?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
block: add basic backup support to block driver
[qemu-kvm.git] / target-i386 / ops_sse.h
blobeb24b5f721102cf28918feb6f2e54b5cc6a3c7a2
1 /*
2 * MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4/PNI support
3 *
4 * Copyright (c) 2005 Fabrice Bellard
5 * Copyright (c) 2008 Intel Corporation <andrew.zaborowski@intel.com>
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
11 *
12 * This library is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
19 */
20 #if SHIFT == 0
21 #define Reg MMXReg
22 #define XMM_ONLY(...)
23 #define B(n) MMX_B(n)
24 #define W(n) MMX_W(n)
25 #define L(n) MMX_L(n)
26 #define Q(n) q
27 #define SUFFIX _mmx
28 #else
29 #define Reg XMMReg
30 #define XMM_ONLY(...) __VA_ARGS__
31 #define B(n) XMM_B(n)
32 #define W(n) XMM_W(n)
33 #define L(n) XMM_L(n)
34 #define Q(n) XMM_Q(n)
35 #define SUFFIX _xmm
36 #endif
37
38 void glue(helper_psrlw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
39 {
40 int shift;
41
42 if (s->Q(0) > 15) {
43 d->Q(0) = 0;
44 #if SHIFT == 1
45 d->Q(1) = 0;
46 #endif
47 } else {
48 shift = s->B(0);
49 d->W(0) >>= shift;
50 d->W(1) >>= shift;
51 d->W(2) >>= shift;
52 d->W(3) >>= shift;
53 #if SHIFT == 1
54 d->W(4) >>= shift;
55 d->W(5) >>= shift;
56 d->W(6) >>= shift;
57 d->W(7) >>= shift;
58 #endif
59 }
60 }
61
62 void glue(helper_psraw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
63 {
64 int shift;
65
66 if (s->Q(0) > 15) {
67 shift = 15;
68 } else {
69 shift = s->B(0);
70 }
71 d->W(0) = (int16_t)d->W(0) >> shift;
72 d->W(1) = (int16_t)d->W(1) >> shift;
73 d->W(2) = (int16_t)d->W(2) >> shift;
74 d->W(3) = (int16_t)d->W(3) >> shift;
75 #if SHIFT == 1
76 d->W(4) = (int16_t)d->W(4) >> shift;
77 d->W(5) = (int16_t)d->W(5) >> shift;
78 d->W(6) = (int16_t)d->W(6) >> shift;
79 d->W(7) = (int16_t)d->W(7) >> shift;
80 #endif
81 }
82
83 void glue(helper_psllw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
84 {
85 int shift;
86
87 if (s->Q(0) > 15) {
88 d->Q(0) = 0;
89 #if SHIFT == 1
90 d->Q(1) = 0;
91 #endif
92 } else {
93 shift = s->B(0);
94 d->W(0) <<= shift;
95 d->W(1) <<= shift;
96 d->W(2) <<= shift;
97 d->W(3) <<= shift;
98 #if SHIFT == 1
99 d->W(4) <<= shift;
100 d->W(5) <<= shift;
101 d->W(6) <<= shift;
102 d->W(7) <<= shift;
103 #endif
104 }
105 }
106
107 void glue(helper_psrld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
108 {
109 int shift;
110
111 if (s->Q(0) > 31) {
112 d->Q(0) = 0;
113 #if SHIFT == 1
114 d->Q(1) = 0;
115 #endif
116 } else {
117 shift = s->B(0);
118 d->L(0) >>= shift;
119 d->L(1) >>= shift;
120 #if SHIFT == 1
121 d->L(2) >>= shift;
122 d->L(3) >>= shift;
123 #endif
124 }
125 }
126
127 void glue(helper_psrad, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
128 {
129 int shift;
130
131 if (s->Q(0) > 31) {
132 shift = 31;
133 } else {
134 shift = s->B(0);
135 }
136 d->L(0) = (int32_t)d->L(0) >> shift;
137 d->L(1) = (int32_t)d->L(1) >> shift;
138 #if SHIFT == 1
139 d->L(2) = (int32_t)d->L(2) >> shift;
140 d->L(3) = (int32_t)d->L(3) >> shift;
141 #endif
142 }
143
144 void glue(helper_pslld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
145 {
146 int shift;
147
148 if (s->Q(0) > 31) {
149 d->Q(0) = 0;
150 #if SHIFT == 1
151 d->Q(1) = 0;
152 #endif
153 } else {
154 shift = s->B(0);
155 d->L(0) <<= shift;
156 d->L(1) <<= shift;
157 #if SHIFT == 1
158 d->L(2) <<= shift;
159 d->L(3) <<= shift;
160 #endif
161 }
162 }
163
164 void glue(helper_psrlq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
165 {
166 int shift;
167
168 if (s->Q(0) > 63) {
169 d->Q(0) = 0;
170 #if SHIFT == 1
171 d->Q(1) = 0;
172 #endif
173 } else {
174 shift = s->B(0);
175 d->Q(0) >>= shift;
176 #if SHIFT == 1
177 d->Q(1) >>= shift;
178 #endif
179 }
180 }
181
182 void glue(helper_psllq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
183 {
184 int shift;
185
186 if (s->Q(0) > 63) {
187 d->Q(0) = 0;
188 #if SHIFT == 1
189 d->Q(1) = 0;
190 #endif
191 } else {
192 shift = s->B(0);
193 d->Q(0) <<= shift;
194 #if SHIFT == 1
195 d->Q(1) <<= shift;
196 #endif
197 }
198 }
199
200 #if SHIFT == 1
201 void glue(helper_psrldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
202 {
203 int shift, i;
204
205 shift = s->L(0);
206 if (shift > 16) {
207 shift = 16;
208 }
209 for (i = 0; i < 16 - shift; i++) {
210 d->B(i) = d->B(i + shift);
211 }
212 for (i = 16 - shift; i < 16; i++) {
213 d->B(i) = 0;
214 }
215 }
216
217 void glue(helper_pslldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
218 {
219 int shift, i;
220
221 shift = s->L(0);
222 if (shift > 16) {
223 shift = 16;
224 }
225 for (i = 15; i >= shift; i--) {
226 d->B(i) = d->B(i - shift);
227 }
228 for (i = 0; i < shift; i++) {
229 d->B(i) = 0;
230 }
231 }
232 #endif
233
234 #define SSE_HELPER_B(name, F) \
235 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
236 { \
237 d->B(0) = F(d->B(0), s->B(0)); \
238 d->B(1) = F(d->B(1), s->B(1)); \
239 d->B(2) = F(d->B(2), s->B(2)); \
240 d->B(3) = F(d->B(3), s->B(3)); \
241 d->B(4) = F(d->B(4), s->B(4)); \
242 d->B(5) = F(d->B(5), s->B(5)); \
243 d->B(6) = F(d->B(6), s->B(6)); \
244 d->B(7) = F(d->B(7), s->B(7)); \
245 XMM_ONLY( \
246 d->B(8) = F(d->B(8), s->B(8)); \
247 d->B(9) = F(d->B(9), s->B(9)); \
248 d->B(10) = F(d->B(10), s->B(10)); \
249 d->B(11) = F(d->B(11), s->B(11)); \
250 d->B(12) = F(d->B(12), s->B(12)); \
251 d->B(13) = F(d->B(13), s->B(13)); \
252 d->B(14) = F(d->B(14), s->B(14)); \
253 d->B(15) = F(d->B(15), s->B(15)); \
254 ) \
255 }
256
257 #define SSE_HELPER_W(name, F) \
258 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
259 { \
260 d->W(0) = F(d->W(0), s->W(0)); \
261 d->W(1) = F(d->W(1), s->W(1)); \
262 d->W(2) = F(d->W(2), s->W(2)); \
263 d->W(3) = F(d->W(3), s->W(3)); \
264 XMM_ONLY( \
265 d->W(4) = F(d->W(4), s->W(4)); \
266 d->W(5) = F(d->W(5), s->W(5)); \
267 d->W(6) = F(d->W(6), s->W(6)); \
268 d->W(7) = F(d->W(7), s->W(7)); \
269 ) \
270 }
271
272 #define SSE_HELPER_L(name, F) \
273 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
274 { \
275 d->L(0) = F(d->L(0), s->L(0)); \
276 d->L(1) = F(d->L(1), s->L(1)); \
277 XMM_ONLY( \
278 d->L(2) = F(d->L(2), s->L(2)); \
279 d->L(3) = F(d->L(3), s->L(3)); \
280 ) \
281 }
282
283 #define SSE_HELPER_Q(name, F) \
284 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
285 { \
286 d->Q(0) = F(d->Q(0), s->Q(0)); \
287 XMM_ONLY( \
288 d->Q(1) = F(d->Q(1), s->Q(1)); \
289 ) \
290 }
291
292 #if SHIFT == 0
293 static inline int satub(int x)
294 {
295 if (x < 0) {
296 return 0;
297 } else if (x > 255) {
298 return 255;
299 } else {
300 return x;
301 }
302 }
303
304 static inline int satuw(int x)
305 {
306 if (x < 0) {
307 return 0;
308 } else if (x > 65535) {
309 return 65535;
310 } else {
311 return x;
312 }
313 }
314
315 static inline int satsb(int x)
316 {
317 if (x < -128) {
318 return -128;
319 } else if (x > 127) {
320 return 127;
321 } else {
322 return x;
323 }
324 }
325
326 static inline int satsw(int x)
327 {
328 if (x < -32768) {
329 return -32768;
330 } else if (x > 32767) {
331 return 32767;
332 } else {
333 return x;
334 }
335 }
336
337 #define FADD(a, b) ((a) + (b))
338 #define FADDUB(a, b) satub((a) + (b))
339 #define FADDUW(a, b) satuw((a) + (b))
340 #define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
341 #define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
342
343 #define FSUB(a, b) ((a) - (b))
344 #define FSUBUB(a, b) satub((a) - (b))
345 #define FSUBUW(a, b) satuw((a) - (b))
346 #define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
347 #define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
348 #define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
349 #define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
350 #define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
351 #define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
352
353 #define FAND(a, b) ((a) & (b))
354 #define FANDN(a, b) ((~(a)) & (b))
355 #define FOR(a, b) ((a) | (b))
356 #define FXOR(a, b) ((a) ^ (b))
357
358 #define FCMPGTB(a, b) ((int8_t)(a) > (int8_t)(b) ? -1 : 0)
359 #define FCMPGTW(a, b) ((int16_t)(a) > (int16_t)(b) ? -1 : 0)
360 #define FCMPGTL(a, b) ((int32_t)(a) > (int32_t)(b) ? -1 : 0)
361 #define FCMPEQ(a, b) ((a) == (b) ? -1 : 0)
362
363 #define FMULLW(a, b) ((a) * (b))
364 #define FMULHRW(a, b) (((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16)
365 #define FMULHUW(a, b) ((a) * (b) >> 16)
366 #define FMULHW(a, b) ((int16_t)(a) * (int16_t)(b) >> 16)
367
368 #define FAVG(a, b) (((a) + (b) + 1) >> 1)
369 #endif
370
371 SSE_HELPER_B(helper_paddb, FADD)
372 SSE_HELPER_W(helper_paddw, FADD)
373 SSE_HELPER_L(helper_paddl, FADD)
374 SSE_HELPER_Q(helper_paddq, FADD)
375
376 SSE_HELPER_B(helper_psubb, FSUB)
377 SSE_HELPER_W(helper_psubw, FSUB)
378 SSE_HELPER_L(helper_psubl, FSUB)
379 SSE_HELPER_Q(helper_psubq, FSUB)
380
381 SSE_HELPER_B(helper_paddusb, FADDUB)
382 SSE_HELPER_B(helper_paddsb, FADDSB)
383 SSE_HELPER_B(helper_psubusb, FSUBUB)
384 SSE_HELPER_B(helper_psubsb, FSUBSB)
385
386 SSE_HELPER_W(helper_paddusw, FADDUW)
387 SSE_HELPER_W(helper_paddsw, FADDSW)
388 SSE_HELPER_W(helper_psubusw, FSUBUW)
389 SSE_HELPER_W(helper_psubsw, FSUBSW)
390
391 SSE_HELPER_B(helper_pminub, FMINUB)
392 SSE_HELPER_B(helper_pmaxub, FMAXUB)
393
394 SSE_HELPER_W(helper_pminsw, FMINSW)
395 SSE_HELPER_W(helper_pmaxsw, FMAXSW)
396
397 SSE_HELPER_Q(helper_pand, FAND)
398 SSE_HELPER_Q(helper_pandn, FANDN)
399 SSE_HELPER_Q(helper_por, FOR)
400 SSE_HELPER_Q(helper_pxor, FXOR)
401
402 SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
403 SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
404 SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
405
406 SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
407 SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
408 SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
409
410 SSE_HELPER_W(helper_pmullw, FMULLW)
411 #if SHIFT == 0
412 SSE_HELPER_W(helper_pmulhrw, FMULHRW)
413 #endif
414 SSE_HELPER_W(helper_pmulhuw, FMULHUW)
415 SSE_HELPER_W(helper_pmulhw, FMULHW)
416
417 SSE_HELPER_B(helper_pavgb, FAVG)
418 SSE_HELPER_W(helper_pavgw, FAVG)
419
420 void glue(helper_pmuludq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
421 {
422 d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
423 #if SHIFT == 1
424 d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
425 #endif
426 }
427
428 void glue(helper_pmaddwd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
429 {
430 int i;
431
432 for (i = 0; i < (2 << SHIFT); i++) {
433 d->L(i) = (int16_t)s->W(2 * i) * (int16_t)d->W(2 * i) +
434 (int16_t)s->W(2 * i + 1) * (int16_t)d->W(2 * i + 1);
435 }
436 }
437
438 #if SHIFT == 0
439 static inline int abs1(int a)
440 {
441 if (a < 0) {
442 return -a;
443 } else {
444 return a;
445 }
446 }
447 #endif
448 void glue(helper_psadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
449 {
450 unsigned int val;
451
452 val = 0;
453 val += abs1(d->B(0) - s->B(0));
454 val += abs1(d->B(1) - s->B(1));
455 val += abs1(d->B(2) - s->B(2));
456 val += abs1(d->B(3) - s->B(3));
457 val += abs1(d->B(4) - s->B(4));
458 val += abs1(d->B(5) - s->B(5));
459 val += abs1(d->B(6) - s->B(6));
460 val += abs1(d->B(7) - s->B(7));
461 d->Q(0) = val;
462 #if SHIFT == 1
463 val = 0;
464 val += abs1(d->B(8) - s->B(8));
465 val += abs1(d->B(9) - s->B(9));
466 val += abs1(d->B(10) - s->B(10));
467 val += abs1(d->B(11) - s->B(11));
468 val += abs1(d->B(12) - s->B(12));
469 val += abs1(d->B(13) - s->B(13));
470 val += abs1(d->B(14) - s->B(14));
471 val += abs1(d->B(15) - s->B(15));
472 d->Q(1) = val;
473 #endif
474 }
475
476 void glue(helper_maskmov, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
477 target_ulong a0)
478 {
479 int i;
480
481 for (i = 0; i < (8 << SHIFT); i++) {
482 if (s->B(i) & 0x80) {
483 cpu_stb_data(env, a0 + i, d->B(i));
484 }
485 }
486 }
487
488 void glue(helper_movl_mm_T0, SUFFIX)(Reg *d, uint32_t val)
489 {
490 d->L(0) = val;
491 d->L(1) = 0;
492 #if SHIFT == 1
493 d->Q(1) = 0;
494 #endif
495 }
496
497 #ifdef TARGET_X86_64
498 void glue(helper_movq_mm_T0, SUFFIX)(Reg *d, uint64_t val)
499 {
500 d->Q(0) = val;
501 #if SHIFT == 1
502 d->Q(1) = 0;
503 #endif
504 }
505 #endif
506
507 #if SHIFT == 0
508 void glue(helper_pshufw, SUFFIX)(Reg *d, Reg *s, int order)
509 {
510 Reg r;
511
512 r.W(0) = s->W(order & 3);
513 r.W(1) = s->W((order >> 2) & 3);
514 r.W(2) = s->W((order >> 4) & 3);
515 r.W(3) = s->W((order >> 6) & 3);
516 *d = r;
517 }
518 #else
519 void helper_shufps(Reg *d, Reg *s, int order)
520 {
521 Reg r;
522
523 r.L(0) = d->L(order & 3);
524 r.L(1) = d->L((order >> 2) & 3);
525 r.L(2) = s->L((order >> 4) & 3);
526 r.L(3) = s->L((order >> 6) & 3);
527 *d = r;
528 }
529
530 void helper_shufpd(Reg *d, Reg *s, int order)
531 {
532 Reg r;
533
534 r.Q(0) = d->Q(order & 1);
535 r.Q(1) = s->Q((order >> 1) & 1);
536 *d = r;
537 }
538
539 void glue(helper_pshufd, SUFFIX)(Reg *d, Reg *s, int order)
540 {
541 Reg r;
542
543 r.L(0) = s->L(order & 3);
544 r.L(1) = s->L((order >> 2) & 3);
545 r.L(2) = s->L((order >> 4) & 3);
546 r.L(3) = s->L((order >> 6) & 3);
547 *d = r;
548 }
549
550 void glue(helper_pshuflw, SUFFIX)(Reg *d, Reg *s, int order)
551 {
552 Reg r;
553
554 r.W(0) = s->W(order & 3);
555 r.W(1) = s->W((order >> 2) & 3);
556 r.W(2) = s->W((order >> 4) & 3);
557 r.W(3) = s->W((order >> 6) & 3);
558 r.Q(1) = s->Q(1);
559 *d = r;
560 }
561
562 void glue(helper_pshufhw, SUFFIX)(Reg *d, Reg *s, int order)
563 {
564 Reg r;
565
566 r.Q(0) = s->Q(0);
567 r.W(4) = s->W(4 + (order & 3));
568 r.W(5) = s->W(4 + ((order >> 2) & 3));
569 r.W(6) = s->W(4 + ((order >> 4) & 3));
570 r.W(7) = s->W(4 + ((order >> 6) & 3));
571 *d = r;
572 }
573 #endif
574
575 #if SHIFT == 1
576 /* FPU ops */
577 /* XXX: not accurate */
578
579 #define SSE_HELPER_S(name, F) \
580 void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s) \
581 { \
582 d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0)); \
583 d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1)); \
584 d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2)); \
585 d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3)); \
586 } \
587 \
588 void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s) \
589 { \
590 d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0)); \
591 } \
592 \
593 void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s) \
594 { \
595 d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0)); \
596 d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1)); \
597 } \
598 \
599 void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s) \
600 { \
601 d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0)); \
602 }
603
604 #define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
605 #define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
606 #define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
607 #define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
608 #define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
609
610 /* Note that the choice of comparison op here is important to get the
611 * special cases right: for min and max Intel specifies that (-0,0),
612 * (NaN, anything) and (anything, NaN) return the second argument.
613 */
614 #define FPU_MIN(size, a, b) \
615 (float ## size ## _lt(a, b, &env->sse_status) ? (a) : (b))
616 #define FPU_MAX(size, a, b) \
617 (float ## size ## _lt(b, a, &env->sse_status) ? (a) : (b))
618
619 SSE_HELPER_S(add, FPU_ADD)
620 SSE_HELPER_S(sub, FPU_SUB)
621 SSE_HELPER_S(mul, FPU_MUL)
622 SSE_HELPER_S(div, FPU_DIV)
623 SSE_HELPER_S(min, FPU_MIN)
624 SSE_HELPER_S(max, FPU_MAX)
625 SSE_HELPER_S(sqrt, FPU_SQRT)
626
627
628 /* float to float conversions */
629 void helper_cvtps2pd(CPUX86State *env, Reg *d, Reg *s)
630 {
631 float32 s0, s1;
632
633 s0 = s->XMM_S(0);
634 s1 = s->XMM_S(1);
635 d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
636 d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
637 }
638
639 void helper_cvtpd2ps(CPUX86State *env, Reg *d, Reg *s)
640 {
641 d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
642 d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
643 d->Q(1) = 0;
644 }
645
646 void helper_cvtss2sd(CPUX86State *env, Reg *d, Reg *s)
647 {
648 d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
649 }
650
651 void helper_cvtsd2ss(CPUX86State *env, Reg *d, Reg *s)
652 {
653 d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
654 }
655
656 /* integer to float */
657 void helper_cvtdq2ps(CPUX86State *env, Reg *d, Reg *s)
658 {
659 d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
660 d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
661 d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
662 d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
663 }
664
665 void helper_cvtdq2pd(CPUX86State *env, Reg *d, Reg *s)
666 {
667 int32_t l0, l1;
668
669 l0 = (int32_t)s->XMM_L(0);
670 l1 = (int32_t)s->XMM_L(1);
671 d->XMM_D(0) = int32_to_float64(l0, &env->sse_status);
672 d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
673 }
674
675 void helper_cvtpi2ps(CPUX86State *env, XMMReg *d, MMXReg *s)
676 {
677 d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
678 d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
679 }
680
681 void helper_cvtpi2pd(CPUX86State *env, XMMReg *d, MMXReg *s)
682 {
683 d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
684 d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
685 }
686
687 void helper_cvtsi2ss(CPUX86State *env, XMMReg *d, uint32_t val)
688 {
689 d->XMM_S(0) = int32_to_float32(val, &env->sse_status);
690 }
691
692 void helper_cvtsi2sd(CPUX86State *env, XMMReg *d, uint32_t val)
693 {
694 d->XMM_D(0) = int32_to_float64(val, &env->sse_status);
695 }
696
697 #ifdef TARGET_X86_64
698 void helper_cvtsq2ss(CPUX86State *env, XMMReg *d, uint64_t val)
699 {
700 d->XMM_S(0) = int64_to_float32(val, &env->sse_status);
701 }
702
703 void helper_cvtsq2sd(CPUX86State *env, XMMReg *d, uint64_t val)
704 {
705 d->XMM_D(0) = int64_to_float64(val, &env->sse_status);
706 }
707 #endif
708
709 /* float to integer */
710 void helper_cvtps2dq(CPUX86State *env, XMMReg *d, XMMReg *s)
711 {
712 d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
713 d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
714 d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
715 d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
716 }
717
718 void helper_cvtpd2dq(CPUX86State *env, XMMReg *d, XMMReg *s)
719 {
720 d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
721 d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
722 d->XMM_Q(1) = 0;
723 }
724
725 void helper_cvtps2pi(CPUX86State *env, MMXReg *d, XMMReg *s)
726 {
727 d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
728 d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
729 }
730
731 void helper_cvtpd2pi(CPUX86State *env, MMXReg *d, XMMReg *s)
732 {
733 d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
734 d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
735 }
736
737 int32_t helper_cvtss2si(CPUX86State *env, XMMReg *s)
738 {
739 return float32_to_int32(s->XMM_S(0), &env->sse_status);
740 }
741
742 int32_t helper_cvtsd2si(CPUX86State *env, XMMReg *s)
743 {
744 return float64_to_int32(s->XMM_D(0), &env->sse_status);
745 }
746
747 #ifdef TARGET_X86_64
748 int64_t helper_cvtss2sq(CPUX86State *env, XMMReg *s)
749 {
750 return float32_to_int64(s->XMM_S(0), &env->sse_status);
751 }
752
753 int64_t helper_cvtsd2sq(CPUX86State *env, XMMReg *s)
754 {
755 return float64_to_int64(s->XMM_D(0), &env->sse_status);
756 }
757 #endif
758
759 /* float to integer truncated */
760 void helper_cvttps2dq(CPUX86State *env, XMMReg *d, XMMReg *s)
761 {
762 d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
763 d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
764 d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
765 d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
766 }
767
768 void helper_cvttpd2dq(CPUX86State *env, XMMReg *d, XMMReg *s)
769 {
770 d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
771 d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
772 d->XMM_Q(1) = 0;
773 }
774
775 void helper_cvttps2pi(CPUX86State *env, MMXReg *d, XMMReg *s)
776 {
777 d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
778 d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
779 }
780
781 void helper_cvttpd2pi(CPUX86State *env, MMXReg *d, XMMReg *s)
782 {
783 d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
784 d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
785 }
786
787 int32_t helper_cvttss2si(CPUX86State *env, XMMReg *s)
788 {
789 return float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
790 }
791
792 int32_t helper_cvttsd2si(CPUX86State *env, XMMReg *s)
793 {
794 return float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
795 }
796
797 #ifdef TARGET_X86_64
798 int64_t helper_cvttss2sq(CPUX86State *env, XMMReg *s)
799 {
800 return float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
801 }
802
803 int64_t helper_cvttsd2sq(CPUX86State *env, XMMReg *s)
804 {
805 return float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
806 }
807 #endif
808
809 void helper_rsqrtps(CPUX86State *env, XMMReg *d, XMMReg *s)
810 {
811 d->XMM_S(0) = float32_div(float32_one,
812 float32_sqrt(s->XMM_S(0), &env->sse_status),
813 &env->sse_status);
814 d->XMM_S(1) = float32_div(float32_one,
815 float32_sqrt(s->XMM_S(1), &env->sse_status),
816 &env->sse_status);
817 d->XMM_S(2) = float32_div(float32_one,
818 float32_sqrt(s->XMM_S(2), &env->sse_status),
819 &env->sse_status);
820 d->XMM_S(3) = float32_div(float32_one,
821 float32_sqrt(s->XMM_S(3), &env->sse_status),
822 &env->sse_status);
823 }
824
825 void helper_rsqrtss(CPUX86State *env, XMMReg *d, XMMReg *s)
826 {
827 d->XMM_S(0) = float32_div(float32_one,
828 float32_sqrt(s->XMM_S(0), &env->sse_status),
829 &env->sse_status);
830 }
831
832 void helper_rcpps(CPUX86State *env, XMMReg *d, XMMReg *s)
833 {
834 d->XMM_S(0) = float32_div(float32_one, s->XMM_S(0), &env->sse_status);
835 d->XMM_S(1) = float32_div(float32_one, s->XMM_S(1), &env->sse_status);
836 d->XMM_S(2) = float32_div(float32_one, s->XMM_S(2), &env->sse_status);
837 d->XMM_S(3) = float32_div(float32_one, s->XMM_S(3), &env->sse_status);
838 }
839
840 void helper_rcpss(CPUX86State *env, XMMReg *d, XMMReg *s)
841 {
842 d->XMM_S(0) = float32_div(float32_one, s->XMM_S(0), &env->sse_status);
843 }
844
845 static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
846 {
847 uint64_t mask;
848
849 if (len == 0) {
850 mask = ~0LL;
851 } else {
852 mask = (1ULL << len) - 1;
853 }
854 return (src >> shift) & mask;
855 }
856
857 void helper_extrq_r(CPUX86State *env, XMMReg *d, XMMReg *s)
858 {
859 d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), s->XMM_B(1), s->XMM_B(0));
860 }
861
862 void helper_extrq_i(CPUX86State *env, XMMReg *d, int index, int length)
863 {
864 d->XMM_Q(0) = helper_extrq(d->XMM_Q(0), index, length);
865 }
866
867 static inline uint64_t helper_insertq(uint64_t src, int shift, int len)
868 {
869 uint64_t mask;
870
871 if (len == 0) {
872 mask = ~0ULL;
873 } else {
874 mask = (1ULL << len) - 1;
875 }
876 return (src & ~(mask << shift)) | ((src & mask) << shift);
877 }
878
879 void helper_insertq_r(CPUX86State *env, XMMReg *d, XMMReg *s)
880 {
881 d->XMM_Q(0) = helper_insertq(s->XMM_Q(0), s->XMM_B(9), s->XMM_B(8));
882 }
883
884 void helper_insertq_i(CPUX86State *env, XMMReg *d, int index, int length)
885 {
886 d->XMM_Q(0) = helper_insertq(d->XMM_Q(0), index, length);
887 }
888
889 void helper_haddps(CPUX86State *env, XMMReg *d, XMMReg *s)
890 {
891 XMMReg r;
892
893 r.XMM_S(0) = float32_add(d->XMM_S(0), d->XMM_S(1), &env->sse_status);
894 r.XMM_S(1) = float32_add(d->XMM_S(2), d->XMM_S(3), &env->sse_status);
895 r.XMM_S(2) = float32_add(s->XMM_S(0), s->XMM_S(1), &env->sse_status);
896 r.XMM_S(3) = float32_add(s->XMM_S(2), s->XMM_S(3), &env->sse_status);
897 *d = r;
898 }
899
900 void helper_haddpd(CPUX86State *env, XMMReg *d, XMMReg *s)
901 {
902 XMMReg r;
903
904 r.XMM_D(0) = float64_add(d->XMM_D(0), d->XMM_D(1), &env->sse_status);
905 r.XMM_D(1) = float64_add(s->XMM_D(0), s->XMM_D(1), &env->sse_status);
906 *d = r;
907 }
908
909 void helper_hsubps(CPUX86State *env, XMMReg *d, XMMReg *s)
910 {
911 XMMReg r;
912
913 r.XMM_S(0) = float32_sub(d->XMM_S(0), d->XMM_S(1), &env->sse_status);
914 r.XMM_S(1) = float32_sub(d->XMM_S(2), d->XMM_S(3), &env->sse_status);
915 r.XMM_S(2) = float32_sub(s->XMM_S(0), s->XMM_S(1), &env->sse_status);
916 r.XMM_S(3) = float32_sub(s->XMM_S(2), s->XMM_S(3), &env->sse_status);
917 *d = r;
918 }
919
920 void helper_hsubpd(CPUX86State *env, XMMReg *d, XMMReg *s)
921 {
922 XMMReg r;
923
924 r.XMM_D(0) = float64_sub(d->XMM_D(0), d->XMM_D(1), &env->sse_status);
925 r.XMM_D(1) = float64_sub(s->XMM_D(0), s->XMM_D(1), &env->sse_status);
926 *d = r;
927 }
928
929 void helper_addsubps(CPUX86State *env, XMMReg *d, XMMReg *s)
930 {
931 d->XMM_S(0) = float32_sub(d->XMM_S(0), s->XMM_S(0), &env->sse_status);
932 d->XMM_S(1) = float32_add(d->XMM_S(1), s->XMM_S(1), &env->sse_status);
933 d->XMM_S(2) = float32_sub(d->XMM_S(2), s->XMM_S(2), &env->sse_status);
934 d->XMM_S(3) = float32_add(d->XMM_S(3), s->XMM_S(3), &env->sse_status);
935 }
936
937 void helper_addsubpd(CPUX86State *env, XMMReg *d, XMMReg *s)
938 {
939 d->XMM_D(0) = float64_sub(d->XMM_D(0), s->XMM_D(0), &env->sse_status);
940 d->XMM_D(1) = float64_add(d->XMM_D(1), s->XMM_D(1), &env->sse_status);
941 }
942
943 /* XXX: unordered */
944 #define SSE_HELPER_CMP(name, F) \
945 void helper_ ## name ## ps(CPUX86State *env, Reg *d, Reg *s) \
946 { \
947 d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0)); \
948 d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1)); \
949 d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2)); \
950 d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3)); \
951 } \
952 \
953 void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *s) \
954 { \
955 d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0)); \
956 } \
957 \
958 void helper_ ## name ## pd(CPUX86State *env, Reg *d, Reg *s) \
959 { \
960 d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0)); \
961 d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1)); \
962 } \
963 \
964 void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *s) \
965 { \
966 d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0)); \
967 }
968
969 #define FPU_CMPEQ(size, a, b) \
970 (float ## size ## _eq_quiet(a, b, &env->sse_status) ? -1 : 0)
971 #define FPU_CMPLT(size, a, b) \
972 (float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0)
973 #define FPU_CMPLE(size, a, b) \
974 (float ## size ## _le(a, b, &env->sse_status) ? -1 : 0)
975 #define FPU_CMPUNORD(size, a, b) \
976 (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? -1 : 0)
977 #define FPU_CMPNEQ(size, a, b) \
978 (float ## size ## _eq_quiet(a, b, &env->sse_status) ? 0 : -1)
979 #define FPU_CMPNLT(size, a, b) \
980 (float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1)
981 #define FPU_CMPNLE(size, a, b) \
982 (float ## size ## _le(a, b, &env->sse_status) ? 0 : -1)
983 #define FPU_CMPORD(size, a, b) \
984 (float ## size ## _unordered_quiet(a, b, &env->sse_status) ? 0 : -1)
985
986 SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
987 SSE_HELPER_CMP(cmplt, FPU_CMPLT)
988 SSE_HELPER_CMP(cmple, FPU_CMPLE)
989 SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
990 SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
991 SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
992 SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
993 SSE_HELPER_CMP(cmpord, FPU_CMPORD)
994
995 static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
996
997 void helper_ucomiss(CPUX86State *env, Reg *d, Reg *s)
998 {
999 int ret;
1000 float32 s0, s1;
1001
1002 s0 = d->XMM_S(0);
1003 s1 = s->XMM_S(0);
1004 ret = float32_compare_quiet(s0, s1, &env->sse_status);
1005 CC_SRC = comis_eflags[ret + 1];
1006 }
1007
1008 void helper_comiss(CPUX86State *env, Reg *d, Reg *s)
1009 {
1010 int ret;
1011 float32 s0, s1;
1012
1013 s0 = d->XMM_S(0);
1014 s1 = s->XMM_S(0);
1015 ret = float32_compare(s0, s1, &env->sse_status);
1016 CC_SRC = comis_eflags[ret + 1];
1017 }
1018
1019 void helper_ucomisd(CPUX86State *env, Reg *d, Reg *s)
1020 {
1021 int ret;
1022 float64 d0, d1;
1023
1024 d0 = d->XMM_D(0);
1025 d1 = s->XMM_D(0);
1026 ret = float64_compare_quiet(d0, d1, &env->sse_status);
1027 CC_SRC = comis_eflags[ret + 1];
1028 }
1029
1030 void helper_comisd(CPUX86State *env, Reg *d, Reg *s)
1031 {
1032 int ret;
1033 float64 d0, d1;
1034
1035 d0 = d->XMM_D(0);
1036 d1 = s->XMM_D(0);
1037 ret = float64_compare(d0, d1, &env->sse_status);
1038 CC_SRC = comis_eflags[ret + 1];
1039 }
1040
1041 uint32_t helper_movmskps(CPUX86State *env, Reg *s)
1042 {
1043 int b0, b1, b2, b3;
1044
1045 b0 = s->XMM_L(0) >> 31;
1046 b1 = s->XMM_L(1) >> 31;
1047 b2 = s->XMM_L(2) >> 31;
1048 b3 = s->XMM_L(3) >> 31;
1049 return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
1050 }
1051
1052 uint32_t helper_movmskpd(CPUX86State *env, Reg *s)
1053 {
1054 int b0, b1;
1055
1056 b0 = s->XMM_L(1) >> 31;
1057 b1 = s->XMM_L(3) >> 31;
1058 return b0 | (b1 << 1);
1059 }
1060
1061 #endif
1062
1063 uint32_t glue(helper_pmovmskb, SUFFIX)(CPUX86State *env, Reg *s)
1064 {
1065 uint32_t val;
1066
1067 val = 0;
1068 val |= (s->B(0) >> 7);
1069 val |= (s->B(1) >> 6) & 0x02;
1070 val |= (s->B(2) >> 5) & 0x04;
1071 val |= (s->B(3) >> 4) & 0x08;
1072 val |= (s->B(4) >> 3) & 0x10;
1073 val |= (s->B(5) >> 2) & 0x20;
1074 val |= (s->B(6) >> 1) & 0x40;
1075 val |= (s->B(7)) & 0x80;
1076 #if SHIFT == 1
1077 val |= (s->B(8) << 1) & 0x0100;
1078 val |= (s->B(9) << 2) & 0x0200;
1079 val |= (s->B(10) << 3) & 0x0400;
1080 val |= (s->B(11) << 4) & 0x0800;
1081 val |= (s->B(12) << 5) & 0x1000;
1082 val |= (s->B(13) << 6) & 0x2000;
1083 val |= (s->B(14) << 7) & 0x4000;
1084 val |= (s->B(15) << 8) & 0x8000;
1085 #endif
1086 return val;
1087 }
1088
1089 void glue(helper_packsswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1090 {
1091 Reg r;
1092
1093 r.B(0) = satsb((int16_t)d->W(0));
1094 r.B(1) = satsb((int16_t)d->W(1));
1095 r.B(2) = satsb((int16_t)d->W(2));
1096 r.B(3) = satsb((int16_t)d->W(3));
1097 #if SHIFT == 1
1098 r.B(4) = satsb((int16_t)d->W(4));
1099 r.B(5) = satsb((int16_t)d->W(5));
1100 r.B(6) = satsb((int16_t)d->W(6));
1101 r.B(7) = satsb((int16_t)d->W(7));
1102 #endif
1103 r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1104 r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1105 r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1106 r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1107 #if SHIFT == 1
1108 r.B(12) = satsb((int16_t)s->W(4));
1109 r.B(13) = satsb((int16_t)s->W(5));
1110 r.B(14) = satsb((int16_t)s->W(6));
1111 r.B(15) = satsb((int16_t)s->W(7));
1112 #endif
1113 *d = r;
1114 }
1115
1116 void glue(helper_packuswb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1117 {
1118 Reg r;
1119
1120 r.B(0) = satub((int16_t)d->W(0));
1121 r.B(1) = satub((int16_t)d->W(1));
1122 r.B(2) = satub((int16_t)d->W(2));
1123 r.B(3) = satub((int16_t)d->W(3));
1124 #if SHIFT == 1
1125 r.B(4) = satub((int16_t)d->W(4));
1126 r.B(5) = satub((int16_t)d->W(5));
1127 r.B(6) = satub((int16_t)d->W(6));
1128 r.B(7) = satub((int16_t)d->W(7));
1129 #endif
1130 r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1131 r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1132 r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1133 r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1134 #if SHIFT == 1
1135 r.B(12) = satub((int16_t)s->W(4));
1136 r.B(13) = satub((int16_t)s->W(5));
1137 r.B(14) = satub((int16_t)s->W(6));
1138 r.B(15) = satub((int16_t)s->W(7));
1139 #endif
1140 *d = r;
1141 }
1142
1143 void glue(helper_packssdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1144 {
1145 Reg r;
1146
1147 r.W(0) = satsw(d->L(0));
1148 r.W(1) = satsw(d->L(1));
1149 #if SHIFT == 1
1150 r.W(2) = satsw(d->L(2));
1151 r.W(3) = satsw(d->L(3));
1152 #endif
1153 r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1154 r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1155 #if SHIFT == 1
1156 r.W(6) = satsw(s->L(2));
1157 r.W(7) = satsw(s->L(3));
1158 #endif
1159 *d = r;
1160 }
1161
1162 #define UNPCK_OP(base_name, base) \
1163 \
1164 void glue(helper_punpck ## base_name ## bw, SUFFIX)(CPUX86State *env,\
1165 Reg *d, Reg *s) \
1166 { \
1167 Reg r; \
1168 \
1169 r.B(0) = d->B((base << (SHIFT + 2)) + 0); \
1170 r.B(1) = s->B((base << (SHIFT + 2)) + 0); \
1171 r.B(2) = d->B((base << (SHIFT + 2)) + 1); \
1172 r.B(3) = s->B((base << (SHIFT + 2)) + 1); \
1173 r.B(4) = d->B((base << (SHIFT + 2)) + 2); \
1174 r.B(5) = s->B((base << (SHIFT + 2)) + 2); \
1175 r.B(6) = d->B((base << (SHIFT + 2)) + 3); \
1176 r.B(7) = s->B((base << (SHIFT + 2)) + 3); \
1177 XMM_ONLY( \
1178 r.B(8) = d->B((base << (SHIFT + 2)) + 4); \
1179 r.B(9) = s->B((base << (SHIFT + 2)) + 4); \
1180 r.B(10) = d->B((base << (SHIFT + 2)) + 5); \
1181 r.B(11) = s->B((base << (SHIFT + 2)) + 5); \
1182 r.B(12) = d->B((base << (SHIFT + 2)) + 6); \
1183 r.B(13) = s->B((base << (SHIFT + 2)) + 6); \
1184 r.B(14) = d->B((base << (SHIFT + 2)) + 7); \
1185 r.B(15) = s->B((base << (SHIFT + 2)) + 7); \
1186 ) \
1187 *d = r; \
1188 } \
1189 \
1190 void glue(helper_punpck ## base_name ## wd, SUFFIX)(CPUX86State *env,\
1191 Reg *d, Reg *s) \
1192 { \
1193 Reg r; \
1194 \
1195 r.W(0) = d->W((base << (SHIFT + 1)) + 0); \
1196 r.W(1) = s->W((base << (SHIFT + 1)) + 0); \
1197 r.W(2) = d->W((base << (SHIFT + 1)) + 1); \
1198 r.W(3) = s->W((base << (SHIFT + 1)) + 1); \
1199 XMM_ONLY( \
1200 r.W(4) = d->W((base << (SHIFT + 1)) + 2); \
1201 r.W(5) = s->W((base << (SHIFT + 1)) + 2); \
1202 r.W(6) = d->W((base << (SHIFT + 1)) + 3); \
1203 r.W(7) = s->W((base << (SHIFT + 1)) + 3); \
1204 ) \
1205 *d = r; \
1206 } \
1207 \
1208 void glue(helper_punpck ## base_name ## dq, SUFFIX)(CPUX86State *env,\
1209 Reg *d, Reg *s) \
1210 { \
1211 Reg r; \
1212 \
1213 r.L(0) = d->L((base << SHIFT) + 0); \
1214 r.L(1) = s->L((base << SHIFT) + 0); \
1215 XMM_ONLY( \
1216 r.L(2) = d->L((base << SHIFT) + 1); \
1217 r.L(3) = s->L((base << SHIFT) + 1); \
1218 ) \
1219 *d = r; \
1220 } \
1221 \
1222 XMM_ONLY( \
1223 void glue(helper_punpck ## base_name ## qdq, SUFFIX)(CPUX86State \
1224 *env, \
1225 Reg *d, \
1226 Reg *s) \
1227 { \
1228 Reg r; \
1229 \
1230 r.Q(0) = d->Q(base); \
1231 r.Q(1) = s->Q(base); \
1232 *d = r; \
1233 } \
1234 )
1235
1236 UNPCK_OP(l, 0)
1237 UNPCK_OP(h, 1)
1238
1239 /* 3DNow! float ops */
1240 #if SHIFT == 0
1241 void helper_pi2fd(CPUX86State *env, MMXReg *d, MMXReg *s)
1242 {
1243 d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
1244 d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
1245 }
1246
1247 void helper_pi2fw(CPUX86State *env, MMXReg *d, MMXReg *s)
1248 {
1249 d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
1250 d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
1251 }
1252
1253 void helper_pf2id(CPUX86State *env, MMXReg *d, MMXReg *s)
1254 {
1255 d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
1256 d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
1257 }
1258
1259 void helper_pf2iw(CPUX86State *env, MMXReg *d, MMXReg *s)
1260 {
1261 d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0),
1262 &env->mmx_status));
1263 d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1),
1264 &env->mmx_status));
1265 }
1266
1267 void helper_pfacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1268 {
1269 MMXReg r;
1270
1271 r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1272 r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1273 *d = r;
1274 }
1275
1276 void helper_pfadd(CPUX86State *env, MMXReg *d, MMXReg *s)
1277 {
1278 d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1279 d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1280 }
1281
1282 void helper_pfcmpeq(CPUX86State *env, MMXReg *d, MMXReg *s)
1283 {
1284 d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0),
1285 &env->mmx_status) ? -1 : 0;
1286 d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1),
1287 &env->mmx_status) ? -1 : 0;
1288 }
1289
1290 void helper_pfcmpge(CPUX86State *env, MMXReg *d, MMXReg *s)
1291 {
1292 d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0),
1293 &env->mmx_status) ? -1 : 0;
1294 d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1),
1295 &env->mmx_status) ? -1 : 0;
1296 }
1297
1298 void helper_pfcmpgt(CPUX86State *env, MMXReg *d, MMXReg *s)
1299 {
1300 d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0),
1301 &env->mmx_status) ? -1 : 0;
1302 d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1),
1303 &env->mmx_status) ? -1 : 0;
1304 }
1305
1306 void helper_pfmax(CPUX86State *env, MMXReg *d, MMXReg *s)
1307 {
1308 if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status)) {
1309 d->MMX_S(0) = s->MMX_S(0);
1310 }
1311 if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status)) {
1312 d->MMX_S(1) = s->MMX_S(1);
1313 }
1314 }
1315
1316 void helper_pfmin(CPUX86State *env, MMXReg *d, MMXReg *s)
1317 {
1318 if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status)) {
1319 d->MMX_S(0) = s->MMX_S(0);
1320 }
1321 if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status)) {
1322 d->MMX_S(1) = s->MMX_S(1);
1323 }
1324 }
1325
1326 void helper_pfmul(CPUX86State *env, MMXReg *d, MMXReg *s)
1327 {
1328 d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1329 d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1330 }
1331
1332 void helper_pfnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1333 {
1334 MMXReg r;
1335
1336 r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1337 r.MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1338 *d = r;
1339 }
1340
1341 void helper_pfpnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
1342 {
1343 MMXReg r;
1344
1345 r.MMX_S(0) = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
1346 r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
1347 *d = r;
1348 }
1349
1350 void helper_pfrcp(CPUX86State *env, MMXReg *d, MMXReg *s)
1351 {
1352 d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
1353 d->MMX_S(1) = d->MMX_S(0);
1354 }
1355
1356 void helper_pfrsqrt(CPUX86State *env, MMXReg *d, MMXReg *s)
1357 {
1358 d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
1359 d->MMX_S(1) = float32_div(float32_one,
1360 float32_sqrt(d->MMX_S(1), &env->mmx_status),
1361 &env->mmx_status);
1362 d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
1363 d->MMX_L(0) = d->MMX_L(1);
1364 }
1365
1366 void helper_pfsub(CPUX86State *env, MMXReg *d, MMXReg *s)
1367 {
1368 d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
1369 d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
1370 }
1371
1372 void helper_pfsubr(CPUX86State *env, MMXReg *d, MMXReg *s)
1373 {
1374 d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
1375 d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
1376 }
1377
1378 void helper_pswapd(CPUX86State *env, MMXReg *d, MMXReg *s)
1379 {
1380 MMXReg r;
1381
1382 r.MMX_L(0) = s->MMX_L(1);
1383 r.MMX_L(1) = s->MMX_L(0);
1384 *d = r;
1385 }
1386 #endif
1387
1388 /* SSSE3 op helpers */
1389 void glue(helper_pshufb, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1390 {
1391 int i;
1392 Reg r;
1393
1394 for (i = 0; i < (8 << SHIFT); i++) {
1395 r.B(i) = (s->B(i) & 0x80) ? 0 : (d->B(s->B(i) & ((8 << SHIFT) - 1)));
1396 }
1397
1398 *d = r;
1399 }
1400
1401 void glue(helper_phaddw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1402 {
1403 d->W(0) = (int16_t)d->W(0) + (int16_t)d->W(1);
1404 d->W(1) = (int16_t)d->W(2) + (int16_t)d->W(3);
1405 XMM_ONLY(d->W(2) = (int16_t)d->W(4) + (int16_t)d->W(5));
1406 XMM_ONLY(d->W(3) = (int16_t)d->W(6) + (int16_t)d->W(7));
1407 d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) + (int16_t)s->W(1);
1408 d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) + (int16_t)s->W(3);
1409 XMM_ONLY(d->W(6) = (int16_t)s->W(4) + (int16_t)s->W(5));
1410 XMM_ONLY(d->W(7) = (int16_t)s->W(6) + (int16_t)s->W(7));
1411 }
1412
1413 void glue(helper_phaddd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1414 {
1415 d->L(0) = (int32_t)d->L(0) + (int32_t)d->L(1);
1416 XMM_ONLY(d->L(1) = (int32_t)d->L(2) + (int32_t)d->L(3));
1417 d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) + (int32_t)s->L(1);
1418 XMM_ONLY(d->L(3) = (int32_t)s->L(2) + (int32_t)s->L(3));
1419 }
1420
1421 void glue(helper_phaddsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1422 {
1423 d->W(0) = satsw((int16_t)d->W(0) + (int16_t)d->W(1));
1424 d->W(1) = satsw((int16_t)d->W(2) + (int16_t)d->W(3));
1425 XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) + (int16_t)d->W(5)));
1426 XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) + (int16_t)d->W(7)));
1427 d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) + (int16_t)s->W(1));
1428 d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) + (int16_t)s->W(3));
1429 XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) + (int16_t)s->W(5)));
1430 XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) + (int16_t)s->W(7)));
1431 }
1432
1433 void glue(helper_pmaddubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1434 {
1435 d->W(0) = satsw((int8_t)s->B(0) * (uint8_t)d->B(0) +
1436 (int8_t)s->B(1) * (uint8_t)d->B(1));
1437 d->W(1) = satsw((int8_t)s->B(2) * (uint8_t)d->B(2) +
1438 (int8_t)s->B(3) * (uint8_t)d->B(3));
1439 d->W(2) = satsw((int8_t)s->B(4) * (uint8_t)d->B(4) +
1440 (int8_t)s->B(5) * (uint8_t)d->B(5));
1441 d->W(3) = satsw((int8_t)s->B(6) * (uint8_t)d->B(6) +
1442 (int8_t)s->B(7) * (uint8_t)d->B(7));
1443 #if SHIFT == 1
1444 d->W(4) = satsw((int8_t)s->B(8) * (uint8_t)d->B(8) +
1445 (int8_t)s->B(9) * (uint8_t)d->B(9));
1446 d->W(5) = satsw((int8_t)s->B(10) * (uint8_t)d->B(10) +
1447 (int8_t)s->B(11) * (uint8_t)d->B(11));
1448 d->W(6) = satsw((int8_t)s->B(12) * (uint8_t)d->B(12) +
1449 (int8_t)s->B(13) * (uint8_t)d->B(13));
1450 d->W(7) = satsw((int8_t)s->B(14) * (uint8_t)d->B(14) +
1451 (int8_t)s->B(15) * (uint8_t)d->B(15));
1452 #endif
1453 }
1454
1455 void glue(helper_phsubw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1456 {
1457 d->W(0) = (int16_t)d->W(0) - (int16_t)d->W(1);
1458 d->W(1) = (int16_t)d->W(2) - (int16_t)d->W(3);
1459 XMM_ONLY(d->W(2) = (int16_t)d->W(4) - (int16_t)d->W(5));
1460 XMM_ONLY(d->W(3) = (int16_t)d->W(6) - (int16_t)d->W(7));
1461 d->W((2 << SHIFT) + 0) = (int16_t)s->W(0) - (int16_t)s->W(1);
1462 d->W((2 << SHIFT) + 1) = (int16_t)s->W(2) - (int16_t)s->W(3);
1463 XMM_ONLY(d->W(6) = (int16_t)s->W(4) - (int16_t)s->W(5));
1464 XMM_ONLY(d->W(7) = (int16_t)s->W(6) - (int16_t)s->W(7));
1465 }
1466
1467 void glue(helper_phsubd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1468 {
1469 d->L(0) = (int32_t)d->L(0) - (int32_t)d->L(1);
1470 XMM_ONLY(d->L(1) = (int32_t)d->L(2) - (int32_t)d->L(3));
1471 d->L((1 << SHIFT) + 0) = (int32_t)s->L(0) - (int32_t)s->L(1);
1472 XMM_ONLY(d->L(3) = (int32_t)s->L(2) - (int32_t)s->L(3));
1473 }
1474
1475 void glue(helper_phsubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1476 {
1477 d->W(0) = satsw((int16_t)d->W(0) - (int16_t)d->W(1));
1478 d->W(1) = satsw((int16_t)d->W(2) - (int16_t)d->W(3));
1479 XMM_ONLY(d->W(2) = satsw((int16_t)d->W(4) - (int16_t)d->W(5)));
1480 XMM_ONLY(d->W(3) = satsw((int16_t)d->W(6) - (int16_t)d->W(7)));
1481 d->W((2 << SHIFT) + 0) = satsw((int16_t)s->W(0) - (int16_t)s->W(1));
1482 d->W((2 << SHIFT) + 1) = satsw((int16_t)s->W(2) - (int16_t)s->W(3));
1483 XMM_ONLY(d->W(6) = satsw((int16_t)s->W(4) - (int16_t)s->W(5)));
1484 XMM_ONLY(d->W(7) = satsw((int16_t)s->W(6) - (int16_t)s->W(7)));
1485 }
1486
1487 #define FABSB(_, x) (x > INT8_MAX ? -(int8_t)x : x)
1488 #define FABSW(_, x) (x > INT16_MAX ? -(int16_t)x : x)
1489 #define FABSL(_, x) (x > INT32_MAX ? -(int32_t)x : x)
1490 SSE_HELPER_B(helper_pabsb, FABSB)
1491 SSE_HELPER_W(helper_pabsw, FABSW)
1492 SSE_HELPER_L(helper_pabsd, FABSL)
1493
1494 #define FMULHRSW(d, s) (((int16_t) d * (int16_t)s + 0x4000) >> 15)
1495 SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
1496
1497 #define FSIGNB(d, s) (s <= INT8_MAX ? s ? d : 0 : -(int8_t)d)
1498 #define FSIGNW(d, s) (s <= INT16_MAX ? s ? d : 0 : -(int16_t)d)
1499 #define FSIGNL(d, s) (s <= INT32_MAX ? s ? d : 0 : -(int32_t)d)
1500 SSE_HELPER_B(helper_psignb, FSIGNB)
1501 SSE_HELPER_W(helper_psignw, FSIGNW)
1502 SSE_HELPER_L(helper_psignd, FSIGNL)
1503
1504 void glue(helper_palignr, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1505 int32_t shift)
1506 {
1507 Reg r;
1508
1509 /* XXX could be checked during translation */
1510 if (shift >= (16 << SHIFT)) {
1511 r.Q(0) = 0;
1512 XMM_ONLY(r.Q(1) = 0);
1513 } else {
1514 shift <<= 3;
1515 #define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
1516 #if SHIFT == 0
1517 r.Q(0) = SHR(s->Q(0), shift - 0) |
1518 SHR(d->Q(0), shift - 64);
1519 #else
1520 r.Q(0) = SHR(s->Q(0), shift - 0) |
1521 SHR(s->Q(1), shift - 64) |
1522 SHR(d->Q(0), shift - 128) |
1523 SHR(d->Q(1), shift - 192);
1524 r.Q(1) = SHR(s->Q(0), shift + 64) |
1525 SHR(s->Q(1), shift - 0) |
1526 SHR(d->Q(0), shift - 64) |
1527 SHR(d->Q(1), shift - 128);
1528 #endif
1529 #undef SHR
1530 }
1531
1532 *d = r;
1533 }
1534
1535 #define XMM0 (env->xmm_regs[0])
1536
1537 #if SHIFT == 1
1538 #define SSE_HELPER_V(name, elem, num, F) \
1539 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
1540 { \
1541 d->elem(0) = F(d->elem(0), s->elem(0), XMM0.elem(0)); \
1542 d->elem(1) = F(d->elem(1), s->elem(1), XMM0.elem(1)); \
1543 if (num > 2) { \
1544 d->elem(2) = F(d->elem(2), s->elem(2), XMM0.elem(2)); \
1545 d->elem(3) = F(d->elem(3), s->elem(3), XMM0.elem(3)); \
1546 if (num > 4) { \
1547 d->elem(4) = F(d->elem(4), s->elem(4), XMM0.elem(4)); \
1548 d->elem(5) = F(d->elem(5), s->elem(5), XMM0.elem(5)); \
1549 d->elem(6) = F(d->elem(6), s->elem(6), XMM0.elem(6)); \
1550 d->elem(7) = F(d->elem(7), s->elem(7), XMM0.elem(7)); \
1551 if (num > 8) { \
1552 d->elem(8) = F(d->elem(8), s->elem(8), XMM0.elem(8)); \
1553 d->elem(9) = F(d->elem(9), s->elem(9), XMM0.elem(9)); \
1554 d->elem(10) = F(d->elem(10), s->elem(10), XMM0.elem(10)); \
1555 d->elem(11) = F(d->elem(11), s->elem(11), XMM0.elem(11)); \
1556 d->elem(12) = F(d->elem(12), s->elem(12), XMM0.elem(12)); \
1557 d->elem(13) = F(d->elem(13), s->elem(13), XMM0.elem(13)); \
1558 d->elem(14) = F(d->elem(14), s->elem(14), XMM0.elem(14)); \
1559 d->elem(15) = F(d->elem(15), s->elem(15), XMM0.elem(15)); \
1560 } \
1561 } \
1562 } \
1563 }
1564
1565 #define SSE_HELPER_I(name, elem, num, F) \
1566 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t imm) \
1567 { \
1568 d->elem(0) = F(d->elem(0), s->elem(0), ((imm >> 0) & 1)); \
1569 d->elem(1) = F(d->elem(1), s->elem(1), ((imm >> 1) & 1)); \
1570 if (num > 2) { \
1571 d->elem(2) = F(d->elem(2), s->elem(2), ((imm >> 2) & 1)); \
1572 d->elem(3) = F(d->elem(3), s->elem(3), ((imm >> 3) & 1)); \
1573 if (num > 4) { \
1574 d->elem(4) = F(d->elem(4), s->elem(4), ((imm >> 4) & 1)); \
1575 d->elem(5) = F(d->elem(5), s->elem(5), ((imm >> 5) & 1)); \
1576 d->elem(6) = F(d->elem(6), s->elem(6), ((imm >> 6) & 1)); \
1577 d->elem(7) = F(d->elem(7), s->elem(7), ((imm >> 7) & 1)); \
1578 if (num > 8) { \
1579 d->elem(8) = F(d->elem(8), s->elem(8), ((imm >> 8) & 1)); \
1580 d->elem(9) = F(d->elem(9), s->elem(9), ((imm >> 9) & 1)); \
1581 d->elem(10) = F(d->elem(10), s->elem(10), \
1582 ((imm >> 10) & 1)); \
1583 d->elem(11) = F(d->elem(11), s->elem(11), \
1584 ((imm >> 11) & 1)); \
1585 d->elem(12) = F(d->elem(12), s->elem(12), \
1586 ((imm >> 12) & 1)); \
1587 d->elem(13) = F(d->elem(13), s->elem(13), \
1588 ((imm >> 13) & 1)); \
1589 d->elem(14) = F(d->elem(14), s->elem(14), \
1590 ((imm >> 14) & 1)); \
1591 d->elem(15) = F(d->elem(15), s->elem(15), \
1592 ((imm >> 15) & 1)); \
1593 } \
1594 } \
1595 } \
1596 }
1597
1598 /* SSE4.1 op helpers */
1599 #define FBLENDVB(d, s, m) ((m & 0x80) ? s : d)
1600 #define FBLENDVPS(d, s, m) ((m & 0x80000000) ? s : d)
1601 #define FBLENDVPD(d, s, m) ((m & 0x8000000000000000LL) ? s : d)
1602 SSE_HELPER_V(helper_pblendvb, B, 16, FBLENDVB)
1603 SSE_HELPER_V(helper_blendvps, L, 4, FBLENDVPS)
1604 SSE_HELPER_V(helper_blendvpd, Q, 2, FBLENDVPD)
1605
1606 void glue(helper_ptest, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1607 {
1608 uint64_t zf = (s->Q(0) & d->Q(0)) | (s->Q(1) & d->Q(1));
1609 uint64_t cf = (s->Q(0) & ~d->Q(0)) | (s->Q(1) & ~d->Q(1));
1610
1611 CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
1612 }
1613
1614 #define SSE_HELPER_F(name, elem, num, F) \
1615 void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
1616 { \
1617 d->elem(0) = F(0); \
1618 d->elem(1) = F(1); \
1619 if (num > 2) { \
1620 d->elem(2) = F(2); \
1621 d->elem(3) = F(3); \
1622 if (num > 4) { \
1623 d->elem(4) = F(4); \
1624 d->elem(5) = F(5); \
1625 d->elem(6) = F(6); \
1626 d->elem(7) = F(7); \
1627 } \
1628 } \
1629 }
1630
1631 SSE_HELPER_F(helper_pmovsxbw, W, 8, (int8_t) s->B)
1632 SSE_HELPER_F(helper_pmovsxbd, L, 4, (int8_t) s->B)
1633 SSE_HELPER_F(helper_pmovsxbq, Q, 2, (int8_t) s->B)
1634 SSE_HELPER_F(helper_pmovsxwd, L, 4, (int16_t) s->W)
1635 SSE_HELPER_F(helper_pmovsxwq, Q, 2, (int16_t) s->W)
1636 SSE_HELPER_F(helper_pmovsxdq, Q, 2, (int32_t) s->L)
1637 SSE_HELPER_F(helper_pmovzxbw, W, 8, s->B)
1638 SSE_HELPER_F(helper_pmovzxbd, L, 4, s->B)
1639 SSE_HELPER_F(helper_pmovzxbq, Q, 2, s->B)
1640 SSE_HELPER_F(helper_pmovzxwd, L, 4, s->W)
1641 SSE_HELPER_F(helper_pmovzxwq, Q, 2, s->W)
1642 SSE_HELPER_F(helper_pmovzxdq, Q, 2, s->L)
1643
1644 void glue(helper_pmuldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1645 {
1646 d->Q(0) = (int64_t)(int32_t) d->L(0) * (int32_t) s->L(0);
1647 d->Q(1) = (int64_t)(int32_t) d->L(2) * (int32_t) s->L(2);
1648 }
1649
1650 #define FCMPEQQ(d, s) (d == s ? -1 : 0)
1651 SSE_HELPER_Q(helper_pcmpeqq, FCMPEQQ)
1652
1653 void glue(helper_packusdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1654 {
1655 d->W(0) = satuw((int32_t) d->L(0));
1656 d->W(1) = satuw((int32_t) d->L(1));
1657 d->W(2) = satuw((int32_t) d->L(2));
1658 d->W(3) = satuw((int32_t) d->L(3));
1659 d->W(4) = satuw((int32_t) s->L(0));
1660 d->W(5) = satuw((int32_t) s->L(1));
1661 d->W(6) = satuw((int32_t) s->L(2));
1662 d->W(7) = satuw((int32_t) s->L(3));
1663 }
1664
1665 #define FMINSB(d, s) MIN((int8_t)d, (int8_t)s)
1666 #define FMINSD(d, s) MIN((int32_t)d, (int32_t)s)
1667 #define FMAXSB(d, s) MAX((int8_t)d, (int8_t)s)
1668 #define FMAXSD(d, s) MAX((int32_t)d, (int32_t)s)
1669 SSE_HELPER_B(helper_pminsb, FMINSB)
1670 SSE_HELPER_L(helper_pminsd, FMINSD)
1671 SSE_HELPER_W(helper_pminuw, MIN)
1672 SSE_HELPER_L(helper_pminud, MIN)
1673 SSE_HELPER_B(helper_pmaxsb, FMAXSB)
1674 SSE_HELPER_L(helper_pmaxsd, FMAXSD)
1675 SSE_HELPER_W(helper_pmaxuw, MAX)
1676 SSE_HELPER_L(helper_pmaxud, MAX)
1677
1678 #define FMULLD(d, s) ((int32_t)d * (int32_t)s)
1679 SSE_HELPER_L(helper_pmulld, FMULLD)
1680
1681 void glue(helper_phminposuw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
1682 {
1683 int idx = 0;
1684
1685 if (s->W(1) < s->W(idx)) {
1686 idx = 1;
1687 }
1688 if (s->W(2) < s->W(idx)) {
1689 idx = 2;
1690 }
1691 if (s->W(3) < s->W(idx)) {
1692 idx = 3;
1693 }
1694 if (s->W(4) < s->W(idx)) {
1695 idx = 4;
1696 }
1697 if (s->W(5) < s->W(idx)) {
1698 idx = 5;
1699 }
1700 if (s->W(6) < s->W(idx)) {
1701 idx = 6;
1702 }
1703 if (s->W(7) < s->W(idx)) {
1704 idx = 7;
1705 }
1706
1707 d->Q(1) = 0;
1708 d->L(1) = 0;
1709 d->W(1) = idx;
1710 d->W(0) = s->W(idx);
1711 }
1712
1713 void glue(helper_roundps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1714 uint32_t mode)
1715 {
1716 signed char prev_rounding_mode;
1717
1718 prev_rounding_mode = env->sse_status.float_rounding_mode;
1719 if (!(mode & (1 << 2))) {
1720 switch (mode & 3) {
1721 case 0:
1722 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1723 break;
1724 case 1:
1725 set_float_rounding_mode(float_round_down, &env->sse_status);
1726 break;
1727 case 2:
1728 set_float_rounding_mode(float_round_up, &env->sse_status);
1729 break;
1730 case 3:
1731 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1732 break;
1733 }
1734 }
1735
1736 d->XMM_S(0) = float32_round_to_int(s->XMM_S(0), &env->sse_status);
1737 d->XMM_S(1) = float32_round_to_int(s->XMM_S(1), &env->sse_status);
1738 d->XMM_S(2) = float32_round_to_int(s->XMM_S(2), &env->sse_status);
1739 d->XMM_S(3) = float32_round_to_int(s->XMM_S(3), &env->sse_status);
1740
1741 #if 0 /* TODO */
1742 if (mode & (1 << 3)) {
1743 set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1744 ~float_flag_inexact,
1745 &env->sse_status);
1746 }
1747 #endif
1748 env->sse_status.float_rounding_mode = prev_rounding_mode;
1749 }
1750
1751 void glue(helper_roundpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1752 uint32_t mode)
1753 {
1754 signed char prev_rounding_mode;
1755
1756 prev_rounding_mode = env->sse_status.float_rounding_mode;
1757 if (!(mode & (1 << 2))) {
1758 switch (mode & 3) {
1759 case 0:
1760 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1761 break;
1762 case 1:
1763 set_float_rounding_mode(float_round_down, &env->sse_status);
1764 break;
1765 case 2:
1766 set_float_rounding_mode(float_round_up, &env->sse_status);
1767 break;
1768 case 3:
1769 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1770 break;
1771 }
1772 }
1773
1774 d->XMM_D(0) = float64_round_to_int(s->XMM_D(0), &env->sse_status);
1775 d->XMM_D(1) = float64_round_to_int(s->XMM_D(1), &env->sse_status);
1776
1777 #if 0 /* TODO */
1778 if (mode & (1 << 3)) {
1779 set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1780 ~float_flag_inexact,
1781 &env->sse_status);
1782 }
1783 #endif
1784 env->sse_status.float_rounding_mode = prev_rounding_mode;
1785 }
1786
1787 void glue(helper_roundss, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1788 uint32_t mode)
1789 {
1790 signed char prev_rounding_mode;
1791
1792 prev_rounding_mode = env->sse_status.float_rounding_mode;
1793 if (!(mode & (1 << 2))) {
1794 switch (mode & 3) {
1795 case 0:
1796 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1797 break;
1798 case 1:
1799 set_float_rounding_mode(float_round_down, &env->sse_status);
1800 break;
1801 case 2:
1802 set_float_rounding_mode(float_round_up, &env->sse_status);
1803 break;
1804 case 3:
1805 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1806 break;
1807 }
1808 }
1809
1810 d->XMM_S(0) = float32_round_to_int(s->XMM_S(0), &env->sse_status);
1811
1812 #if 0 /* TODO */
1813 if (mode & (1 << 3)) {
1814 set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1815 ~float_flag_inexact,
1816 &env->sse_status);
1817 }
1818 #endif
1819 env->sse_status.float_rounding_mode = prev_rounding_mode;
1820 }
1821
1822 void glue(helper_roundsd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1823 uint32_t mode)
1824 {
1825 signed char prev_rounding_mode;
1826
1827 prev_rounding_mode = env->sse_status.float_rounding_mode;
1828 if (!(mode & (1 << 2))) {
1829 switch (mode & 3) {
1830 case 0:
1831 set_float_rounding_mode(float_round_nearest_even, &env->sse_status);
1832 break;
1833 case 1:
1834 set_float_rounding_mode(float_round_down, &env->sse_status);
1835 break;
1836 case 2:
1837 set_float_rounding_mode(float_round_up, &env->sse_status);
1838 break;
1839 case 3:
1840 set_float_rounding_mode(float_round_to_zero, &env->sse_status);
1841 break;
1842 }
1843 }
1844
1845 d->XMM_D(0) = float64_round_to_int(s->XMM_D(0), &env->sse_status);
1846
1847 #if 0 /* TODO */
1848 if (mode & (1 << 3)) {
1849 set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
1850 ~float_flag_inexact,
1851 &env->sse_status);
1852 }
1853 #endif
1854 env->sse_status.float_rounding_mode = prev_rounding_mode;
1855 }
1856
1857 #define FBLENDP(d, s, m) (m ? s : d)
1858 SSE_HELPER_I(helper_blendps, L, 4, FBLENDP)
1859 SSE_HELPER_I(helper_blendpd, Q, 2, FBLENDP)
1860 SSE_HELPER_I(helper_pblendw, W, 8, FBLENDP)
1861
1862 void glue(helper_dpps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1863 {
1864 float32 iresult = float32_zero;
1865
1866 if (mask & (1 << 4)) {
1867 iresult = float32_add(iresult,
1868 float32_mul(d->XMM_S(0), s->XMM_S(0),
1869 &env->sse_status),
1870 &env->sse_status);
1871 }
1872 if (mask & (1 << 5)) {
1873 iresult = float32_add(iresult,
1874 float32_mul(d->XMM_S(1), s->XMM_S(1),
1875 &env->sse_status),
1876 &env->sse_status);
1877 }
1878 if (mask & (1 << 6)) {
1879 iresult = float32_add(iresult,
1880 float32_mul(d->XMM_S(2), s->XMM_S(2),
1881 &env->sse_status),
1882 &env->sse_status);
1883 }
1884 if (mask & (1 << 7)) {
1885 iresult = float32_add(iresult,
1886 float32_mul(d->XMM_S(3), s->XMM_S(3),
1887 &env->sse_status),
1888 &env->sse_status);
1889 }
1890 d->XMM_S(0) = (mask & (1 << 0)) ? iresult : float32_zero;
1891 d->XMM_S(1) = (mask & (1 << 1)) ? iresult : float32_zero;
1892 d->XMM_S(2) = (mask & (1 << 2)) ? iresult : float32_zero;
1893 d->XMM_S(3) = (mask & (1 << 3)) ? iresult : float32_zero;
1894 }
1895
1896 void glue(helper_dppd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, uint32_t mask)
1897 {
1898 float64 iresult = float64_zero;
1899
1900 if (mask & (1 << 4)) {
1901 iresult = float64_add(iresult,
1902 float64_mul(d->XMM_D(0), s->XMM_D(0),
1903 &env->sse_status),
1904 &env->sse_status);
1905 }
1906 if (mask & (1 << 5)) {
1907 iresult = float64_add(iresult,
1908 float64_mul(d->XMM_D(1), s->XMM_D(1),
1909 &env->sse_status),
1910 &env->sse_status);
1911 }
1912 d->XMM_D(0) = (mask & (1 << 0)) ? iresult : float64_zero;
1913 d->XMM_D(1) = (mask & (1 << 1)) ? iresult : float64_zero;
1914 }
1915
1916 void glue(helper_mpsadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
1917 uint32_t offset)
1918 {
1919 int s0 = (offset & 3) << 2;
1920 int d0 = (offset & 4) << 0;
1921 int i;
1922 Reg r;
1923
1924 for (i = 0; i < 8; i++, d0++) {
1925 r.W(i) = 0;
1926 r.W(i) += abs1(d->B(d0 + 0) - s->B(s0 + 0));
1927 r.W(i) += abs1(d->B(d0 + 1) - s->B(s0 + 1));
1928 r.W(i) += abs1(d->B(d0 + 2) - s->B(s0 + 2));
1929 r.W(i) += abs1(d->B(d0 + 3) - s->B(s0 + 3));
1930 }
1931
1932 *d = r;
1933 }
1934
1935 /* SSE4.2 op helpers */
1936 #define FCMPGTQ(d, s) ((int64_t)d > (int64_t)s ? -1 : 0)
1937 SSE_HELPER_Q(helper_pcmpgtq, FCMPGTQ)
1938
1939 static inline int pcmp_elen(CPUX86State *env, int reg, uint32_t ctrl)
1940 {
1941 int val;
1942
1943 /* Presence of REX.W is indicated by a bit higher than 7 set */
1944 if (ctrl >> 8) {
1945 val = abs1((int64_t)env->regs[reg]);
1946 } else {
1947 val = abs1((int32_t)env->regs[reg]);
1948 }
1949
1950 if (ctrl & 1) {
1951 if (val > 8) {
1952 return 8;
1953 }
1954 } else {
1955 if (val > 16) {
1956 return 16;
1957 }
1958 }
1959 return val;
1960 }
1961
1962 static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
1963 {
1964 int val = 0;
1965
1966 if (ctrl & 1) {
1967 while (val < 8 && r->W(val)) {
1968 val++;
1969 }
1970 } else {
1971 while (val < 16 && r->B(val)) {
1972 val++;
1973 }
1974 }
1975
1976 return val;
1977 }
1978
1979 static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
1980 {
1981 switch ((ctrl >> 0) & 3) {
1982 case 0:
1983 return r->B(i);
1984 case 1:
1985 return r->W(i);
1986 case 2:
1987 return (int8_t)r->B(i);
1988 case 3:
1989 default:
1990 return (int16_t)r->W(i);
1991 }
1992 }
1993
1994 static inline unsigned pcmpxstrx(CPUX86State *env, Reg *d, Reg *s,
1995 int8_t ctrl, int valids, int validd)
1996 {
1997 unsigned int res = 0;
1998 int v;
1999 int j, i;
2000 int upper = (ctrl & 1) ? 7 : 15;
2001
2002 valids--;
2003 validd--;
2004
2005 CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
2006
2007 switch ((ctrl >> 2) & 3) {
2008 case 0:
2009 for (j = valids; j >= 0; j--) {
2010 res <<= 1;
2011 v = pcmp_val(s, ctrl, j);
2012 for (i = validd; i >= 0; i--) {
2013 res |= (v == pcmp_val(d, ctrl, i));
2014 }
2015 }
2016 break;
2017 case 1:
2018 for (j = valids; j >= 0; j--) {
2019 res <<= 1;
2020 v = pcmp_val(s, ctrl, j);
2021 for (i = ((validd - 1) | 1); i >= 0; i -= 2) {
2022 res |= (pcmp_val(d, ctrl, i - 0) >= v &&
2023 pcmp_val(d, ctrl, i - 1) <= v);
2024 }
2025 }
2026 break;
2027 case 2:
2028 res = (1 << (upper - MAX(valids, validd))) - 1;
2029 res <<= MAX(valids, validd) - MIN(valids, validd);
2030 for (i = MIN(valids, validd); i >= 0; i--) {
2031 res <<= 1;
2032 v = pcmp_val(s, ctrl, i);
2033 res |= (v == pcmp_val(d, ctrl, i));
2034 }
2035 break;
2036 case 3:
2037 for (j = valids - validd; j >= 0; j--) {
2038 res <<= 1;
2039 v = 1;
2040 for (i = MIN(upper - j, validd); i >= 0; i--) {
2041 v &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
2042 }
2043 res |= v;
2044 }
2045 break;
2046 }
2047
2048 switch ((ctrl >> 4) & 3) {
2049 case 1:
2050 res ^= (2 << upper) - 1;
2051 break;
2052 case 3:
2053 res ^= (1 << (valids + 1)) - 1;
2054 break;
2055 }
2056
2057 if (res) {
2058 CC_SRC |= CC_C;
2059 }
2060 if (res & 1) {
2061 CC_SRC |= CC_O;
2062 }
2063
2064 return res;
2065 }
2066
2067 void glue(helper_pcmpestri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2068 uint32_t ctrl)
2069 {
2070 unsigned int res = pcmpxstrx(env, d, s, ctrl,
2071 pcmp_elen(env, R_EDX, ctrl),
2072 pcmp_elen(env, R_EAX, ctrl));
2073
2074 if (res) {
2075 env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2076 } else {
2077 env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2078 }
2079 }
2080
2081 void glue(helper_pcmpestrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2082 uint32_t ctrl)
2083 {
2084 int i;
2085 unsigned int res = pcmpxstrx(env, d, s, ctrl,
2086 pcmp_elen(env, R_EDX, ctrl),
2087 pcmp_elen(env, R_EAX, ctrl));
2088
2089 if ((ctrl >> 6) & 1) {
2090 if (ctrl & 1) {
2091 for (i = 0; i < 8; i++, res >>= 1) {
2092 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2093 }
2094 } else {
2095 for (i = 0; i < 16; i++, res >>= 1) {
2096 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2097 }
2098 }
2099 } else {
2100 env->xmm_regs[0].Q(1) = 0;
2101 env->xmm_regs[0].Q(0) = res;
2102 }
2103 }
2104
2105 void glue(helper_pcmpistri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2106 uint32_t ctrl)
2107 {
2108 unsigned int res = pcmpxstrx(env, d, s, ctrl,
2109 pcmp_ilen(s, ctrl),
2110 pcmp_ilen(d, ctrl));
2111
2112 if (res) {
2113 env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
2114 } else {
2115 env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
2116 }
2117 }
2118
2119 void glue(helper_pcmpistrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2120 uint32_t ctrl)
2121 {
2122 int i;
2123 unsigned int res = pcmpxstrx(env, d, s, ctrl,
2124 pcmp_ilen(s, ctrl),
2125 pcmp_ilen(d, ctrl));
2126
2127 if ((ctrl >> 6) & 1) {
2128 if (ctrl & 1) {
2129 for (i = 0; i < 8; i++, res >>= 1) {
2130 env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
2131 }
2132 } else {
2133 for (i = 0; i < 16; i++, res >>= 1) {
2134 env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
2135 }
2136 }
2137 } else {
2138 env->xmm_regs[0].Q(1) = 0;
2139 env->xmm_regs[0].Q(0) = res;
2140 }
2141 }
2142
2143 #define CRCPOLY 0x1edc6f41
2144 #define CRCPOLY_BITREV 0x82f63b78
2145 target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
2146 {
2147 target_ulong crc = (msg & ((target_ulong) -1 >>
2148 (TARGET_LONG_BITS - len))) ^ crc1;
2149
2150 while (len--) {
2151 crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
2152 }
2153
2154 return crc;
2155 }
2156
2157 #define POPMASK(i) ((target_ulong) -1 / ((1LL << (1 << i)) + 1))
2158 #define POPCOUNT(n, i) ((n & POPMASK(i)) + ((n >> (1 << i)) & POPMASK(i)))
2159 target_ulong helper_popcnt(CPUX86State *env, target_ulong n, uint32_t type)
2160 {
2161 CC_SRC = n ? 0 : CC_Z;
2162
2163 n = POPCOUNT(n, 0);
2164 n = POPCOUNT(n, 1);
2165 n = POPCOUNT(n, 2);
2166 n = POPCOUNT(n, 3);
2167 if (type == 1) {
2168 return n & 0xff;
2169 }
2170
2171 n = POPCOUNT(n, 4);
2172 #ifndef TARGET_X86_64
2173 return n;
2174 #else
2175 if (type == 2) {
2176 return n & 0xff;
2177 }
2178
2179 return POPCOUNT(n, 5);
2180 #endif
2181 }
2182
2183 void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2184 uint32_t ctrl)
2185 {
2186 uint64_t ah, al, b, resh, resl;
2187
2188 ah = 0;
2189 al = d->Q((ctrl & 1) != 0);
2190 b = s->Q((ctrl & 16) != 0);
2191 resh = resl = 0;
2192
2193 while (b) {
2194 if (b & 1) {
2195 resl ^= al;
2196 resh ^= ah;
2197 }
2198 ah = (ah << 1) | (al >> 63);
2199 al <<= 1;
2200 b >>= 1;
2201 }
2202
2203 d->Q(0) = resl;
2204 d->Q(1) = resh;
2205 }
2206
2207 /* AES-NI op helpers */
2208 static const uint8_t aes_shifts[16] = {
2209 0, 5, 10, 15, 4, 9, 14, 3, 8, 13, 2, 7, 12, 1, 6, 11
2210 };
2211
2212 static const uint8_t aes_ishifts[16] = {
2213 0, 13, 10, 7, 4, 1, 14, 11, 8, 5, 2, 15, 12, 9, 6, 3
2214 };
2215
2216 void glue(helper_aesdec, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2217 {
2218 int i;
2219 Reg st = *d;
2220 Reg rk = *s;
2221
2222 for (i = 0 ; i < 4 ; i++) {
2223 d->L(i) = rk.L(i) ^ bswap32(AES_Td0[st.B(aes_ishifts[4*i+0])] ^
2224 AES_Td1[st.B(aes_ishifts[4*i+1])] ^
2225 AES_Td2[st.B(aes_ishifts[4*i+2])] ^
2226 AES_Td3[st.B(aes_ishifts[4*i+3])]);
2227 }
2228 }
2229
2230 void glue(helper_aesdeclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2231 {
2232 int i;
2233 Reg st = *d;
2234 Reg rk = *s;
2235
2236 for (i = 0; i < 16; i++) {
2237 d->B(i) = rk.B(i) ^ (AES_Td4[st.B(aes_ishifts[i])] & 0xff);
2238 }
2239 }
2240
2241 void glue(helper_aesenc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2242 {
2243 int i;
2244 Reg st = *d;
2245 Reg rk = *s;
2246
2247 for (i = 0 ; i < 4 ; i++) {
2248 d->L(i) = rk.L(i) ^ bswap32(AES_Te0[st.B(aes_shifts[4*i+0])] ^
2249 AES_Te1[st.B(aes_shifts[4*i+1])] ^
2250 AES_Te2[st.B(aes_shifts[4*i+2])] ^
2251 AES_Te3[st.B(aes_shifts[4*i+3])]);
2252 }
2253 }
2254
2255 void glue(helper_aesenclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2256 {
2257 int i;
2258 Reg st = *d;
2259 Reg rk = *s;
2260
2261 for (i = 0; i < 16; i++) {
2262 d->B(i) = rk.B(i) ^ (AES_Te4[st.B(aes_shifts[i])] & 0xff);
2263 }
2264
2265 }
2266
2267 void glue(helper_aesimc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
2268 {
2269 int i;
2270 Reg tmp = *s;
2271
2272 for (i = 0 ; i < 4 ; i++) {
2273 d->L(i) = bswap32(AES_Td0[AES_Te4[tmp.B(4*i+0)] & 0xff] ^
2274 AES_Td1[AES_Te4[tmp.B(4*i+1)] & 0xff] ^
2275 AES_Td2[AES_Te4[tmp.B(4*i+2)] & 0xff] ^
2276 AES_Td3[AES_Te4[tmp.B(4*i+3)] & 0xff]);
2277 }
2278 }
2279
2280 void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
2281 uint32_t ctrl)
2282 {
2283 int i;
2284 Reg tmp = *s;
2285
2286 for (i = 0 ; i < 4 ; i++) {
2287 d->B(i) = AES_Te4[tmp.B(i + 4)] & 0xff;
2288 d->B(i + 8) = AES_Te4[tmp.B(i + 12)] & 0xff;
2289 }
2290 d->L(1) = (d->L(0) << 24 | d->L(0) >> 8) ^ ctrl;
2291 d->L(3) = (d->L(2) << 24 | d->L(2) >> 8) ^ ctrl;
2292 }
2293 #endif
2294
2295 #undef SHIFT
2296 #undef XMM_ONLY
2297 #undef Reg
2298 #undef B
2299 #undef W
2300 #undef L
2301 #undef Q
2302 #undef SUFFIX
QEMU modified to work with kvm