(8) * (uint8_t)d->B(8) +
1519                     (int8_t)s->B(9) * (uint8_t)d->B(9));
1520     d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1521                     (int8_t)s->B(11) * (uint8_t)d->B(11));
1522     d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1523                     (int8_t)s->B(13) * (uint8_t)d->B(13));
1524     d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1525                     (int8_t)s->B(15) * (uint8_t)d->B(15));
1526 #endif
1527 }
1528 
1529 void glue(helper_phsubw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1530 {
1531     d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1532     d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1533     XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1534     XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1535     d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1536     d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1537     XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1538     XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1539 }
1540 
1541 void glue(helper_phsubd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1542 {
1543     d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1544     XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1545     d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1546     XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1547 }
1548 
1549 void glue(helper_phsubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1550 {
1551     d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1552     d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1553     XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1554     XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1555     d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1556     d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1557     XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1558     XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1559 }
1560 
1561 #define FABSB(_, x) (x > INT8_MAX  ? -(int8_t)x : x)
1562 #define FABSW(_, x) (x > INT16_MAX ? -(int16_t)x : x)
1563 #define FABSL(_, x) (x > INT32_MAX ? -(int32_t)x : x)
1564 SSE_HELPER_B(helper_pabsb, FABSB)
1565 SSE_HELPER_W(helper_pabsw, FABSW)
1566 SSE_HELPER_L(helper_pabsd, FABSL)
1567 
1568 #define FMULHRSW(d, s) (((int16_t) d * (int16_t)s + 0x4000) >> 15)
1569 SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1570 
1571 #define FSIGNB(d, s) (s <= INT8_MAX  ? s ? d : 0 : -(int8_t)d)
1572 #define FSIGNW(d, s) (s <= INT16_MAX ? s ? d : 0 : -(int16_t)d)
1573 #define FSIGNL(d, s) (s <= INT32_MAX ? s ? d : 0 : -(int32_t)d)
1574 SSE_HELPER_B(helper_psignb, FSIGNB)
1575 SSE_HELPER_W(helper_psignw, FSIGNW)
1576 SSE_HELPER_L(helper_psignd, FSIGNL)
1577 
1578 void glue(helper_palignr, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1579                                   int32_t shift)
1580 {
1581     Reg r;
1582 
1583     /* XXX could be checked during translation */
1584     if (shift >= (16 << SHIFT)) {
1585         r.Q(0) = 0;
1586         XMM_ONLY(r.Q(1) = 0);
1587     } else {
1588         shift <<= 3;
1589 #define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1590 #if SHIFT == 0
1591         r.Q(0) = SHR(s->Q(0), shift - 0) |
1592             SHR(d->Q(0), shift -  64);
1593 #else
1594         r.Q(0) = SHR(s->Q(0), shift - 0) |
1595             SHR(s->Q(1), shift -  64) |
1596             SHR(d->Q(0), shift - 128) |
1597             SHR(d->Q(1), shift - 192);
1598         r.Q(1) = SHR(s->Q(0), shift + 64) |
1599             SHR(s->Q(1), shift -   0) |
1600             SHR(d->Q(0), shift -  64) |
1601             SHR(d->Q(1), shift - 128);
1602 #endif
1603 #undef SHR
1604     }
1605 
1606     MOVE(*d, r);
1607 }
1608 
1609 #define XMM0 (env->xmm_regs[0])
1610 
1611 #if SHIFT == 1
1612 #define SSE_HELPER_V(name, elem, num, F)                                \
1613     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)           \
1614     {                                                                   \
1615         d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0));           \
1616         d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1));           \
1617         if (num > 2) {                                                  \
1618             d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2));       \
1619             d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3));       \
1620             if (num > 4) {                                              \
1621                 d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4));   \
1622                 d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5));   \
1623                 d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6));   \
1624                 d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7));   \
1625                 if (num > 8) {                                          \
1626                     d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8)); \
1627                     d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9)); \
1628                     d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10)); \
1629                     d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11)); \
1630                     d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12)); \
1631                     d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13)); \
1632                     d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14)); \
1633                     d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15)); \
1634                 }                                                       \
1635             }                                                           \
1636         }                                                               \
1637     }
1638 
1639 #define SSE_HELPER_I(name, elem, num, F)                                \
1640     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t imm) \
1641     {                                                                   \
1642         d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1));       \
1643         d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1));       \
1644         if (num > 2) {                                                  \
1645             d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1));   \
1646             d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1));   \
1647             if (num > 4) {                                              \
1648                 d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1)); \
1649                 d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1)); \
1650                 d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1)); \
1651                 d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1)); \
1652                 if (num > 8) {                                          \
1653                     d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1)); \
1654                     d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1)); \
1655                     d->elem(10) = F(d->elem(10), s->elem(10),           \
1656                                     ((imm >> 10) & 1));                 \
1657                     d->elem(11) = F(d->elem(11), s->elem(11),           \
1658                                     ((imm >> 11) & 1));                 \
1659                     d->elem(12) = F(d->elem(12), s->elem(12),           \
1660                                     ((imm >> 12) & 1));                 \
1661                     d->elem(13) = F(d->elem(13), s->elem(13),           \
1662                                     ((imm >> 13) & 1));                 \
1663                     d->elem(14) = F(d->elem(14), s->elem(14),           \
1664                                     ((imm >> 14) & 1));                 \
1665                     d->elem(15) = F(d->elem(15), s->elem(15),           \
1666                                     ((imm >> 15) & 1));                 \
1667                 }                                                       \
1668             }                                                           \
1669         }                                                               \
1670     }
1671 
1672 /* SSE4.1 op helpers */
1673 #define FBLENDVB(d, s, m) ((m & 0x80) ? s : d)
1674 #define FBLENDVPS(d, s, m) ((m & 0x80000000) ? s : d)
1675 #define FBLENDVPD(d, s, m) ((m & 0x8000000000000000LL) ? s : d)
1676 SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1677 SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1678 SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1679 
1680 void glue(helper_ptest, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1681 {
1682     uint64_t zf = (s->Q(0) &  d->Q(0)) | (s->Q(1) &  d->Q(1));
1683     uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1684 
1685     CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1686 }
1687 
1688 #define SSE_HELPER_F(name, elem, num, F)        \
1689     void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)     \
1690     {                                           \
1691         if (num > 2) {                          \
1692             if (num > 4) {                      \
1693                 d->elem(7) = F(7);              \
1694                 d->elem(6) = F(6);              \
1695                 d->elem(5) = F(5);              \
1696                 d->elem(4) = F(4);              \
1697             }                                   \
1698             d->elem(3) = F(3);                  \
1699             d->elem(2) = F(2);                  \
1700         }                                       \
1701         d->elem(1) = F(1);                      \
1702         d->elem(0) = F(0);                      \
1703     }
1704 
1705 SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1706 SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1707 SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1708 SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1709 SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1710 SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1711 SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1712 SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1713 SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1714 SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1715 SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1716 SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1717 
1718 void glue(helper_pmuldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1719 {
1720     d->Q(0) = (int64_t)(int32_t) d->L(0) * (int32_t) s->L(0);
1721     d->Q(1) = (int64_t)(int32_t) d->L(2) * (int32_t) s->L(2);
1722 }
1723 
1724 #define FCMPEQQ(d, s) (d == s ? -1 : 0)
1725 SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1726 
1727 void glue(helper_packusdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1728 {
1729     Reg r;
1730 
1731     r.W(0) = satuw((int32_t) d->L(0));
1732     r.W(1) = satuw((int32_t) d->L(1));
1733     r.W(2) = satuw((int32_t) d->L(2));
1734     r.W(3) = satuw((int32_t) d->L(3));
1735     r.W(4) = satuw((int32_t) s->L(0));
1736     r.W(5) = satuw((int32_t) s->L(1));
1737     r.W(6) = satuw((int32_t) s->L(2));
1738     r.W(7) = satuw((int32_t) s->L(3));
1739     MOVE(*d, r);
1740 }
1741 
1742 #define FMINSB(d, s) MIN((int8_t)d, (int8_t)s)
1743 #define FMINSD(d, s) MIN((int32_t)d, (int32_t)s)
1744 #define FMAXSB(d, s) MAX((int8_t)d, (int8_t)s)
1745 #define FMAXSD(d, s) MAX((int32_t)d, (int32_t)s)
1746 SSE_HELPER_B(helper_pminsb, FMINSB)
1747 SSE_HELPER_L(helper_pminsd, FMINSD)
1748 SSE_HELPER_W(helper_pminuw, MIN)
1749 SSE_HELPER_L(helper_pminud, MIN)
1750 SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1751 SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1752 SSE_HELPER_W(helper_pmaxuw, MAX)
1753 SSE_HELPER_L(helper_pmaxud, MAX)
1754 
1755 #define FMULLD(d, s) ((int32_t)d * (int32_t)s)
1756 SSE_HELPER_L(helper_pmulld, FMULLD)
1757 
1758 void glue(helper_phminposuw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1759 {
1760     int idx = 0;
1761 
1762     if (s->W(1) < s->W(idx)) {
1763         idx = 1;
1764     }
1765     if (s->W(2) < s->W(idx)) {
1766         idx = 2;
1767     }
1768     if (s->W(3) < s->W(idx)) {
1769         idx = 3;
1770     }
1771     if (s->W(4) < s->W(idx)) {
1772         idx = 4;
1773     }
1774     if (s->W(5) < s->W(idx)) {
1775         idx = 5;
1776     }
1777     if (s->W(6) < s->W(idx)) {
1778         idx = 6;
1779     }
1780     if (s->W(7) < s->W(idx)) {
1781         idx = 7;
1782     }
1783 
1784     d->W(0) = s->W(idx);
1785     d->W(1) = idx;
1786     d->L(1) = 0;
1787     d->Q(1) = 0;
1788 }
1789 
1790 void glue(helper_roundps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1791                                   uint32_t mode)
1792 {
1793     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1794     signed char prev_rounding_mode;
1795 
1796     prev_rounding_mode = env->sse_status.float_rounding_mode;
1797     if (!(mode & (1 << 2))) {
1798         switch (mode & 3) {
1799         case 0:
1800             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1801             break;
1802         case 1:
1803             set_float_rounding_mode(float_round_down, &env->sse_status);
1804             break;
1805         case 2:
1806             set_float_rounding_mode(float_round_up, &env->sse_status);
1807             break;
1808         case 3:
1809             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1810             break;
1811         }
1812     }
1813 
1814     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1815     d->ZMM_S(1) = float32_round_to_int(s->ZMM_S(1), &env->sse_status);
1816     d->ZMM_S(2) = float32_round_to_int(s->ZMM_S(2), &env->sse_status);
1817     d->ZMM_S(3) = float32_round_to_int(s->ZMM_S(3), &env->sse_status);
1818 
1819     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1820         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1821                                   ~float_flag_inexact,
1822                                   &env->sse_status);
1823     }
1824     env->sse_status.float_rounding_mode = prev_rounding_mode;
1825 }
1826 
1827 void glue(helper_roundpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1828                                   uint32_t mode)
1829 {
1830     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1831     signed char prev_rounding_mode;
1832 
1833     prev_rounding_mode = env->sse_status.float_rounding_mode;
1834     if (!(mode & (1 << 2))) {
1835         switch (mode & 3) {
1836         case 0:
1837             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1838             break;
1839         case 1:
1840             set_float_rounding_mode(float_round_down, &env->sse_status);
1841             break;
1842         case 2:
1843             set_float_rounding_mode(float_round_up, &env->sse_status);
1844             break;
1845         case 3:
1846             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1847             break;
1848         }
1849     }
1850 
1851     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1852     d->ZMM_D(1) = float64_round_to_int(s->ZMM_D(1), &env->sse_status);
1853 
1854     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1855         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1856                                   ~float_flag_inexact,
1857                                   &env->sse_status);
1858     }
1859     env->sse_status.float_rounding_mode = prev_rounding_mode;
1860 }
1861 
1862 void glue(helper_roundss, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1863                                   uint32_t mode)
1864 {
1865     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1866     signed char prev_rounding_mode;
1867 
1868     prev_rounding_mode = env->sse_status.float_rounding_mode;
1869     if (!(mode & (1 << 2))) {
1870         switch (mode & 3) {
1871         case 0:
1872             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1873             break;
1874         case 1:
1875             set_float_rounding_mode(float_round_down, &env->sse_status);
1876             break;
1877         case 2:
1878             set_float_rounding_mode(float_round_up, &env->sse_status);
1879             break;
1880         case 3:
1881             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1882             break;
1883         }
1884     }
1885 
1886     d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
1887 
1888     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1889         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1890                                   ~float_flag_inexact,
1891                                   &env->sse_status);
1892     }
1893     env->sse_status.float_rounding_mode = prev_rounding_mode;
1894 }
1895 
1896 void glue(helper_roundsd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1897                                   uint32_t mode)
1898 {
1899     uint8_t old_flags = get_float_exception_flags(&env->sse_status);
1900     signed char prev_rounding_mode;
1901 
1902     prev_rounding_mode = env->sse_status.float_rounding_mode;
1903     if (!(mode & (1 << 2))) {
1904         switch (mode & 3) {
1905         case 0:
1906             set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1907             break;
1908         case 1:
1909             set_float_rounding_mode(float_round_down, &env->sse_status);
1910             break;
1911         case 2:
1912             set_float_rounding_mode(float_round_up, &env->sse_status);
1913             break;
1914         case 3:
1915             set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1916             break;
1917         }
1918     }
1919 
1920     d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
1921 
1922     if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
1923         set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1924                                   ~float_flag_inexact,
1925                                   &env->sse_status);
1926     }
1927     env->sse_status.float_rounding_mode = prev_rounding_mode;
1928 }
1929 
1930 #define FBLENDP(d, s, m) (m ? s : d)
1931 SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1932 SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1933 SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1934 
1935 void glue(helper_dpps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1936 {
1937     float32 iresult = float32_zero;
1938 
1939     if (mask & (1 << 4)) {
1940         iresult = float32_add(iresult,
1941                               float32_mul(d->ZMM_S(0), s->ZMM_S(0),
1942                                           &env->sse_status),
1943                               &env->sse_status);
1944     }
1945     if (mask & (1 << 5)) {
1946         iresult = float32_add(iresult,
1947                               float32_mul(d->ZMM_S(1), s->ZMM_S(1),
1948                                           &env->sse_status),
1949                               &env->sse_status);
1950     }
1951     if (mask & (1 << 6)) {
1952         iresult = float32_add(iresult,
1953                               float32_mul(d->ZMM_S(2), s->ZMM_S(2),
1954                                           &env->sse_status),
1955                               &env->sse_status);
1956     }
1957     if (mask & (1 << 7)) {
1958         iresult = float32_add(iresult,
1959                               float32_mul(d->ZMM_S(3), s->ZMM_S(3),
1960                                           &env->sse_status),
1961                               &env->sse_status);
1962     }
1963     d->ZMM_S(0) = (mask & (1 << 0)) ? iresult : float32_zero;
1964     d->ZMM_S(1) = (mask & (1 << 1)) ? iresult : float32_zero;
1965     d->ZMM_S(2) = (mask & (1 << 2)) ? iresult : float32_zero;
1966     d->ZMM_S(3) = (mask & (1 << 3)) ? iresult : float32_zero;
1967 }
1968 
1969 void glue(helper_dppd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1970 {
1971     float64 iresult = float64_zero;
1972 
1973     if (mask & (1 << 4)) {
1974         iresult = float64_add(iresult,
1975                               float64_mul(d->ZMM_D(0), s->ZMM_D(0),
1976                                           &env->sse_status),
1977                               &env->sse_status);
1978     }
1979     if (mask & (1 << 5)) {
1980         iresult = float64_add(iresult,
1981                               float64_mul(d->ZMM_D(1), s->ZMM_D(1),
1982                                           &env->sse_status),
1983                               &env->sse_status);
1984     }
1985     d->ZMM_D(0) = (mask & (1 << 0)) ? iresult : float64_zero;
1986     d->ZMM_D(1) = (mask & (1 << 1)) ? iresult : float64_zero;
1987 }
1988 
1989 void glue(helper_mpsadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1990                                   uint32_t offset)
1991 {
1992     int s0 = (offset & 3) << 2;
1993     int d0 = (offset & 4) << 0;
1994     int i;
1995     Reg r;
1996 
1997     for (i = 0; i < 8; i++, d0++) {
1998         r.W(i) = 0;
1999         r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
2000         r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
2001         r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
2002         r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
2003     }
2004 
2005     MOVE(*d, r);
2006 }
2007 
2008 /* SSE4.2 op helpers */
2009 #define FCMPGTQ(d, s) ((int64_t)d > (int64_t)s ? -1 : 0)
2010 SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
2011 
2012 static inline int pcmp_elen(CPUX86State *env, int reg, uint32_t ctrl)
2013 {
2014     int val;
2015 
2016     /* Presence of REX.W is indicated by a bit higher than 7 set */
2017     if (ctrl >> 8) {
2018         val = abs1((int64_t)env->regs[reg]);
2019     } else {
2020         val = abs1((int32_t)env->regs[reg]);
2021     }
2022 
2023     if (ctrl & 1) {
2024         if (val > 8) {
2025             return 8;
2026         }
2027     } else {
2028         if (val > 16) {
2029             return 16;
2030         }
2031     }
2032     return val;
2033 }
2034 
2035 static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
2036 {
2037     int val = 0;
2038 
2039     if (ctrl & 1) {
2040         while (val < 8 && r->W(val)) {
2041             val++;
2042         }
2043     } else {
2044         while (val < 16 && r->B(val)) {
2045             val++;
2046         }
2047     }
2048 
2049     return val;
2050 }
2051 
2052 static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
2053 {
2054     switch ((ctrl >> 0) & 3) {
2055     case 0:
2056         return r->B(i);
2057     case 1:
2058         return r->W(i);
2059     case 2:
2060         return (int8_t)r->B(i);
2061     case 3:
2062     default:
2063         return (int16_t)r->W(i);
2064     }
2065 }
2066 
2067 static inline unsigned pcmpxstrx(CPUX86State *env, Reg *d, Reg *s,
2068                                  int8_t ctrl, int valids, int validd)
2069 {
2070     unsigned int res = 0;
2071     int v;
2072     int j, i;
2073     int upper = (ctrl & 1) ? 7 : 15;
2074 
2075     valids--;
2076     validd--;
2077 
2078     CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
2079 
2080     switch ((ctrl >> 2) & 3) {
2081     case 0:
2082         for (j = valids; j >= 0; j--) {
2083             res <<= 1;
2084             v = pcmp_val(s, ctrl, j);
2085             for (i = validd; i >= 0; i--) {
2086                 res |= (v == pcmp_val(d, ctrl, i));
2087             }
2088         }
2089         break;
2090     case 1:
2091         for (j = valids; j >= 0; j--) {
2092             res <<= 1;
2093             v = pcmp_val(s, ctrl, j);
2094             for (i = ((validd - 1) | 1); i >= 0; i -= 2) {
2095                 res |= (pcmp_val(d, ctrl, i - 0) >= v &&
2096                         pcmp_val(d, ctrl, i - 1) <= v);
2097             }
2098         }
2099         break;
2100     case 2:
2101         res = (1 << (upper - MAX(valids, validd))) - 1;
2102         res <<= MAX(valids, validd) - MIN(valids, validd);
2103         for (i = MIN(valids, validd); i >= 0; i--) {
2104             res <<= 1;
2105             v = pcmp_val(s, ctrl, i);
2106             res |= (v == pcmp_val(d, ctrl, i));
2107         }
2108         break;
2109     case 3:
2110         if (validd == -1) {
2111             res = (2 << upper) - 1;
2112             break;
2113         }
2114         for (j = valids == upper ? valids : valids - validd; j >= 0; j--) {
2115             res <<= 1;
2116             v = 1;
2117             for (i = MIN(valids - j, validd); i >= 0; i--) {
2118                 v &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
2119             }
2120             res |= v;
2121         }
2122         break;
2123     }
2124 
2125     switch ((ctrl >> 4) & 3) {
2126     case 1:
2127         res ^= (2 << upper) - 1;
2128         break;
2129     case 3:
2130         res ^= (1 << (valids + 1)) - 1;
2131         break;
2132     }
2133 
2134     if (res) {
2135         CC_SRC |= CC_C;
2136     }
2137     if (res & 1) {
2138         CC_SRC |= CC_O;
2139     }
2140 
2141     return res;
2142 }
2143 
2144 void glue(helper_pcmpestri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2145                                     uint32_t ctrl)
2146 {
2147     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2148                                  pcmp_elen(env, R_EDX, ctrl),
2149                                  pcmp_elen(env, R_EAX, ctrl));
2150 
2151     if (res) {
2152         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2153     } else {
2154         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2155     }
2156 }
2157 
2158 void glue(helper_pcmpestrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2159                                     uint32_t ctrl)
2160 {
2161     int i;
2162     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2163                                  pcmp_elen(env, R_EDX, ctrl),
2164                                  pcmp_elen(env, R_EAX, ctrl));
2165 
2166     if ((ctrl >> 6) & 1) {
2167         if (ctrl & 1) {
2168             for (i = 0; i < 8; i++, res >>= 1) {
2169                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2170             }
2171         } else {
2172             for (i = 0; i < 16; i++, res >>= 1) {
2173                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2174             }
2175         }
2176     } else {
2177         env->xmm_regs[0].Q(1) = 0;
2178         env->xmm_regs[0].Q(0) = res;
2179     }
2180 }
2181 
2182 void glue(helper_pcmpistri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2183                                     uint32_t ctrl)
2184 {
2185     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2186                                  pcmp_ilen(s, ctrl),
2187                                  pcmp_ilen(d, ctrl));
2188 
2189     if (res) {
2190         env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2191     } else {
2192         env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2193     }
2194 }
2195 
2196 void glue(helper_pcmpistrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2197                                     uint32_t ctrl)
2198 {
2199     int i;
2200     unsigned int res = pcmpxstrx(env, d, s, ctrl,
2201                                  pcmp_ilen(s, ctrl),
2202                                  pcmp_ilen(d, ctrl));
2203 
2204     if ((ctrl >> 6) & 1) {
2205         if (ctrl & 1) {
2206             for (i = 0; i < 8; i++, res >>= 1) {
2207                 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2208             }
2209         } else {
2210             for (i = 0; i < 16; i++, res >>= 1) {
2211                 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2212             }
2213         }
2214     } else {
2215         env->xmm_regs[0].Q(1) = 0;
2216         env->xmm_regs[0].Q(0) = res;
2217     }
2218 }
2219 
2220 #define CRCPOLY        0x1edc6f41
2221 #define CRCPOLY_BITREV 0x82f63b78
2222 target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2223 {
2224     target_ulong crc = (msg & ((target_ulong) -1 >>
2225                                (TARGET_LONG_BITS - len))) ^ crc1;
2226 
2227     while (len--) {
2228         crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2229     }
2230 
2231     return crc;
2232 }
2233 
2234 void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2235                                     uint32_t ctrl)
2236 {
2237     uint64_t ah, al, b, resh, resl;
2238 
2239     ah = 0;
2240     al = d->Q((ctrl & 1) != 0);
2241     b = s->Q((ctrl & 16) != 0);
2242     resh = resl = 0;
2243 
2244     while (b) {
2245         if (b & 1) {
2246             resl ^= al;
2247             resh ^= ah;
2248         }
2249         ah = (ah << 1) | (al >> 63);
2250         al <<= 1;
2251         b >>= 1;
2252     }
2253 
2254     d->Q(0) = resl;
2255     d->Q(1) = resh;
2256 }
2257 
2258 void glue(helper_aesdec, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2259 {
2260     int i;
2261     Reg st = *d;
2262     Reg rk = *s;
2263 
2264     for (i = 0 ; i < 4 ; i++) {
2265         d->L(i) = rk.L(i) ^ bswap32(AES_Td0[st.B(AES_ishifts[4*i+0])] ^
2266                                     AES_Td1[st.B(AES_ishifts[4*i+1])] ^
2267                                     AES_Td2[st.B(AES_ishifts[4*i+2])] ^
2268                                     AES_Td3[st.B(AES_ishifts[4*i+3])]);
2269     }
2270 }
2271 
2272 void glue(helper_aesdeclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2273 {
2274     int i;
2275     Reg st = *d;
2276     Reg rk = *s;
2277 
2278     for (i = 0; i < 16; i++) {
2279         d->B(i) = rk.B(i) ^ (AES_isbox[st.B(AES_ishifts[i])]);
2280     }
2281 }
2282 
2283 void glue(helper_aesenc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2284 {
2285     int i;
2286     Reg st = *d;
2287     Reg rk = *s;
2288 
2289     for (i = 0 ; i < 4 ; i++) {
2290         d->L(i) = rk.L(i) ^ bswap32(AES_Te0[st.B(AES_shifts[4*i+0])] ^
2291                                     AES_Te1[st.B(AES_shifts[4*i+1])] ^
2292                                     AES_Te2[st.B(AES_shifts[4*i+2])] ^
2293                                     AES_Te3[st.B(AES_shifts[4*i+3])]);
2294     }
2295 }
2296 
2297 void glue(helper_aesenclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2298 {
2299     int i;
2300     Reg st = *d;
2301     Reg rk = *s;
2302 
2303     for (i = 0; i < 16; i++) {
2304         d->B(i) = rk.B(i) ^ (AES_sbox[st.B(AES_shifts[i])]);
2305     }
2306 
2307 }
2308 
2309 void glue(helper_aesimc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2310 {
2311     int i;
2312     Reg tmp = *s;
2313 
2314     for (i = 0 ; i < 4 ; i++) {
2315         d->L(i) = bswap32(AES_imc[tmp.B(4*i+0)][0] ^
2316                           AES_imc[tmp.B(4*i+1)][1] ^
2317                           AES_imc[tmp.B(4*i+2)][2] ^
2318                           AES_imc[tmp.B(4*i+3)][3]);
2319     }
2320 }
2321 
2322 void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2323                                           uint32_t ctrl)
2324 {
2325     int i;
2326     Reg tmp = *s;
2327 
2328     for (i = 0 ; i < 4 ; i++) {
2329         d->B(i) = AES_sbox[tmp.B(i + 4)];
2330         d->B(i + 8) = AES_sbox[tmp.B(i + 12)];
2331     }
2332     d->L(1) = (d->L(0) << 24 | d->L(0) >> 8) ^ ctrl;
2333     d->L(3) = (d->L(2) << 24 | d->L(2) >> 8) ^ ctrl;
2334 }
2335 #endif
2336 
2337 #undef SHIFT
2338 #undef XMM_ONLY
2339 #undef Reg
2340 #undef B
2341 #undef W
2342 #undef L
2343 #undef Q
2344 #undef SUFFIX
2345 #undef SIZE
2346