35 |
35 |
#define SUFFIX _xmm
|
36 |
36 |
#endif
|
37 |
37 |
|
38 |
|
void OPPROTO glue(op_psrlw, SUFFIX)(void)
|
|
38 |
void glue(helper_psrlw, SUFFIX)(Reg *d, Reg *s)
|
39 |
39 |
{
|
40 |
|
Reg *d, *s;
|
41 |
40 |
int shift;
|
42 |
41 |
|
43 |
|
d = (Reg *)((char *)env + PARAM1);
|
44 |
|
s = (Reg *)((char *)env + PARAM2);
|
45 |
|
|
46 |
42 |
if (s->Q(0) > 15) {
|
47 |
43 |
d->Q(0) = 0;
|
48 |
44 |
#if SHIFT == 1
|
... | ... | |
64 |
60 |
FORCE_RET();
|
65 |
61 |
}
|
66 |
62 |
|
67 |
|
void OPPROTO glue(op_psraw, SUFFIX)(void)
|
|
63 |
void glue(helper_psraw, SUFFIX)(Reg *d, Reg *s)
|
68 |
64 |
{
|
69 |
|
Reg *d, *s;
|
70 |
65 |
int shift;
|
71 |
66 |
|
72 |
|
d = (Reg *)((char *)env + PARAM1);
|
73 |
|
s = (Reg *)((char *)env + PARAM2);
|
74 |
|
|
75 |
67 |
if (s->Q(0) > 15) {
|
76 |
68 |
shift = 15;
|
77 |
69 |
} else {
|
... | ... | |
89 |
81 |
#endif
|
90 |
82 |
}
|
91 |
83 |
|
92 |
|
void OPPROTO glue(op_psllw, SUFFIX)(void)
|
|
84 |
void glue(helper_psllw, SUFFIX)(Reg *d, Reg *s)
|
93 |
85 |
{
|
94 |
|
Reg *d, *s;
|
95 |
86 |
int shift;
|
96 |
87 |
|
97 |
|
d = (Reg *)((char *)env + PARAM1);
|
98 |
|
s = (Reg *)((char *)env + PARAM2);
|
99 |
|
|
100 |
88 |
if (s->Q(0) > 15) {
|
101 |
89 |
d->Q(0) = 0;
|
102 |
90 |
#if SHIFT == 1
|
... | ... | |
118 |
106 |
FORCE_RET();
|
119 |
107 |
}
|
120 |
108 |
|
121 |
|
void OPPROTO glue(op_psrld, SUFFIX)(void)
|
|
109 |
void glue(helper_psrld, SUFFIX)(Reg *d, Reg *s)
|
122 |
110 |
{
|
123 |
|
Reg *d, *s;
|
124 |
111 |
int shift;
|
125 |
112 |
|
126 |
|
d = (Reg *)((char *)env + PARAM1);
|
127 |
|
s = (Reg *)((char *)env + PARAM2);
|
128 |
|
|
129 |
113 |
if (s->Q(0) > 31) {
|
130 |
114 |
d->Q(0) = 0;
|
131 |
115 |
#if SHIFT == 1
|
... | ... | |
143 |
127 |
FORCE_RET();
|
144 |
128 |
}
|
145 |
129 |
|
146 |
|
void OPPROTO glue(op_psrad, SUFFIX)(void)
|
|
130 |
void glue(helper_psrad, SUFFIX)(Reg *d, Reg *s)
|
147 |
131 |
{
|
148 |
|
Reg *d, *s;
|
149 |
132 |
int shift;
|
150 |
133 |
|
151 |
|
d = (Reg *)((char *)env + PARAM1);
|
152 |
|
s = (Reg *)((char *)env + PARAM2);
|
153 |
|
|
154 |
134 |
if (s->Q(0) > 31) {
|
155 |
135 |
shift = 31;
|
156 |
136 |
} else {
|
... | ... | |
164 |
144 |
#endif
|
165 |
145 |
}
|
166 |
146 |
|
167 |
|
void OPPROTO glue(op_pslld, SUFFIX)(void)
|
|
147 |
void glue(helper_pslld, SUFFIX)(Reg *d, Reg *s)
|
168 |
148 |
{
|
169 |
|
Reg *d, *s;
|
170 |
149 |
int shift;
|
171 |
150 |
|
172 |
|
d = (Reg *)((char *)env + PARAM1);
|
173 |
|
s = (Reg *)((char *)env + PARAM2);
|
174 |
|
|
175 |
151 |
if (s->Q(0) > 31) {
|
176 |
152 |
d->Q(0) = 0;
|
177 |
153 |
#if SHIFT == 1
|
... | ... | |
189 |
165 |
FORCE_RET();
|
190 |
166 |
}
|
191 |
167 |
|
192 |
|
void OPPROTO glue(op_psrlq, SUFFIX)(void)
|
|
168 |
void glue(helper_psrlq, SUFFIX)(Reg *d, Reg *s)
|
193 |
169 |
{
|
194 |
|
Reg *d, *s;
|
195 |
170 |
int shift;
|
196 |
171 |
|
197 |
|
d = (Reg *)((char *)env + PARAM1);
|
198 |
|
s = (Reg *)((char *)env + PARAM2);
|
199 |
|
|
200 |
172 |
if (s->Q(0) > 63) {
|
201 |
173 |
d->Q(0) = 0;
|
202 |
174 |
#if SHIFT == 1
|
... | ... | |
212 |
184 |
FORCE_RET();
|
213 |
185 |
}
|
214 |
186 |
|
215 |
|
void OPPROTO glue(op_psllq, SUFFIX)(void)
|
|
187 |
void glue(helper_psllq, SUFFIX)(Reg *d, Reg *s)
|
216 |
188 |
{
|
217 |
|
Reg *d, *s;
|
218 |
189 |
int shift;
|
219 |
190 |
|
220 |
|
d = (Reg *)((char *)env + PARAM1);
|
221 |
|
s = (Reg *)((char *)env + PARAM2);
|
222 |
|
|
223 |
191 |
if (s->Q(0) > 63) {
|
224 |
192 |
d->Q(0) = 0;
|
225 |
193 |
#if SHIFT == 1
|
... | ... | |
236 |
204 |
}
|
237 |
205 |
|
238 |
206 |
#if SHIFT == 1
|
239 |
|
void OPPROTO glue(op_psrldq, SUFFIX)(void)
|
|
207 |
void glue(helper_psrldq, SUFFIX)(Reg *d, Reg *s)
|
240 |
208 |
{
|
241 |
|
Reg *d, *s;
|
242 |
209 |
int shift, i;
|
243 |
210 |
|
244 |
|
d = (Reg *)((char *)env + PARAM1);
|
245 |
|
s = (Reg *)((char *)env + PARAM2);
|
246 |
211 |
shift = s->L(0);
|
247 |
212 |
if (shift > 16)
|
248 |
213 |
shift = 16;
|
... | ... | |
253 |
218 |
FORCE_RET();
|
254 |
219 |
}
|
255 |
220 |
|
256 |
|
void OPPROTO glue(op_pslldq, SUFFIX)(void)
|
|
221 |
void glue(helper_pslldq, SUFFIX)(Reg *d, Reg *s)
|
257 |
222 |
{
|
258 |
|
Reg *d, *s;
|
259 |
223 |
int shift, i;
|
260 |
224 |
|
261 |
|
d = (Reg *)((char *)env + PARAM1);
|
262 |
|
s = (Reg *)((char *)env + PARAM2);
|
263 |
225 |
shift = s->L(0);
|
264 |
226 |
if (shift > 16)
|
265 |
227 |
shift = 16;
|
... | ... | |
271 |
233 |
}
|
272 |
234 |
#endif
|
273 |
235 |
|
274 |
|
#define SSE_OP_B(name, F)\
|
275 |
|
void OPPROTO glue(name, SUFFIX) (void)\
|
|
236 |
#define SSE_HELPER_B(name, F)\
|
|
237 |
void glue(name, SUFFIX) (Reg *d, Reg *s)\
|
276 |
238 |
{\
|
277 |
|
Reg *d, *s;\
|
278 |
|
d = (Reg *)((char *)env + PARAM1);\
|
279 |
|
s = (Reg *)((char *)env + PARAM2);\
|
280 |
239 |
d->B(0) = F(d->B(0), s->B(0));\
|
281 |
240 |
d->B(1) = F(d->B(1), s->B(1));\
|
282 |
241 |
d->B(2) = F(d->B(2), s->B(2));\
|
... | ... | |
297 |
256 |
)\
|
298 |
257 |
}
|
299 |
258 |
|
300 |
|
#define SSE_OP_W(name, F)\
|
301 |
|
void OPPROTO glue(name, SUFFIX) (void)\
|
|
259 |
#define SSE_HELPER_W(name, F)\
|
|
260 |
void glue(name, SUFFIX) (Reg *d, Reg *s)\
|
302 |
261 |
{\
|
303 |
|
Reg *d, *s;\
|
304 |
|
d = (Reg *)((char *)env + PARAM1);\
|
305 |
|
s = (Reg *)((char *)env + PARAM2);\
|
306 |
262 |
d->W(0) = F(d->W(0), s->W(0));\
|
307 |
263 |
d->W(1) = F(d->W(1), s->W(1));\
|
308 |
264 |
d->W(2) = F(d->W(2), s->W(2));\
|
... | ... | |
315 |
271 |
)\
|
316 |
272 |
}
|
317 |
273 |
|
318 |
|
#define SSE_OP_L(name, F)\
|
319 |
|
void OPPROTO glue(name, SUFFIX) (void)\
|
|
274 |
#define SSE_HELPER_L(name, F)\
|
|
275 |
void glue(name, SUFFIX) (Reg *d, Reg *s)\
|
320 |
276 |
{\
|
321 |
|
Reg *d, *s;\
|
322 |
|
d = (Reg *)((char *)env + PARAM1);\
|
323 |
|
s = (Reg *)((char *)env + PARAM2);\
|
324 |
277 |
d->L(0) = F(d->L(0), s->L(0));\
|
325 |
278 |
d->L(1) = F(d->L(1), s->L(1));\
|
326 |
279 |
XMM_ONLY(\
|
... | ... | |
329 |
282 |
)\
|
330 |
283 |
}
|
331 |
284 |
|
332 |
|
#define SSE_OP_Q(name, F)\
|
333 |
|
void OPPROTO glue(name, SUFFIX) (void)\
|
|
285 |
#define SSE_HELPER_Q(name, F)\
|
|
286 |
void glue(name, SUFFIX) (Reg *d, Reg *s)\
|
334 |
287 |
{\
|
335 |
|
Reg *d, *s;\
|
336 |
|
d = (Reg *)((char *)env + PARAM1);\
|
337 |
|
s = (Reg *)((char *)env + PARAM2);\
|
338 |
288 |
d->Q(0) = F(d->Q(0), s->Q(0));\
|
339 |
289 |
XMM_ONLY(\
|
340 |
290 |
d->Q(1) = F(d->Q(1), s->Q(1));\
|
... | ... | |
416 |
366 |
#define FAVG(a, b) ((a) + (b) + 1) >> 1
|
417 |
367 |
#endif
|
418 |
368 |
|
419 |
|
SSE_OP_B(op_paddb, FADD)
|
420 |
|
SSE_OP_W(op_paddw, FADD)
|
421 |
|
SSE_OP_L(op_paddl, FADD)
|
422 |
|
SSE_OP_Q(op_paddq, FADD)
|
|
369 |
SSE_HELPER_B(helper_paddb, FADD)
|
|
370 |
SSE_HELPER_W(helper_paddw, FADD)
|
|
371 |
SSE_HELPER_L(helper_paddl, FADD)
|
|
372 |
SSE_HELPER_Q(helper_paddq, FADD)
|
423 |
373 |
|
424 |
|
SSE_OP_B(op_psubb, FSUB)
|
425 |
|
SSE_OP_W(op_psubw, FSUB)
|
426 |
|
SSE_OP_L(op_psubl, FSUB)
|
427 |
|
SSE_OP_Q(op_psubq, FSUB)
|
|
374 |
SSE_HELPER_B(helper_psubb, FSUB)
|
|
375 |
SSE_HELPER_W(helper_psubw, FSUB)
|
|
376 |
SSE_HELPER_L(helper_psubl, FSUB)
|
|
377 |
SSE_HELPER_Q(helper_psubq, FSUB)
|
428 |
378 |
|
429 |
|
SSE_OP_B(op_paddusb, FADDUB)
|
430 |
|
SSE_OP_B(op_paddsb, FADDSB)
|
431 |
|
SSE_OP_B(op_psubusb, FSUBUB)
|
432 |
|
SSE_OP_B(op_psubsb, FSUBSB)
|
|
379 |
SSE_HELPER_B(helper_paddusb, FADDUB)
|
|
380 |
SSE_HELPER_B(helper_paddsb, FADDSB)
|
|
381 |
SSE_HELPER_B(helper_psubusb, FSUBUB)
|
|
382 |
SSE_HELPER_B(helper_psubsb, FSUBSB)
|
433 |
383 |
|
434 |
|
SSE_OP_W(op_paddusw, FADDUW)
|
435 |
|
SSE_OP_W(op_paddsw, FADDSW)
|
436 |
|
SSE_OP_W(op_psubusw, FSUBUW)
|
437 |
|
SSE_OP_W(op_psubsw, FSUBSW)
|
|
384 |
SSE_HELPER_W(helper_paddusw, FADDUW)
|
|
385 |
SSE_HELPER_W(helper_paddsw, FADDSW)
|
|
386 |
SSE_HELPER_W(helper_psubusw, FSUBUW)
|
|
387 |
SSE_HELPER_W(helper_psubsw, FSUBSW)
|
438 |
388 |
|
439 |
|
SSE_OP_B(op_pminub, FMINUB)
|
440 |
|
SSE_OP_B(op_pmaxub, FMAXUB)
|
|
389 |
SSE_HELPER_B(helper_pminub, FMINUB)
|
|
390 |
SSE_HELPER_B(helper_pmaxub, FMAXUB)
|
441 |
391 |
|
442 |
|
SSE_OP_W(op_pminsw, FMINSW)
|
443 |
|
SSE_OP_W(op_pmaxsw, FMAXSW)
|
|
392 |
SSE_HELPER_W(helper_pminsw, FMINSW)
|
|
393 |
SSE_HELPER_W(helper_pmaxsw, FMAXSW)
|
444 |
394 |
|
445 |
|
SSE_OP_Q(op_pand, FAND)
|
446 |
|
SSE_OP_Q(op_pandn, FANDN)
|
447 |
|
SSE_OP_Q(op_por, FOR)
|
448 |
|
SSE_OP_Q(op_pxor, FXOR)
|
|
395 |
SSE_HELPER_Q(helper_pand, FAND)
|
|
396 |
SSE_HELPER_Q(helper_pandn, FANDN)
|
|
397 |
SSE_HELPER_Q(helper_por, FOR)
|
|
398 |
SSE_HELPER_Q(helper_pxor, FXOR)
|
449 |
399 |
|
450 |
|
SSE_OP_B(op_pcmpgtb, FCMPGTB)
|
451 |
|
SSE_OP_W(op_pcmpgtw, FCMPGTW)
|
452 |
|
SSE_OP_L(op_pcmpgtl, FCMPGTL)
|
|
400 |
SSE_HELPER_B(helper_pcmpgtb, FCMPGTB)
|
|
401 |
SSE_HELPER_W(helper_pcmpgtw, FCMPGTW)
|
|
402 |
SSE_HELPER_L(helper_pcmpgtl, FCMPGTL)
|
453 |
403 |
|
454 |
|
SSE_OP_B(op_pcmpeqb, FCMPEQ)
|
455 |
|
SSE_OP_W(op_pcmpeqw, FCMPEQ)
|
456 |
|
SSE_OP_L(op_pcmpeql, FCMPEQ)
|
|
404 |
SSE_HELPER_B(helper_pcmpeqb, FCMPEQ)
|
|
405 |
SSE_HELPER_W(helper_pcmpeqw, FCMPEQ)
|
|
406 |
SSE_HELPER_L(helper_pcmpeql, FCMPEQ)
|
457 |
407 |
|
458 |
|
SSE_OP_W(op_pmullw, FMULLW)
|
|
408 |
SSE_HELPER_W(helper_pmullw, FMULLW)
|
459 |
409 |
#if SHIFT == 0
|
460 |
|
SSE_OP_W(op_pmulhrw, FMULHRW)
|
|
410 |
SSE_HELPER_W(helper_pmulhrw, FMULHRW)
|
461 |
411 |
#endif
|
462 |
|
SSE_OP_W(op_pmulhuw, FMULHUW)
|
463 |
|
SSE_OP_W(op_pmulhw, FMULHW)
|
|
412 |
SSE_HELPER_W(helper_pmulhuw, FMULHUW)
|
|
413 |
SSE_HELPER_W(helper_pmulhw, FMULHW)
|
464 |
414 |
|
465 |
|
SSE_OP_B(op_pavgb, FAVG)
|
466 |
|
SSE_OP_W(op_pavgw, FAVG)
|
|
415 |
SSE_HELPER_B(helper_pavgb, FAVG)
|
|
416 |
SSE_HELPER_W(helper_pavgw, FAVG)
|
467 |
417 |
|
468 |
|
void OPPROTO glue(op_pmuludq, SUFFIX) (void)
|
|
418 |
void glue(helper_pmuludq, SUFFIX) (Reg *d, Reg *s)
|
469 |
419 |
{
|
470 |
|
Reg *d, *s;
|
471 |
|
d = (Reg *)((char *)env + PARAM1);
|
472 |
|
s = (Reg *)((char *)env + PARAM2);
|
473 |
|
|
474 |
420 |
d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
|
475 |
421 |
#if SHIFT == 1
|
476 |
422 |
d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
|
477 |
423 |
#endif
|
478 |
424 |
}
|
479 |
425 |
|
480 |
|
void OPPROTO glue(op_pmaddwd, SUFFIX) (void)
|
|
426 |
void glue(helper_pmaddwd, SUFFIX) (Reg *d, Reg *s)
|
481 |
427 |
{
|
482 |
428 |
int i;
|
483 |
|
Reg *d, *s;
|
484 |
|
d = (Reg *)((char *)env + PARAM1);
|
485 |
|
s = (Reg *)((char *)env + PARAM2);
|
486 |
429 |
|
487 |
430 |
for(i = 0; i < (2 << SHIFT); i++) {
|
488 |
431 |
d->L(i) = (int16_t)s->W(2*i) * (int16_t)d->W(2*i) +
|
... | ... | |
500 |
443 |
return a;
|
501 |
444 |
}
|
502 |
445 |
#endif
|
503 |
|
void OPPROTO glue(op_psadbw, SUFFIX) (void)
|
|
446 |
void glue(helper_psadbw, SUFFIX) (Reg *d, Reg *s)
|
504 |
447 |
{
|
505 |
448 |
unsigned int val;
|
506 |
|
Reg *d, *s;
|
507 |
|
d = (Reg *)((char *)env + PARAM1);
|
508 |
|
s = (Reg *)((char *)env + PARAM2);
|
509 |
449 |
|
510 |
450 |
val = 0;
|
511 |
451 |
val += abs1(d->B(0) - s->B(0));
|
... | ... | |
531 |
471 |
#endif
|
532 |
472 |
}
|
533 |
473 |
|
534 |
|
void OPPROTO glue(op_maskmov, SUFFIX) (void)
|
|
474 |
void glue(helper_maskmov, SUFFIX) (Reg *d, Reg *s)
|
535 |
475 |
{
|
536 |
476 |
int i;
|
537 |
|
Reg *d, *s;
|
538 |
|
d = (Reg *)((char *)env + PARAM1);
|
539 |
|
s = (Reg *)((char *)env + PARAM2);
|
540 |
477 |
for(i = 0; i < (8 << SHIFT); i++) {
|
541 |
478 |
if (s->B(i) & 0x80)
|
542 |
479 |
stb(A0 + i, d->B(i));
|
... | ... | |
544 |
481 |
FORCE_RET();
|
545 |
482 |
}
|
546 |
483 |
|
547 |
|
void OPPROTO glue(op_movl_mm_T0, SUFFIX) (void)
|
|
484 |
void glue(helper_movl_mm_T0, SUFFIX) (Reg *d, uint32_t val)
|
548 |
485 |
{
|
549 |
|
Reg *d;
|
550 |
|
d = (Reg *)((char *)env + PARAM1);
|
551 |
|
d->L(0) = T0;
|
|
486 |
d->L(0) = val;
|
552 |
487 |
d->L(1) = 0;
|
553 |
488 |
#if SHIFT == 1
|
554 |
489 |
d->Q(1) = 0;
|
555 |
490 |
#endif
|
556 |
491 |
}
|
557 |
492 |
|
558 |
|
void OPPROTO glue(op_movl_T0_mm, SUFFIX) (void)
|
559 |
|
{
|
560 |
|
Reg *s;
|
561 |
|
s = (Reg *)((char *)env + PARAM1);
|
562 |
|
T0 = s->L(0);
|
563 |
|
}
|
564 |
|
|
565 |
493 |
#ifdef TARGET_X86_64
|
566 |
|
void OPPROTO glue(op_movq_mm_T0, SUFFIX) (void)
|
|
494 |
void glue(helper_movq_mm_T0, SUFFIX) (Reg *d, uint64_t val)
|
567 |
495 |
{
|
568 |
|
Reg *d;
|
569 |
|
d = (Reg *)((char *)env + PARAM1);
|
570 |
|
d->Q(0) = T0;
|
|
496 |
d->Q(0) = val;
|
571 |
497 |
#if SHIFT == 1
|
572 |
498 |
d->Q(1) = 0;
|
573 |
499 |
#endif
|
574 |
500 |
}
|
575 |
|
|
576 |
|
void OPPROTO glue(op_movq_T0_mm, SUFFIX) (void)
|
577 |
|
{
|
578 |
|
Reg *s;
|
579 |
|
s = (Reg *)((char *)env + PARAM1);
|
580 |
|
T0 = s->Q(0);
|
581 |
|
}
|
582 |
501 |
#endif
|
583 |
502 |
|
584 |
503 |
#if SHIFT == 0
|
585 |
|
void OPPROTO glue(op_pshufw, SUFFIX) (void)
|
|
504 |
void glue(helper_pshufw, SUFFIX) (Reg *d, Reg *s, int order)
|
586 |
505 |
{
|
587 |
|
Reg r, *d, *s;
|
588 |
|
int order;
|
589 |
|
d = (Reg *)((char *)env + PARAM1);
|
590 |
|
s = (Reg *)((char *)env + PARAM2);
|
591 |
|
order = PARAM3;
|
|
506 |
Reg r;
|
592 |
507 |
r.W(0) = s->W(order & 3);
|
593 |
508 |
r.W(1) = s->W((order >> 2) & 3);
|
594 |
509 |
r.W(2) = s->W((order >> 4) & 3);
|
... | ... | |
596 |
511 |
*d = r;
|
597 |
512 |
}
|
598 |
513 |
#else
|
599 |
|
void OPPROTO op_shufps(void)
|
|
514 |
void helper_shufps(Reg *d, Reg *s, int order)
|
600 |
515 |
{
|
601 |
|
Reg r, *d, *s;
|
602 |
|
int order;
|
603 |
|
d = (Reg *)((char *)env + PARAM1);
|
604 |
|
s = (Reg *)((char *)env + PARAM2);
|
605 |
|
order = PARAM3;
|
|
516 |
Reg r;
|
606 |
517 |
r.L(0) = d->L(order & 3);
|
607 |
518 |
r.L(1) = d->L((order >> 2) & 3);
|
608 |
519 |
r.L(2) = s->L((order >> 4) & 3);
|
... | ... | |
610 |
521 |
*d = r;
|
611 |
522 |
}
|
612 |
523 |
|
613 |
|
void OPPROTO op_shufpd(void)
|
|
524 |
void helper_shufpd(Reg *d, Reg *s, int order)
|
614 |
525 |
{
|
615 |
|
Reg r, *d, *s;
|
616 |
|
int order;
|
617 |
|
d = (Reg *)((char *)env + PARAM1);
|
618 |
|
s = (Reg *)((char *)env + PARAM2);
|
619 |
|
order = PARAM3;
|
|
526 |
Reg r;
|
620 |
527 |
r.Q(0) = d->Q(order & 1);
|
621 |
528 |
r.Q(1) = s->Q((order >> 1) & 1);
|
622 |
529 |
*d = r;
|
623 |
530 |
}
|
624 |
531 |
|
625 |
|
void OPPROTO glue(op_pshufd, SUFFIX) (void)
|
|
532 |
void glue(helper_pshufd, SUFFIX) (Reg *d, Reg *s, int order)
|
626 |
533 |
{
|
627 |
|
Reg r, *d, *s;
|
628 |
|
int order;
|
629 |
|
d = (Reg *)((char *)env + PARAM1);
|
630 |
|
s = (Reg *)((char *)env + PARAM2);
|
631 |
|
order = PARAM3;
|
|
534 |
Reg r;
|
632 |
535 |
r.L(0) = s->L(order & 3);
|
633 |
536 |
r.L(1) = s->L((order >> 2) & 3);
|
634 |
537 |
r.L(2) = s->L((order >> 4) & 3);
|
... | ... | |
636 |
539 |
*d = r;
|
637 |
540 |
}
|
638 |
541 |
|
639 |
|
void OPPROTO glue(op_pshuflw, SUFFIX) (void)
|
|
542 |
void glue(helper_pshuflw, SUFFIX) (Reg *d, Reg *s, int order)
|
640 |
543 |
{
|
641 |
|
Reg r, *d, *s;
|
642 |
|
int order;
|
643 |
|
d = (Reg *)((char *)env + PARAM1);
|
644 |
|
s = (Reg *)((char *)env + PARAM2);
|
645 |
|
order = PARAM3;
|
|
544 |
Reg r;
|
646 |
545 |
r.W(0) = s->W(order & 3);
|
647 |
546 |
r.W(1) = s->W((order >> 2) & 3);
|
648 |
547 |
r.W(2) = s->W((order >> 4) & 3);
|
... | ... | |
651 |
550 |
*d = r;
|
652 |
551 |
}
|
653 |
552 |
|
654 |
|
void OPPROTO glue(op_pshufhw, SUFFIX) (void)
|
|
553 |
void glue(helper_pshufhw, SUFFIX) (Reg *d, Reg *s, int order)
|
655 |
554 |
{
|
656 |
|
Reg r, *d, *s;
|
657 |
|
int order;
|
658 |
|
d = (Reg *)((char *)env + PARAM1);
|
659 |
|
s = (Reg *)((char *)env + PARAM2);
|
660 |
|
order = PARAM3;
|
|
555 |
Reg r;
|
661 |
556 |
r.Q(0) = s->Q(0);
|
662 |
557 |
r.W(4) = s->W(4 + (order & 3));
|
663 |
558 |
r.W(5) = s->W(4 + ((order >> 2) & 3));
|
... | ... | |
671 |
566 |
/* FPU ops */
|
672 |
567 |
/* XXX: not accurate */
|
673 |
568 |
|
674 |
|
#define SSE_OP_S(name, F)\
|
675 |
|
void OPPROTO op_ ## name ## ps (void)\
|
|
569 |
#define SSE_HELPER_S(name, F)\
|
|
570 |
void helper_ ## name ## ps (Reg *d, Reg *s)\
|
676 |
571 |
{\
|
677 |
|
Reg *d, *s;\
|
678 |
|
d = (Reg *)((char *)env + PARAM1);\
|
679 |
|
s = (Reg *)((char *)env + PARAM2);\
|
680 |
572 |
d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
|
681 |
573 |
d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
|
682 |
574 |
d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
|
683 |
575 |
d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
|
684 |
576 |
}\
|
685 |
577 |
\
|
686 |
|
void OPPROTO op_ ## name ## ss (void)\
|
|
578 |
void helper_ ## name ## ss (Reg *d, Reg *s)\
|
687 |
579 |
{\
|
688 |
|
Reg *d, *s;\
|
689 |
|
d = (Reg *)((char *)env + PARAM1);\
|
690 |
|
s = (Reg *)((char *)env + PARAM2);\
|
691 |
580 |
d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
|
692 |
581 |
}\
|
693 |
|
void OPPROTO op_ ## name ## pd (void)\
|
|
582 |
void helper_ ## name ## pd (Reg *d, Reg *s)\
|
694 |
583 |
{\
|
695 |
|
Reg *d, *s;\
|
696 |
|
d = (Reg *)((char *)env + PARAM1);\
|
697 |
|
s = (Reg *)((char *)env + PARAM2);\
|
698 |
584 |
d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
|
699 |
585 |
d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
|
700 |
586 |
}\
|
701 |
587 |
\
|
702 |
|
void OPPROTO op_ ## name ## sd (void)\
|
|
588 |
void helper_ ## name ## sd (Reg *d, Reg *s)\
|
703 |
589 |
{\
|
704 |
|
Reg *d, *s;\
|
705 |
|
d = (Reg *)((char *)env + PARAM1);\
|
706 |
|
s = (Reg *)((char *)env + PARAM2);\
|
707 |
590 |
d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
|
708 |
591 |
}
|
709 |
592 |
|
... | ... | |
715 |
598 |
#define FPU_MAX(size, a, b) (a) > (b) ? (a) : (b)
|
716 |
599 |
#define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
|
717 |
600 |
|
718 |
|
SSE_OP_S(add, FPU_ADD)
|
719 |
|
SSE_OP_S(sub, FPU_SUB)
|
720 |
|
SSE_OP_S(mul, FPU_MUL)
|
721 |
|
SSE_OP_S(div, FPU_DIV)
|
722 |
|
SSE_OP_S(min, FPU_MIN)
|
723 |
|
SSE_OP_S(max, FPU_MAX)
|
724 |
|
SSE_OP_S(sqrt, FPU_SQRT)
|
|
601 |
SSE_HELPER_S(add, FPU_ADD)
|
|
602 |
SSE_HELPER_S(sub, FPU_SUB)
|
|
603 |
SSE_HELPER_S(mul, FPU_MUL)
|
|
604 |
SSE_HELPER_S(div, FPU_DIV)
|
|
605 |
SSE_HELPER_S(min, FPU_MIN)
|
|
606 |
SSE_HELPER_S(max, FPU_MAX)
|
|
607 |
SSE_HELPER_S(sqrt, FPU_SQRT)
|
725 |
608 |
|
726 |
609 |
|
727 |
610 |
/* float to float conversions */
|
728 |
|
void OPPROTO op_cvtps2pd(void)
|
|
611 |
void helper_cvtps2pd(Reg *d, Reg *s)
|
729 |
612 |
{
|
730 |
613 |
float32 s0, s1;
|
731 |
|
Reg *d, *s;
|
732 |
|
d = (Reg *)((char *)env + PARAM1);
|
733 |
|
s = (Reg *)((char *)env + PARAM2);
|
734 |
614 |
s0 = s->XMM_S(0);
|
735 |
615 |
s1 = s->XMM_S(1);
|
736 |
616 |
d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
|
737 |
617 |
d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
|
738 |
618 |
}
|
739 |
619 |
|
740 |
|
void OPPROTO op_cvtpd2ps(void)
|
|
620 |
void helper_cvtpd2ps(Reg *d, Reg *s)
|
741 |
621 |
{
|
742 |
|
Reg *d, *s;
|
743 |
|
d = (Reg *)((char *)env + PARAM1);
|
744 |
|
s = (Reg *)((char *)env + PARAM2);
|
745 |
622 |
d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
|
746 |
623 |
d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
|
747 |
624 |
d->Q(1) = 0;
|
748 |
625 |
}
|
749 |
626 |
|
750 |
|
void OPPROTO op_cvtss2sd(void)
|
|
627 |
void helper_cvtss2sd(Reg *d, Reg *s)
|
751 |
628 |
{
|
752 |
|
Reg *d, *s;
|
753 |
|
d = (Reg *)((char *)env + PARAM1);
|
754 |
|
s = (Reg *)((char *)env + PARAM2);
|
755 |
629 |
d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
|
756 |
630 |
}
|
757 |
631 |
|
758 |
|
void OPPROTO op_cvtsd2ss(void)
|
|
632 |
void helper_cvtsd2ss(Reg *d, Reg *s)
|
759 |
633 |
{
|
760 |
|
Reg *d, *s;
|
761 |
|
d = (Reg *)((char *)env + PARAM1);
|
762 |
|
s = (Reg *)((char *)env + PARAM2);
|
763 |
634 |
d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
|
764 |
635 |
}
|
765 |
636 |
|
766 |
637 |
/* integer to float */
|
767 |
|
void OPPROTO op_cvtdq2ps(void)
|
|
638 |
void helper_cvtdq2ps(Reg *d, Reg *s)
|
768 |
639 |
{
|
769 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
770 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
771 |
640 |
d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
|
772 |
641 |
d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
|
773 |
642 |
d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
|
774 |
643 |
d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
|
775 |
644 |
}
|
776 |
645 |
|
777 |
|
void OPPROTO op_cvtdq2pd(void)
|
|
646 |
void helper_cvtdq2pd(Reg *d, Reg *s)
|
778 |
647 |
{
|
779 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
780 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
781 |
648 |
int32_t l0, l1;
|
782 |
649 |
l0 = (int32_t)s->XMM_L(0);
|
783 |
650 |
l1 = (int32_t)s->XMM_L(1);
|
... | ... | |
785 |
652 |
d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
|
786 |
653 |
}
|
787 |
654 |
|
788 |
|
void OPPROTO op_cvtpi2ps(void)
|
|
655 |
void helper_cvtpi2ps(XMMReg *d, MMXReg *s)
|
789 |
656 |
{
|
790 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
791 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
792 |
657 |
d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
|
793 |
658 |
d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
|
794 |
659 |
}
|
795 |
660 |
|
796 |
|
void OPPROTO op_cvtpi2pd(void)
|
|
661 |
void helper_cvtpi2pd(XMMReg *d, MMXReg *s)
|
797 |
662 |
{
|
798 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
799 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
800 |
663 |
d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
|
801 |
664 |
d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
|
802 |
665 |
}
|
803 |
666 |
|
804 |
|
void OPPROTO op_cvtsi2ss(void)
|
|
667 |
void helper_cvtsi2ss(XMMReg *d, uint32_t val)
|
805 |
668 |
{
|
806 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
807 |
|
d->XMM_S(0) = int32_to_float32(T0, &env->sse_status);
|
|
669 |
d->XMM_S(0) = int32_to_float32(val, &env->sse_status);
|
808 |
670 |
}
|
809 |
671 |
|
810 |
|
void OPPROTO op_cvtsi2sd(void)
|
|
672 |
void helper_cvtsi2sd(XMMReg *d, uint32_t val)
|
811 |
673 |
{
|
812 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
813 |
|
d->XMM_D(0) = int32_to_float64(T0, &env->sse_status);
|
|
674 |
d->XMM_D(0) = int32_to_float64(val, &env->sse_status);
|
814 |
675 |
}
|
815 |
676 |
|
816 |
677 |
#ifdef TARGET_X86_64
|
817 |
|
void OPPROTO op_cvtsq2ss(void)
|
|
678 |
void helper_cvtsq2ss(XMMReg *d, uint64_t val)
|
818 |
679 |
{
|
819 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
820 |
|
d->XMM_S(0) = int64_to_float32(T0, &env->sse_status);
|
|
680 |
d->XMM_S(0) = int64_to_float32(val, &env->sse_status);
|
821 |
681 |
}
|
822 |
682 |
|
823 |
|
void OPPROTO op_cvtsq2sd(void)
|
|
683 |
void helper_cvtsq2sd(XMMReg *d, uint64_t val)
|
824 |
684 |
{
|
825 |
|
XMMReg *d = (Reg *)((char *)env + PARAM1);
|
826 |
|
d->XMM_D(0) = int64_to_float64(T0, &env->sse_status);
|
|
685 |
d->XMM_D(0) = int64_to_float64(val, &env->sse_status);
|
827 |
686 |
}
|
828 |
687 |
#endif
|
829 |
688 |
|
830 |
689 |
/* float to integer */
|
831 |
|
void OPPROTO op_cvtps2dq(void)
|
|
690 |
void helper_cvtps2dq(XMMReg *d, XMMReg *s)
|
832 |
691 |
{
|
833 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
834 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
835 |
692 |
d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
|
836 |
693 |
d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
|
837 |
694 |
d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
|
838 |
695 |
d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
|
839 |
696 |
}
|
840 |
697 |
|
841 |
|
void OPPROTO op_cvtpd2dq(void)
|
|
698 |
void helper_cvtpd2dq(XMMReg *d, XMMReg *s)
|
842 |
699 |
{
|
843 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
844 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
845 |
700 |
d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
|
846 |
701 |
d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
|
847 |
702 |
d->XMM_Q(1) = 0;
|
848 |
703 |
}
|
849 |
704 |
|
850 |
|
void OPPROTO op_cvtps2pi(void)
|
|
705 |
void helper_cvtps2pi(MMXReg *d, XMMReg *s)
|
851 |
706 |
{
|
852 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
853 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
854 |
707 |
d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
|
855 |
708 |
d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
|
856 |
709 |
}
|
857 |
710 |
|
858 |
|
void OPPROTO op_cvtpd2pi(void)
|
|
711 |
void helper_cvtpd2pi(MMXReg *d, XMMReg *s)
|
859 |
712 |
{
|
860 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
861 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
862 |
713 |
d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
|
863 |
714 |
d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
|
864 |
715 |
}
|
865 |
716 |
|
866 |
|
void OPPROTO op_cvtss2si(void)
|
|
717 |
int32_t helper_cvtss2si(XMMReg *s)
|
867 |
718 |
{
|
868 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
869 |
|
T0 = float32_to_int32(s->XMM_S(0), &env->sse_status);
|
|
719 |
return float32_to_int32(s->XMM_S(0), &env->sse_status);
|
870 |
720 |
}
|
871 |
721 |
|
872 |
|
void OPPROTO op_cvtsd2si(void)
|
|
722 |
int32_t helper_cvtsd2si(XMMReg *s)
|
873 |
723 |
{
|
874 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
875 |
|
T0 = float64_to_int32(s->XMM_D(0), &env->sse_status);
|
|
724 |
return float64_to_int32(s->XMM_D(0), &env->sse_status);
|
876 |
725 |
}
|
877 |
726 |
|
878 |
727 |
#ifdef TARGET_X86_64
|
879 |
|
void OPPROTO op_cvtss2sq(void)
|
|
728 |
int64_t helper_cvtss2sq(XMMReg *s)
|
880 |
729 |
{
|
881 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
882 |
|
T0 = float32_to_int64(s->XMM_S(0), &env->sse_status);
|
|
730 |
return float32_to_int64(s->XMM_S(0), &env->sse_status);
|
883 |
731 |
}
|
884 |
732 |
|
885 |
|
void OPPROTO op_cvtsd2sq(void)
|
|
733 |
int64_t helper_cvtsd2sq(XMMReg *s)
|
886 |
734 |
{
|
887 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
888 |
|
T0 = float64_to_int64(s->XMM_D(0), &env->sse_status);
|
|
735 |
return float64_to_int64(s->XMM_D(0), &env->sse_status);
|
889 |
736 |
}
|
890 |
737 |
#endif
|
891 |
738 |
|
892 |
739 |
/* float to integer truncated */
|
893 |
|
void OPPROTO op_cvttps2dq(void)
|
|
740 |
void helper_cvttps2dq(XMMReg *d, XMMReg *s)
|
894 |
741 |
{
|
895 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
896 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
897 |
742 |
d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
|
898 |
743 |
d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
|
899 |
744 |
d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
|
900 |
745 |
d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
|
901 |
746 |
}
|
902 |
747 |
|
903 |
|
void OPPROTO op_cvttpd2dq(void)
|
|
748 |
void helper_cvttpd2dq(XMMReg *d, XMMReg *s)
|
904 |
749 |
{
|
905 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
906 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
907 |
750 |
d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
|
908 |
751 |
d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
|
909 |
752 |
d->XMM_Q(1) = 0;
|
910 |
753 |
}
|
911 |
754 |
|
912 |
|
void OPPROTO op_cvttps2pi(void)
|
|
755 |
void helper_cvttps2pi(MMXReg *d, XMMReg *s)
|
913 |
756 |
{
|
914 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
915 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
916 |
757 |
d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
|
917 |
758 |
d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
|
918 |
759 |
}
|
919 |
760 |
|
920 |
|
void OPPROTO op_cvttpd2pi(void)
|
|
761 |
void helper_cvttpd2pi(MMXReg *d, XMMReg *s)
|
921 |
762 |
{
|
922 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
923 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
924 |
763 |
d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
|
925 |
764 |
d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
|
926 |
765 |
}
|
927 |
766 |
|
928 |
|
void OPPROTO op_cvttss2si(void)
|
|
767 |
int32_t helper_cvttss2si(XMMReg *s)
|
929 |
768 |
{
|
930 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
931 |
|
T0 = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
|
|
769 |
return float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
|
932 |
770 |
}
|
933 |
771 |
|
934 |
|
void OPPROTO op_cvttsd2si(void)
|
|
772 |
int32_t helper_cvttsd2si(XMMReg *s)
|
935 |
773 |
{
|
936 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
937 |
|
T0 = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
|
|
774 |
return float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
|
938 |
775 |
}
|
939 |
776 |
|
940 |
777 |
#ifdef TARGET_X86_64
|
941 |
|
void OPPROTO op_cvttss2sq(void)
|
|
778 |
int64_t helper_cvttss2sq(XMMReg *s)
|
942 |
779 |
{
|
943 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
944 |
|
T0 = float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
|
|
780 |
return float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
|
945 |
781 |
}
|
946 |
782 |
|
947 |
|
void OPPROTO op_cvttsd2sq(void)
|
|
783 |
int64_t helper_cvttsd2sq(XMMReg *s)
|
948 |
784 |
{
|
949 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM1);
|
950 |
|
T0 = float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
|
|
785 |
return float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
|
951 |
786 |
}
|
952 |
787 |
#endif
|
953 |
788 |
|
954 |
|
void OPPROTO op_rsqrtps(void)
|
|
789 |
void helper_rsqrtps(XMMReg *d, XMMReg *s)
|
955 |
790 |
{
|
956 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
957 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
958 |
791 |
d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
|
959 |
792 |
d->XMM_S(1) = approx_rsqrt(s->XMM_S(1));
|
960 |
793 |
d->XMM_S(2) = approx_rsqrt(s->XMM_S(2));
|
961 |
794 |
d->XMM_S(3) = approx_rsqrt(s->XMM_S(3));
|
962 |
795 |
}
|
963 |
796 |
|
964 |
|
void OPPROTO op_rsqrtss(void)
|
|
797 |
void helper_rsqrtss(XMMReg *d, XMMReg *s)
|
965 |
798 |
{
|
966 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
967 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
968 |
799 |
d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
|
969 |
800 |
}
|
970 |
801 |
|
971 |
|
void OPPROTO op_rcpps(void)
|
|
802 |
void helper_rcpps(XMMReg *d, XMMReg *s)
|
972 |
803 |
{
|
973 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
974 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
975 |
804 |
d->XMM_S(0) = approx_rcp(s->XMM_S(0));
|
976 |
805 |
d->XMM_S(1) = approx_rcp(s->XMM_S(1));
|
977 |
806 |
d->XMM_S(2) = approx_rcp(s->XMM_S(2));
|
978 |
807 |
d->XMM_S(3) = approx_rcp(s->XMM_S(3));
|
979 |
808 |
}
|
980 |
809 |
|
981 |
|
void OPPROTO op_rcpss(void)
|
|
810 |
void helper_rcpss(XMMReg *d, XMMReg *s)
|
982 |
811 |
{
|
983 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
984 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
985 |
812 |
d->XMM_S(0) = approx_rcp(s->XMM_S(0));
|
986 |
813 |
}
|
987 |
814 |
|
988 |
|
void OPPROTO op_haddps(void)
|
|
815 |
void helper_haddps(XMMReg *d, XMMReg *s)
|
989 |
816 |
{
|
990 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
991 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
992 |
817 |
XMMReg r;
|
993 |
818 |
r.XMM_S(0) = d->XMM_S(0) + d->XMM_S(1);
|
994 |
819 |
r.XMM_S(1) = d->XMM_S(2) + d->XMM_S(3);
|
... | ... | |
997 |
822 |
*d = r;
|
998 |
823 |
}
|
999 |
824 |
|
1000 |
|
void OPPROTO op_haddpd(void)
|
|
825 |
void helper_haddpd(XMMReg *d, XMMReg *s)
|
1001 |
826 |
{
|
1002 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
1003 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
1004 |
827 |
XMMReg r;
|
1005 |
828 |
r.XMM_D(0) = d->XMM_D(0) + d->XMM_D(1);
|
1006 |
829 |
r.XMM_D(1) = s->XMM_D(0) + s->XMM_D(1);
|
1007 |
830 |
*d = r;
|
1008 |
831 |
}
|
1009 |
832 |
|
1010 |
|
void OPPROTO op_hsubps(void)
|
|
833 |
void helper_hsubps(XMMReg *d, XMMReg *s)
|
1011 |
834 |
{
|
1012 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
1013 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
1014 |
835 |
XMMReg r;
|
1015 |
836 |
r.XMM_S(0) = d->XMM_S(0) - d->XMM_S(1);
|
1016 |
837 |
r.XMM_S(1) = d->XMM_S(2) - d->XMM_S(3);
|
... | ... | |
1019 |
840 |
*d = r;
|
1020 |
841 |
}
|
1021 |
842 |
|
1022 |
|
void OPPROTO op_hsubpd(void)
|
|
843 |
void helper_hsubpd(XMMReg *d, XMMReg *s)
|
1023 |
844 |
{
|
1024 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
1025 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
1026 |
845 |
XMMReg r;
|
1027 |
846 |
r.XMM_D(0) = d->XMM_D(0) - d->XMM_D(1);
|
1028 |
847 |
r.XMM_D(1) = s->XMM_D(0) - s->XMM_D(1);
|
1029 |
848 |
*d = r;
|
1030 |
849 |
}
|
1031 |
850 |
|
1032 |
|
void OPPROTO op_addsubps(void)
|
|
851 |
void helper_addsubps(XMMReg *d, XMMReg *s)
|
1033 |
852 |
{
|
1034 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
1035 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
1036 |
853 |
d->XMM_S(0) = d->XMM_S(0) - s->XMM_S(0);
|
1037 |
854 |
d->XMM_S(1) = d->XMM_S(1) + s->XMM_S(1);
|
1038 |
855 |
d->XMM_S(2) = d->XMM_S(2) - s->XMM_S(2);
|
1039 |
856 |
d->XMM_S(3) = d->XMM_S(3) + s->XMM_S(3);
|
1040 |
857 |
}
|
1041 |
858 |
|
1042 |
|
void OPPROTO op_addsubpd(void)
|
|
859 |
void helper_addsubpd(XMMReg *d, XMMReg *s)
|
1043 |
860 |
{
|
1044 |
|
XMMReg *d = (XMMReg *)((char *)env + PARAM1);
|
1045 |
|
XMMReg *s = (XMMReg *)((char *)env + PARAM2);
|
1046 |
861 |
d->XMM_D(0) = d->XMM_D(0) - s->XMM_D(0);
|
1047 |
862 |
d->XMM_D(1) = d->XMM_D(1) + s->XMM_D(1);
|
1048 |
863 |
}
|
1049 |
864 |
|
1050 |
865 |
/* XXX: unordered */
|
1051 |
|
#define SSE_OP_CMP(name, F)\
|
1052 |
|
void OPPROTO op_ ## name ## ps (void)\
|
|
866 |
#define SSE_HELPER_CMP(name, F)\
|
|
867 |
void helper_ ## name ## ps (Reg *d, Reg *s)\
|
1053 |
868 |
{\
|
1054 |
|
Reg *d, *s;\
|
1055 |
|
d = (Reg *)((char *)env + PARAM1);\
|
1056 |
|
s = (Reg *)((char *)env + PARAM2);\
|
1057 |
869 |
d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
|
1058 |
870 |
d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
|
1059 |
871 |
d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
|
1060 |
872 |
d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
|
1061 |
873 |
}\
|
1062 |
874 |
\
|
1063 |
|
void OPPROTO op_ ## name ## ss (void)\
|
|
875 |
void helper_ ## name ## ss (Reg *d, Reg *s)\
|
1064 |
876 |
{\
|
1065 |
|
Reg *d, *s;\
|
1066 |
|
d = (Reg *)((char *)env + PARAM1);\
|
1067 |
|
s = (Reg *)((char *)env + PARAM2);\
|
1068 |
877 |
d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
|
1069 |
878 |
}\
|
1070 |
|
void OPPROTO op_ ## name ## pd (void)\
|
|
879 |
void helper_ ## name ## pd (Reg *d, Reg *s)\
|
1071 |
880 |
{\
|
1072 |
|
Reg *d, *s;\
|
1073 |
|
d = (Reg *)((char *)env + PARAM1);\
|
1074 |
|
s = (Reg *)((char *)env + PARAM2);\
|
1075 |
881 |
d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
|
1076 |
882 |
d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
|
1077 |
883 |
}\
|
1078 |
884 |
\
|
1079 |
|
void OPPROTO op_ ## name ## sd (void)\
|
|
885 |
void helper_ ## name ## sd (Reg *d, Reg *s)\
|
1080 |
886 |
{\
|
1081 |
|
Reg *d, *s;\
|
1082 |
|
d = (Reg *)((char *)env + PARAM1);\
|
1083 |
|
s = (Reg *)((char *)env + PARAM2);\
|
1084 |
887 |
d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
|
1085 |
888 |
}
|
1086 |
889 |
|
... | ... | |
1093 |
896 |
#define FPU_CMPNLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? 0 : -1
|
1094 |
897 |
#define FPU_CMPORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? 0 : -1
|
1095 |
898 |
|
1096 |
|
SSE_OP_CMP(cmpeq, FPU_CMPEQ)
|
1097 |
|
SSE_OP_CMP(cmplt, FPU_CMPLT)
|
1098 |
|
SSE_OP_CMP(cmple, FPU_CMPLE)
|
1099 |
|
SSE_OP_CMP(cmpunord, FPU_CMPUNORD)
|
1100 |
|
SSE_OP_CMP(cmpneq, FPU_CMPNEQ)
|
1101 |
|
SSE_OP_CMP(cmpnlt, FPU_CMPNLT)
|
1102 |
|
SSE_OP_CMP(cmpnle, FPU_CMPNLE)
|
1103 |
|
SSE_OP_CMP(cmpord, FPU_CMPORD)
|
|
899 |
SSE_HELPER_CMP(cmpeq, FPU_CMPEQ)
|
|
900 |
SSE_HELPER_CMP(cmplt, FPU_CMPLT)
|
|
901 |
SSE_HELPER_CMP(cmple, FPU_CMPLE)
|
|
902 |
SSE_HELPER_CMP(cmpunord, FPU_CMPUNORD)
|
|
903 |
SSE_HELPER_CMP(cmpneq, FPU_CMPNEQ)
|
|
904 |
SSE_HELPER_CMP(cmpnlt, FPU_CMPNLT)
|
|
905 |
SSE_HELPER_CMP(cmpnle, FPU_CMPNLE)
|
|
906 |
SSE_HELPER_CMP(cmpord, FPU_CMPORD)
|
1104 |
907 |
|
1105 |
908 |
const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
|
1106 |
909 |
|
1107 |
|
void OPPROTO op_ucomiss(void)
|
|
910 |
void helper_ucomiss(Reg *d, Reg *s)
|
1108 |
911 |
{
|
1109 |
912 |
int ret;
|
1110 |
913 |
float32 s0, s1;
|
1111 |
|
Reg *d, *s;
|
1112 |
|
d = (Reg *)((char *)env + PARAM1);
|
1113 |
|
s = (Reg *)((char *)env + PARAM2);
|
1114 |
914 |
|
1115 |
915 |
s0 = d->XMM_S(0);
|
1116 |
916 |
s1 = s->XMM_S(0);
|
... | ... | |
1119 |
919 |
FORCE_RET();
|
1120 |
920 |
}
|
1121 |
921 |
|
1122 |
|
void OPPROTO op_comiss(void)
|
|
922 |
void helper_comiss(Reg *d, Reg *s)
|
1123 |
923 |
{
|
1124 |
924 |
int ret;
|
1125 |
925 |
float32 s0, s1;
|
1126 |
|
Reg *d, *s;
|
1127 |
|
d = (Reg *)((char *)env + PARAM1);
|
1128 |
|
s = (Reg *)((char *)env + PARAM2);
|
1129 |
926 |
|
1130 |
927 |
s0 = d->XMM_S(0);
|
1131 |
928 |
s1 = s->XMM_S(0);
|
... | ... | |
1134 |
931 |
FORCE_RET();
|
1135 |
932 |
}
|
1136 |
933 |
|
1137 |
|
void OPPROTO op_ucomisd(void)
|
|
934 |
void helper_ucomisd(Reg *d, Reg *s)
|
1138 |
935 |
{
|
1139 |
936 |
int ret;
|
1140 |
937 |
float64 d0, d1;
|
1141 |
|
Reg *d, *s;
|
1142 |
|
d = (Reg *)((char *)env + PARAM1);
|
1143 |
|
s = (Reg *)((char *)env + PARAM2);
|
1144 |
938 |
|
1145 |
939 |
d0 = d->XMM_D(0);
|
1146 |
940 |
d1 = s->XMM_D(0);
|
... | ... | |
1149 |
943 |
FORCE_RET();
|
1150 |
944 |
}
|
1151 |
945 |
|
1152 |
|
void OPPROTO op_comisd(void)
|
|
946 |
void helper_comisd(Reg *d, Reg *s)
|
1153 |
947 |
{
|
1154 |
948 |
int ret;
|
1155 |
949 |
float64 d0, d1;
|
1156 |
|
Reg *d, *s;
|
1157 |
|
d = (Reg *)((char *)env + PARAM1);
|
1158 |
|
s = (Reg *)((char *)env + PARAM2);
|
1159 |
950 |
|
1160 |
951 |
d0 = d->XMM_D(0);
|
1161 |
952 |
d1 = s->XMM_D(0);
|
... | ... | |
1164 |
955 |
FORCE_RET();
|
1165 |
956 |
}
|
1166 |
957 |
|
1167 |
|
void OPPROTO op_movmskps(void)
|
|
958 |
uint32_t helper_movmskps(Reg *s)
|
1168 |
959 |
{
|
1169 |
960 |
int b0, b1, b2, b3;
|
1170 |
|
Reg *s;
|
1171 |
|
s = (Reg *)((char *)env + PARAM1);
|
1172 |
961 |
b0 = s->XMM_L(0) >> 31;
|
1173 |
962 |
b1 = s->XMM_L(1) >> 31;
|
1174 |
963 |
b2 = s->XMM_L(2) >> 31;
|
1175 |
964 |
b3 = s->XMM_L(3) >> 31;
|
1176 |
|
T0 = b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
|
|
965 |
return b0 | (b1 << 1) | (b2 << 2) | (b3 << 3);
|
1177 |
966 |
}
|
1178 |
967 |
|
1179 |
|
void OPPROTO op_movmskpd(void)
|
|
968 |
uint32_t helper_movmskpd(Reg *s)
|
1180 |
969 |
{
|
1181 |
970 |
int b0, b1;
|
1182 |
|
Reg *s;
|
1183 |
|
s = (Reg *)((char *)env + PARAM1);
|
1184 |
971 |
b0 = s->XMM_L(1) >> 31;
|
1185 |
972 |
b1 = s->XMM_L(3) >> 31;
|
1186 |
|
T0 = b0 | (b1 << 1);
|
|
973 |
return b0 | (b1 << 1);
|
1187 |
974 |
}
|
1188 |
975 |
|
1189 |
976 |
#endif
|
1190 |
977 |
|
1191 |
|
void OPPROTO glue(op_pmovmskb, SUFFIX)(void)
|
1192 |
|
{
|
1193 |
|
Reg *s;
|
1194 |
|
s = (Reg *)((char *)env + PARAM1);
|
1195 |
|
T0 = 0;
|
1196 |
|
T0 |= (s->XMM_B(0) >> 7);
|
1197 |
|
T0 |= (s->XMM_B(1) >> 6) & 0x02;
|
1198 |
|
T0 |= (s->XMM_B(2) >> 5) & 0x04;
|
1199 |
|
T0 |= (s->XMM_B(3) >> 4) & 0x08;
|
1200 |
|
T0 |= (s->XMM_B(4) >> 3) & 0x10;
|
1201 |
|
T0 |= (s->XMM_B(5) >> 2) & 0x20;
|
1202 |
|
T0 |= (s->XMM_B(6) >> 1) & 0x40;
|
1203 |
|
T0 |= (s->XMM_B(7)) & 0x80;
|
|
978 |
uint32_t glue(helper_pmovmskb, SUFFIX)(Reg *s)
|
|
979 |
{
|
|
980 |
uint32_t val;
|
|
981 |
val = 0;
|
|
982 |
val |= (s->XMM_B(0) >> 7);
|
|
983 |
val |= (s->XMM_B(1) >> 6) & 0x02;
|
|
984 |
val |= (s->XMM_B(2) >> 5) & 0x04;
|
|
985 |
val |= (s->XMM_B(3) >> 4) & 0x08;
|
|
986 |
val |= (s->XMM_B(4) >> 3) & 0x10;
|
|
987 |
val |= (s->XMM_B(5) >> 2) & 0x20;
|
|
988 |
val |= (s->XMM_B(6) >> 1) & 0x40;
|
|
989 |
val |= (s->XMM_B(7)) & 0x80;
|
1204 |
990 |
#if SHIFT == 1
|
1205 |
|
T0 |= (s->XMM_B(8) << 1) & 0x0100;
|
1206 |
|
T0 |= (s->XMM_B(9) << 2) & 0x0200;
|
1207 |
|
T0 |= (s->XMM_B(10) << 3) & 0x0400;
|
1208 |
|
T0 |= (s->XMM_B(11) << 4) & 0x0800;
|
1209 |
|
T0 |= (s->XMM_B(12) << 5) & 0x1000;
|
1210 |
|
T0 |= (s->XMM_B(13) << 6) & 0x2000;
|
1211 |
|
T0 |= (s->XMM_B(14) << 7) & 0x4000;
|
1212 |
|
T0 |= (s->XMM_B(15) << 8) & 0x8000;
|
|
991 |
val |= (s->XMM_B(8) << 1) & 0x0100;
|
|
992 |
val |= (s->XMM_B(9) << 2) & 0x0200;
|
|
993 |
val |= (s->XMM_B(10) << 3) & 0x0400;
|
|
994 |
val |= (s->XMM_B(11) << 4) & 0x0800;
|
|
995 |
val |= (s->XMM_B(12) << 5) & 0x1000;
|
|
996 |
val |= (s->XMM_B(13) << 6) & 0x2000;
|
|
997 |
val |= (s->XMM_B(14) << 7) & 0x4000;
|
|
998 |
val |= (s->XMM_B(15) << 8) & 0x8000;
|
1213 |
999 |
#endif
|
|
1000 |
return val;
|
1214 |
1001 |
}
|
1215 |
1002 |
|
1216 |
|
void OPPROTO glue(op_pinsrw, SUFFIX) (void)
|
1217 |
|
{
|
1218 |
|
Reg *d = (Reg *)((char *)env + PARAM1);
|
1219 |
|
int pos = PARAM2;
|
1220 |
|
|
1221 |
|
d->W(pos) = T0;
|
1222 |
|
}
|
1223 |
|
|
1224 |
|
void OPPROTO glue(op_pextrw, SUFFIX) (void)
|
1225 |
|
{
|
1226 |
|
Reg *s = (Reg *)((char *)env + PARAM1);
|
1227 |
|
int pos = PARAM2;
|
1228 |
|
|
1229 |
|
T0 = s->W(pos);
|
1230 |
|
}
|
1231 |
|
|
1232 |
|
void OPPROTO glue(op_packsswb, SUFFIX) (void)
|
|
1003 |
void glue(helper_packsswb, SUFFIX) (Reg *d, Reg *s)
|
1233 |
1004 |
{
|
1234 |
|
Reg r, *d, *s;
|
1235 |
|
d = (Reg *)((char *)env + PARAM1);
|
1236 |
|
s = (Reg *)((char *)env + PARAM2);
|
|
1005 |
Reg r;
|
1237 |
1006 |
|
1238 |
1007 |
r.B(0) = satsb((int16_t)d->W(0));
|
1239 |
1008 |
r.B(1) = satsb((int16_t)d->W(1));
|
... | ... | |
1258 |
1027 |
*d = r;
|
1259 |
1028 |
}
|
1260 |
1029 |
|
1261 |
|
void OPPROTO glue(op_packuswb, SUFFIX) (void)
|
|
1030 |
void glue(helper_packuswb, SUFFIX) (Reg *d, Reg *s)
|
1262 |
1031 |
{
|
1263 |
|
Reg r, *d, *s;
|
1264 |
|
d = (Reg *)((char *)env + PARAM1);
|
1265 |
|
s = (Reg *)((char *)env + PARAM2);
|
|
1032 |
Reg r;
|
1266 |
1033 |
|
1267 |
1034 |
r.B(0) = satub((int16_t)d->W(0));
|
1268 |
1035 |
r.B(1) = satub((int16_t)d->W(1));
|
... | ... | |
1287 |
1054 |
*d = r;
|
1288 |
1055 |
}
|
1289 |
1056 |
|
1290 |
|
void OPPROTO glue(op_packssdw, SUFFIX) (void)
|
|
1057 |
void glue(helper_packssdw, SUFFIX) (Reg *d, Reg *s)
|
1291 |
1058 |
{
|
1292 |
|
Reg r, *d, *s;
|
1293 |
|
d = (Reg *)((char *)env + PARAM1);
|
1294 |
|
s = (Reg *)((char *)env + PARAM2);
|
|
1059 |
Reg r;
|
1295 |
1060 |
|
1296 |
1061 |
r.W(0) = satsw(d->L(0));
|
1297 |
1062 |
r.W(1) = satsw(d->L(1));
|
... | ... | |
1310 |
1075 |
|
1311 |
1076 |
#define UNPCK_OP(base_name, base) \
|
1312 |
1077 |
\
|
1313 |
|
void OPPROTO glue(op_punpck ## base_name ## bw, SUFFIX) (void) \
|
|
1078 |
void glue(helper_punpck ## base_name ## bw, SUFFIX) (Reg *d, Reg *s) \
|
1314 |
1079 |
{ \
|
1315 |
|
Reg r, *d, *s; \
|
1316 |
|
d = (Reg *)((char *)env + PARAM1); \
|
1317 |
|
s = (Reg *)((char *)env + PARAM2); \
|
|
1080 |
Reg r; \
|
1318 |
1081 |
\
|
1319 |
1082 |
r.B(0) = d->B((base << (SHIFT + 2)) + 0); \
|
1320 |
1083 |
r.B(1) = s->B((base << (SHIFT + 2)) + 0); \
|
... | ... | |
1337 |
1100 |
*d = r; \
|
1338 |
1101 |
} \
|
1339 |
1102 |
\
|
1340 |
|
void OPPROTO glue(op_punpck ## base_name ## wd, SUFFIX) (void) \
|
|
1103 |
void glue(helper_punpck ## base_name ## wd, SUFFIX) (Reg *d, Reg *s) \
|
1341 |
1104 |
{ \
|
1342 |
|
Reg r, *d, *s; \
|
1343 |
|
d = (Reg *)((char *)env + PARAM1); \
|
1344 |
|
s = (Reg *)((char *)env + PARAM2); \
|
|
1105 |
Reg r; \
|
1345 |
1106 |
\
|
1346 |
1107 |
r.W(0) = d->W((base << (SHIFT + 1)) + 0); \
|
1347 |
1108 |
r.W(1) = s->W((base << (SHIFT + 1)) + 0); \
|
... | ... | |
1356 |
1117 |
*d = r; \
|
1357 |
1118 |
} \
|
1358 |
1119 |
\
|
1359 |
|
void OPPROTO glue(op_punpck ## base_name ## dq, SUFFIX) (void) \
|
|
1120 |
void glue(helper_punpck ## base_name ## dq, SUFFIX) (Reg *d, Reg *s) \
|
1360 |
1121 |
{ \
|
1361 |
|
Reg r, *d, *s; \
|
1362 |
|
d = (Reg *)((char *)env + PARAM1); \
|
1363 |
|
s = (Reg *)((char *)env + PARAM2); \
|
|
1122 |
Reg r; \
|
1364 |
1123 |
\
|
1365 |
1124 |
r.L(0) = d->L((base << SHIFT) + 0); \
|
1366 |
1125 |
r.L(1) = s->L((base << SHIFT) + 0); \
|
... | ... | |
1372 |
1131 |
} \
|
1373 |
1132 |
\
|
1374 |
1133 |
XMM_ONLY( \
|
1375 |
|
void OPPROTO glue(op_punpck ## base_name ## qdq, SUFFIX) (void) \
|
|
1134 |
void glue(helper_punpck ## base_name ## qdq, SUFFIX) (Reg *d, Reg *s) \
|
1376 |
1135 |
{ \
|
1377 |
|
Reg r, *d, *s; \
|
1378 |
|
d = (Reg *)((char *)env + PARAM1); \
|
1379 |
|
s = (Reg *)((char *)env + PARAM2); \
|
|
1136 |
Reg r; \
|
1380 |
1137 |
\
|
1381 |
1138 |
r.Q(0) = d->Q(base); \
|
1382 |
1139 |
r.Q(1) = s->Q(base); \
|
... | ... | |
1389 |
1146 |
|
1390 |
1147 |
/* 3DNow! float ops */
|
1391 |
1148 |
#if SHIFT == 0
|
1392 |
|
void OPPROTO op_pi2fd(void)
|
|
1149 |
void helper_pi2fd(MMXReg *d, MMXReg *s)
|
1393 |
1150 |
{
|
1394 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1395 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1396 |
1151 |
d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
|
1397 |
1152 |
d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
|
1398 |
1153 |
}
|
1399 |
1154 |
|
1400 |
|
void OPPROTO op_pi2fw(void)
|
|
1155 |
void helper_pi2fw(MMXReg *d, MMXReg *s)
|
1401 |
1156 |
{
|
1402 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1403 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1404 |
1157 |
d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
|
1405 |
1158 |
d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
|
1406 |
1159 |
}
|
1407 |
1160 |
|
1408 |
|
void OPPROTO op_pf2id(void)
|
|
1161 |
void helper_pf2id(MMXReg *d, MMXReg *s)
|
1409 |
1162 |
{
|
1410 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1411 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1412 |
1163 |
d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
|
1413 |
1164 |
d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
|
1414 |
1165 |
}
|
1415 |
1166 |
|
1416 |
|
void OPPROTO op_pf2iw(void)
|
|
1167 |
void helper_pf2iw(MMXReg *d, MMXReg *s)
|
1417 |
1168 |
{
|
1418 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1419 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1420 |
1169 |
d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status));
|
1421 |
1170 |
d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status));
|
1422 |
1171 |
}
|
1423 |
1172 |
|
1424 |
|
void OPPROTO op_pfacc(void)
|
|
1173 |
void helper_pfacc(MMXReg *d, MMXReg *s)
|
1425 |
1174 |
{
|
1426 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1427 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1428 |
1175 |
MMXReg r;
|
1429 |
1176 |
r.MMX_S(0) = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
|
1430 |
1177 |
r.MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
|
1431 |
1178 |
*d = r;
|
1432 |
1179 |
}
|
1433 |
1180 |
|
1434 |
|
void OPPROTO op_pfadd(void)
|
|
1181 |
void helper_pfadd(MMXReg *d, MMXReg *s)
|
1435 |
1182 |
{
|
1436 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1437 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1438 |
1183 |
d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
|
1439 |
1184 |
d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
|
1440 |
1185 |
}
|
1441 |
1186 |
|
1442 |
|
void OPPROTO op_pfcmpeq(void)
|
|
1187 |
void helper_pfcmpeq(MMXReg *d, MMXReg *s)
|
1443 |
1188 |
{
|
1444 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1445 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1446 |
1189 |
d->MMX_L(0) = float32_eq(d->MMX_S(0), s->MMX_S(0), &env->mmx_status) ? -1 : 0;
|
1447 |
1190 |
d->MMX_L(1) = float32_eq(d->MMX_S(1), s->MMX_S(1), &env->mmx_status) ? -1 : 0;
|
1448 |
1191 |
}
|
1449 |
1192 |
|
1450 |
|
void OPPROTO op_pfcmpge(void)
|
|
1193 |
void helper_pfcmpge(MMXReg *d, MMXReg *s)
|
1451 |
1194 |
{
|
1452 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1453 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1454 |
1195 |
d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
|
1455 |
1196 |
d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
|
1456 |
1197 |
}
|
1457 |
1198 |
|
1458 |
|
void OPPROTO op_pfcmpgt(void)
|
|
1199 |
void helper_pfcmpgt(MMXReg *d, MMXReg *s)
|
1459 |
1200 |
{
|
1460 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1461 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1462 |
1201 |
d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status) ? -1 : 0;
|
1463 |
1202 |
d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status) ? -1 : 0;
|
1464 |
1203 |
}
|
1465 |
1204 |
|
1466 |
|
void OPPROTO op_pfmax(void)
|
|
1205 |
void helper_pfmax(MMXReg *d, MMXReg *s)
|
1467 |
1206 |
{
|
1468 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1469 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1470 |
1207 |
if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status))
|
1471 |
1208 |
d->MMX_S(0) = s->MMX_S(0);
|
1472 |
1209 |
if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status))
|
1473 |
1210 |
d->MMX_S(1) = s->MMX_S(1);
|
1474 |
1211 |
}
|
1475 |
1212 |
|
1476 |
|
void OPPROTO op_pfmin(void)
|
|
1213 |
void helper_pfmin(MMXReg *d, MMXReg *s)
|
1477 |
1214 |
{
|
1478 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1479 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1480 |
1215 |
if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status))
|
1481 |
1216 |
d->MMX_S(0) = s->MMX_S(0);
|
1482 |
1217 |
if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status))
|
1483 |
1218 |
d->MMX_S(1) = s->MMX_S(1);
|
1484 |
1219 |
}
|
1485 |
1220 |
|
1486 |
|
void OPPROTO op_pfmul(void)
|
|
1221 |
void helper_pfmul(MMXReg *d, MMXReg *s)
|
1487 |
1222 |
{
|
1488 |
|
MMXReg *d = (MMXReg *)((char *)env + PARAM1);
|
1489 |
|
MMXReg *s = (MMXReg *)((char *)env + PARAM2);
|
1490 |
1223 |
d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
|
1491 |
1224 |
d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
|
1492 |
1225 |
}
|