ops_sse.h@ 13480

最後變更在這個檔案從13480是 11982,由 vboxsync 提交於 16 年前
All: license header changes for 2.0 (OSE headers, add Sun GPL/LGPL disclaimer)
屬性 svn:eol-style 設為 `native` 屬性 svn:keywords 設為 `Author Date Id Revision`
檔案大小: 38.0 KB

行
1	/*
2	* MMX/SSE/SSE2/PNI support
3	*
4	* Copyright (c) 2005 Fabrice Bellard
5	*
6	* This library is free software; you can redistribute it and/or
7	* modify it under the terms of the GNU Lesser General Public
8	* License as published by the Free Software Foundation; either
9	* version 2 of the License, or (at your option) any later version.
10	*
11	* This library is distributed in the hope that it will be useful,
12	* but WITHOUT ANY WARRANTY; without even the implied warranty of
13	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14	* Lesser General Public License for more details.
15	*
16	* You should have received a copy of the GNU Lesser General Public
17	* License along with this library; if not, write to the Free Software
18	* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19	*/
20
21	/*
22	* Sun LGPL Disclaimer: For the avoidance of doubt, except that if any license choice
23	* other than GPL or LGPL is available it will apply instead, Sun elects to use only
24	* the Lesser General Public License version 2.1 (LGPLv2) at this time for any software where
25	* a choice of LGPL license versions is made available with the language indicating
26	* that LGPLv2 or any later version may be used, or where a choice of which version
27	* of the LGPL is applied is otherwise unspecified.
28	*/
29	#if SHIFT == 0
30	#define Reg MMXReg
31	#define XMM_ONLY(x...)
32	#define B(n) MMX_B(n)
33	#define W(n) MMX_W(n)
34	#define L(n) MMX_L(n)
35	#define Q(n) q
36	#define SUFFIX _mmx
37	#else
38	#define Reg XMMReg
39	#define XMM_ONLY(x...) x
40	#define B(n) XMM_B(n)
41	#define W(n) XMM_W(n)
42	#define L(n) XMM_L(n)
43	#define Q(n) XMM_Q(n)
44	#define SUFFIX _xmm
45	#endif
46
47	void OPPROTO glue(op_psrlw, SUFFIX)(void)
48	{
49	Reg d, s;
50	int shift;
51
52	d = (Reg )((char )env + PARAM1);
53	s = (Reg )((char )env + PARAM2);
54
55	if (s->Q(0) > 15) {
56	d->Q(0) = 0;
57	#if SHIFT == 1
58	d->Q(1) = 0;
59	#endif
60	} else {
61	shift = s->B(0);
62	d->W(0) >>= shift;
63	d->W(1) >>= shift;
64	d->W(2) >>= shift;
65	d->W(3) >>= shift;
66	#if SHIFT == 1
67	d->W(4) >>= shift;
68	d->W(5) >>= shift;
69	d->W(6) >>= shift;
70	d->W(7) >>= shift;
71	#endif
72	}
73	FORCE_RET();
74	}
75
76	void OPPROTO glue(op_psraw, SUFFIX)(void)
77	{
78	Reg d, s;
79	int shift;
80
81	d = (Reg )((char )env + PARAM1);
82	s = (Reg )((char )env + PARAM2);
83
84	if (s->Q(0) > 15) {
85	shift = 15;
86	} else {
87	shift = s->B(0);
88	}
89	d->W(0) = (int16_t)d->W(0) >> shift;
90	d->W(1) = (int16_t)d->W(1) >> shift;
91	d->W(2) = (int16_t)d->W(2) >> shift;
92	d->W(3) = (int16_t)d->W(3) >> shift;
93	#if SHIFT == 1
94	d->W(4) = (int16_t)d->W(4) >> shift;
95	d->W(5) = (int16_t)d->W(5) >> shift;
96	d->W(6) = (int16_t)d->W(6) >> shift;
97	d->W(7) = (int16_t)d->W(7) >> shift;
98	#endif
99	}
100
101	void OPPROTO glue(op_psllw, SUFFIX)(void)
102	{
103	Reg d, s;
104	int shift;
105
106	d = (Reg )((char )env + PARAM1);
107	s = (Reg )((char )env + PARAM2);
108
109	if (s->Q(0) > 15) {
110	d->Q(0) = 0;
111	#if SHIFT == 1
112	d->Q(1) = 0;
113	#endif
114	} else {
115	shift = s->B(0);
116	d->W(0) <<= shift;
117	d->W(1) <<= shift;
118	d->W(2) <<= shift;
119	d->W(3) <<= shift;
120	#if SHIFT == 1
121	d->W(4) <<= shift;
122	d->W(5) <<= shift;
123	d->W(6) <<= shift;
124	d->W(7) <<= shift;
125	#endif
126	}
127	FORCE_RET();
128	}
129
130	void OPPROTO glue(op_psrld, SUFFIX)(void)
131	{
132	Reg d, s;
133	int shift;
134
135	d = (Reg )((char )env + PARAM1);
136	s = (Reg )((char )env + PARAM2);
137
138	if (s->Q(0) > 31) {
139	d->Q(0) = 0;
140	#if SHIFT == 1
141	d->Q(1) = 0;
142	#endif
143	} else {
144	shift = s->B(0);
145	d->L(0) >>= shift;
146	d->L(1) >>= shift;
147	#if SHIFT == 1
148	d->L(2) >>= shift;
149	d->L(3) >>= shift;
150	#endif
151	}
152	FORCE_RET();
153	}
154
155	void OPPROTO glue(op_psrad, SUFFIX)(void)
156	{
157	Reg d, s;
158	int shift;
159
160	d = (Reg )((char )env + PARAM1);
161	s = (Reg )((char )env + PARAM2);
162
163	if (s->Q(0) > 31) {
164	shift = 31;
165	} else {
166	shift = s->B(0);
167	}
168	d->L(0) = (int32_t)d->L(0) >> shift;
169	d->L(1) = (int32_t)d->L(1) >> shift;
170	#if SHIFT == 1
171	d->L(2) = (int32_t)d->L(2) >> shift;
172	d->L(3) = (int32_t)d->L(3) >> shift;
173	#endif
174	}
175
176	void OPPROTO glue(op_pslld, SUFFIX)(void)
177	{
178	Reg d, s;
179	int shift;
180
181	d = (Reg )((char )env + PARAM1);
182	s = (Reg )((char )env + PARAM2);
183
184	if (s->Q(0) > 31) {
185	d->Q(0) = 0;
186	#if SHIFT == 1
187	d->Q(1) = 0;
188	#endif
189	} else {
190	shift = s->B(0);
191	d->L(0) <<= shift;
192	d->L(1) <<= shift;
193	#if SHIFT == 1
194	d->L(2) <<= shift;
195	d->L(3) <<= shift;
196	#endif
197	}
198	FORCE_RET();
199	}
200
201	void OPPROTO glue(op_psrlq, SUFFIX)(void)
202	{
203	Reg d, s;
204	int shift;
205
206	d = (Reg )((char )env + PARAM1);
207	s = (Reg )((char )env + PARAM2);
208
209	if (s->Q(0) > 63) {
210	d->Q(0) = 0;
211	#if SHIFT == 1
212	d->Q(1) = 0;
213	#endif
214	} else {
215	shift = s->B(0);
216	d->Q(0) >>= shift;
217	#if SHIFT == 1
218	d->Q(1) >>= shift;
219	#endif
220	}
221	FORCE_RET();
222	}
223
224	void OPPROTO glue(op_psllq, SUFFIX)(void)
225	{
226	Reg d, s;
227	int shift;
228
229	d = (Reg )((char )env + PARAM1);
230	s = (Reg )((char )env + PARAM2);
231
232	if (s->Q(0) > 63) {
233	d->Q(0) = 0;
234	#if SHIFT == 1
235	d->Q(1) = 0;
236	#endif
237	} else {
238	shift = s->B(0);
239	d->Q(0) <<= shift;
240	#if SHIFT == 1
241	d->Q(1) <<= shift;
242	#endif
243	}
244	FORCE_RET();
245	}
246
247	#if SHIFT == 1
248	void OPPROTO glue(op_psrldq, SUFFIX)(void)
249	{
250	Reg d, s;
251	int shift, i;
252
253	d = (Reg )((char )env + PARAM1);
254	s = (Reg )((char )env + PARAM2);
255	shift = s->L(0);
256	if (shift > 16)
257	shift = 16;
258	for(i = 0; i < 16 - shift; i++)
259	d->B(i) = d->B(i + shift);
260	for(i = 16 - shift; i < 16; i++)
261	d->B(i) = 0;
262	FORCE_RET();
263	}
264
265	void OPPROTO glue(op_pslldq, SUFFIX)(void)
266	{
267	Reg d, s;
268	int shift, i;
269
270	d = (Reg )((char )env + PARAM1);
271	s = (Reg )((char )env + PARAM2);
272	shift = s->L(0);
273	if (shift > 16)
274	shift = 16;
275	for(i = 15; i >= shift; i--)
276	d->B(i) = d->B(i - shift);
277	for(i = 0; i < shift; i++)
278	d->B(i) = 0;
279	FORCE_RET();
280	}
281	#endif
282
283	#define SSE_OP_B(name, F)\
284	void OPPROTO glue(name, SUFFIX) (void)\
285	{\
286	Reg d, s;\
287	d = (Reg )((char )env + PARAM1);\
288	s = (Reg )((char )env + PARAM2);\
289	d->B(0) = F(d->B(0), s->B(0));\
290	d->B(1) = F(d->B(1), s->B(1));\
291	d->B(2) = F(d->B(2), s->B(2));\
292	d->B(3) = F(d->B(3), s->B(3));\
293	d->B(4) = F(d->B(4), s->B(4));\
294	d->B(5) = F(d->B(5), s->B(5));\
295	d->B(6) = F(d->B(6), s->B(6));\
296	d->B(7) = F(d->B(7), s->B(7));\
297	XMM_ONLY(\
298	d->B(8) = F(d->B(8), s->B(8));\
299	d->B(9) = F(d->B(9), s->B(9));\
300	d->B(10) = F(d->B(10), s->B(10));\
301	d->B(11) = F(d->B(11), s->B(11));\
302	d->B(12) = F(d->B(12), s->B(12));\
303	d->B(13) = F(d->B(13), s->B(13));\
304	d->B(14) = F(d->B(14), s->B(14));\
305	d->B(15) = F(d->B(15), s->B(15));\
306	)\
307	}
308
309	#define SSE_OP_W(name, F)\
310	void OPPROTO glue(name, SUFFIX) (void)\
311	{\
312	Reg d, s;\
313	d = (Reg )((char )env + PARAM1);\
314	s = (Reg )((char )env + PARAM2);\
315	d->W(0) = F(d->W(0), s->W(0));\
316	d->W(1) = F(d->W(1), s->W(1));\
317	d->W(2) = F(d->W(2), s->W(2));\
318	d->W(3) = F(d->W(3), s->W(3));\
319	XMM_ONLY(\
320	d->W(4) = F(d->W(4), s->W(4));\
321	d->W(5) = F(d->W(5), s->W(5));\
322	d->W(6) = F(d->W(6), s->W(6));\
323	d->W(7) = F(d->W(7), s->W(7));\
324	)\
325	}
326
327	#define SSE_OP_L(name, F)\
328	void OPPROTO glue(name, SUFFIX) (void)\
329	{\
330	Reg d, s;\
331	d = (Reg )((char )env + PARAM1);\
332	s = (Reg )((char )env + PARAM2);\
333	d->L(0) = F(d->L(0), s->L(0));\
334	d->L(1) = F(d->L(1), s->L(1));\
335	XMM_ONLY(\
336	d->L(2) = F(d->L(2), s->L(2));\
337	d->L(3) = F(d->L(3), s->L(3));\
338	)\
339	}
340
341	#define SSE_OP_Q(name, F)\
342	void OPPROTO glue(name, SUFFIX) (void)\
343	{\
344	Reg d, s;\
345	d = (Reg )((char )env + PARAM1);\
346	s = (Reg )((char )env + PARAM2);\
347	d->Q(0) = F(d->Q(0), s->Q(0));\
348	XMM_ONLY(\
349	d->Q(1) = F(d->Q(1), s->Q(1));\
350	)\
351	}
352
353	#if SHIFT == 0
354	static inline int satub(int x)
355	{
356	if (x < 0)
357	return 0;
358	else if (x > 255)
359	return 255;
360	else
361	return x;
362	}
363
364	static inline int satuw(int x)
365	{
366	if (x < 0)
367	return 0;
368	else if (x > 65535)
369	return 65535;
370	else
371	return x;
372	}
373
374	static inline int satsb(int x)
375	{
376	if (x < -128)
377	return -128;
378	else if (x > 127)
379	return 127;
380	else
381	return x;
382	}
383
384	static inline int satsw(int x)
385	{
386	if (x < -32768)
387	return -32768;
388	else if (x > 32767)
389	return 32767;
390	else
391	return x;
392	}
393
394	#define FADD(a, b) ((a) + (b))
395	#define FADDUB(a, b) satub((a) + (b))
396	#define FADDUW(a, b) satuw((a) + (b))
397	#define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
398	#define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
399
400	#define FSUB(a, b) ((a) - (b))
401	#define FSUBUB(a, b) satub((a) - (b))
402	#define FSUBUW(a, b) satuw((a) - (b))
403	#define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
404	#define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
405	#define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
406	#define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
407	#define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
408	#define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
409
410	#define FAND(a, b) (a) & (b)
411	#define FANDN(a, b) ((~(a)) & (b))
412	#define FOR(a, b) (a) \| (b)
413	#define FXOR(a, b) (a) ^ (b)
414
415	#define FCMPGTB(a, b) (int8_t)(a) > (int8_t)(b) ? -1 : 0
416	#define FCMPGTW(a, b) (int16_t)(a) > (int16_t)(b) ? -1 : 0
417	#define FCMPGTL(a, b) (int32_t)(a) > (int32_t)(b) ? -1 : 0
418	#define FCMPEQ(a, b) (a) == (b) ? -1 : 0
419
420	#define FMULLW(a, b) (a) * (b)
421	#define FMULHUW(a, b) (a) * (b) >> 16
422	#define FMULHW(a, b) (int16_t)(a) * (int16_t)(b) >> 16
423
424	#define FAVG(a, b) ((a) + (b) + 1) >> 1
425	#endif
426
427	SSE_OP_B(op_paddb, FADD)
428	SSE_OP_W(op_paddw, FADD)
429	SSE_OP_L(op_paddl, FADD)
430	SSE_OP_Q(op_paddq, FADD)
431
432	SSE_OP_B(op_psubb, FSUB)
433	SSE_OP_W(op_psubw, FSUB)
434	SSE_OP_L(op_psubl, FSUB)
435	SSE_OP_Q(op_psubq, FSUB)
436
437	SSE_OP_B(op_paddusb, FADDUB)
438	SSE_OP_B(op_paddsb, FADDSB)
439	SSE_OP_B(op_psubusb, FSUBUB)
440	SSE_OP_B(op_psubsb, FSUBSB)
441
442	SSE_OP_W(op_paddusw, FADDUW)
443	SSE_OP_W(op_paddsw, FADDSW)
444	SSE_OP_W(op_psubusw, FSUBUW)
445	SSE_OP_W(op_psubsw, FSUBSW)
446
447	SSE_OP_B(op_pminub, FMINUB)
448	SSE_OP_B(op_pmaxub, FMAXUB)
449
450	SSE_OP_W(op_pminsw, FMINSW)
451	SSE_OP_W(op_pmaxsw, FMAXSW)
452
453	SSE_OP_Q(op_pand, FAND)
454	SSE_OP_Q(op_pandn, FANDN)
455	SSE_OP_Q(op_por, FOR)
456	SSE_OP_Q(op_pxor, FXOR)
457
458	SSE_OP_B(op_pcmpgtb, FCMPGTB)
459	SSE_OP_W(op_pcmpgtw, FCMPGTW)
460	SSE_OP_L(op_pcmpgtl, FCMPGTL)
461
462	SSE_OP_B(op_pcmpeqb, FCMPEQ)
463	SSE_OP_W(op_pcmpeqw, FCMPEQ)
464	SSE_OP_L(op_pcmpeql, FCMPEQ)
465
466	SSE_OP_W(op_pmullw, FMULLW)
467	SSE_OP_W(op_pmulhuw, FMULHUW)
468	SSE_OP_W(op_pmulhw, FMULHW)
469
470	SSE_OP_B(op_pavgb, FAVG)
471	SSE_OP_W(op_pavgw, FAVG)
472
473	void OPPROTO glue(op_pmuludq, SUFFIX) (void)
474	{
475	Reg d, s;
476	d = (Reg )((char )env + PARAM1);
477	s = (Reg )((char )env + PARAM2);
478
479	d->Q(0) = (uint64_t)s->L(0) * (uint64_t)d->L(0);
480	#if SHIFT == 1
481	d->Q(1) = (uint64_t)s->L(2) * (uint64_t)d->L(2);
482	#endif
483	}
484
485	void OPPROTO glue(op_pmaddwd, SUFFIX) (void)
486	{
487	int i;
488	Reg d, s;
489	d = (Reg )((char )env + PARAM1);
490	s = (Reg )((char )env + PARAM2);
491
492	for(i = 0; i < (2 << SHIFT); i++) {
493	d->L(i) = (int16_t)s->W(2i) (int16_t)d->W(2*i) +
494	(int16_t)s->W(2i+1) (int16_t)d->W(2*i+1);
495	}
496	FORCE_RET();
497	}
498
499	#if SHIFT == 0
500	static inline int abs1(int a)
501	{
502	if (a < 0)
503	return -a;
504	else
505	return a;
506	}
507	#endif
508	void OPPROTO glue(op_psadbw, SUFFIX) (void)
509	{
510	unsigned int val;
511	Reg d, s;
512	d = (Reg )((char )env + PARAM1);
513	s = (Reg )((char )env + PARAM2);
514
515	val = 0;
516	val += abs1(d->B(0) - s->B(0));
517	val += abs1(d->B(1) - s->B(1));
518	val += abs1(d->B(2) - s->B(2));
519	val += abs1(d->B(3) - s->B(3));
520	val += abs1(d->B(4) - s->B(4));
521	val += abs1(d->B(5) - s->B(5));
522	val += abs1(d->B(6) - s->B(6));
523	val += abs1(d->B(7) - s->B(7));
524	d->Q(0) = val;
525	#if SHIFT == 1
526	val = 0;
527	val += abs1(d->B(8) - s->B(8));
528	val += abs1(d->B(9) - s->B(9));
529	val += abs1(d->B(10) - s->B(10));
530	val += abs1(d->B(11) - s->B(11));
531	val += abs1(d->B(12) - s->B(12));
532	val += abs1(d->B(13) - s->B(13));
533	val += abs1(d->B(14) - s->B(14));
534	val += abs1(d->B(15) - s->B(15));
535	d->Q(1) = val;
536	#endif
537	}
538
539	void OPPROTO glue(op_maskmov, SUFFIX) (void)
540	{
541	int i;
542	Reg d, s;
543	d = (Reg )((char )env + PARAM1);
544	s = (Reg )((char )env + PARAM2);
545	for(i = 0; i < (8 << SHIFT); i++) {
546	if (s->B(i) & 0x80)
547	stb(A0 + i, d->B(i));
548	}
549	FORCE_RET();
550	}
551
552	void OPPROTO glue(op_movl_mm_T0, SUFFIX) (void)
553	{
554	Reg *d;
555	d = (Reg )((char )env + PARAM1);
556	d->L(0) = T0;
557	d->L(1) = 0;
558	#if SHIFT == 1
559	d->Q(1) = 0;
560	#endif
561	}
562
563	void OPPROTO glue(op_movl_T0_mm, SUFFIX) (void)
564	{
565	Reg *s;
566	s = (Reg )((char )env + PARAM1);
567	T0 = s->L(0);
568	}
569
570	#ifdef TARGET_X86_64
571	void OPPROTO glue(op_movq_mm_T0, SUFFIX) (void)
572	{
573	Reg *d;
574	d = (Reg )((char )env + PARAM1);
575	d->Q(0) = T0;
576	#if SHIFT == 1
577	d->Q(1) = 0;
578	#endif
579	}
580
581	void OPPROTO glue(op_movq_T0_mm, SUFFIX) (void)
582	{
583	Reg *s;
584	s = (Reg )((char )env + PARAM1);
585	T0 = s->Q(0);
586	}
587	#endif
588
589	#if SHIFT == 0
590	void OPPROTO glue(op_pshufw, SUFFIX) (void)
591	{
592	#if __GCC__ == 3 \|\| defined(RT_ARCH_AMD64) /* VBOX hack in #else */
593	Reg r, d, s;
594	int order;
595	d = (Reg )((char )env + PARAM1);
596	s = (Reg )((char )env + PARAM2);
597	order = PARAM3;
598	r.W(0) = s->W(order & 3);
599	r.W(1) = s->W((order >> 2) & 3);
600	r.W(2) = s->W((order >> 4) & 3);
601	r.W(3) = s->W((order >> 6) & 3);
602	*d = r;
603	#else
604	Reg *s;
605	int order;
606	uint32_t l0, l1;
607	s = (Reg )((char )env + PARAM2);
608	order = PARAM3;
609	l0 = s->W(order & 3);
610	l0 \|= (uint32_t)s->W((order >> 2) & 3) << 16;
611	l1 = s->W((order >> 4) & 3);
612	l1 \|= (uint32_t)s->W((order >> 6) & 3) << 16;
613
614	s = (Reg )((char )env + PARAM1);
615	s->_l[0] = l0;
616	s->_l[1] = l1;
617	#endif
618	}
619	#else
620	void OPPROTO op_shufps(void)
621	{
622	Reg r, d, s;
623	int order;
624	d = (Reg )((char )env + PARAM1);
625	s = (Reg )((char )env + PARAM2);
626	order = PARAM3;
627	r.L(0) = d->L(order & 3);
628	r.L(1) = d->L((order >> 2) & 3);
629	r.L(2) = s->L((order >> 4) & 3);
630	r.L(3) = s->L((order >> 6) & 3);
631	*d = r;
632	}
633
634	void OPPROTO op_shufpd(void)
635	{
636	Reg r, d, s;
637	int order;
638	d = (Reg )((char )env + PARAM1);
639	s = (Reg )((char )env + PARAM2);
640	order = PARAM3;
641	r.Q(0) = d->Q(order & 1);
642	r.Q(1) = s->Q((order >> 1) & 1);
643	*d = r;
644	}
645
646	void OPPROTO glue(op_pshufd, SUFFIX) (void)
647	{
648	Reg r, d, s;
649	int order;
650	d = (Reg )((char )env + PARAM1);
651	s = (Reg )((char )env + PARAM2);
652	order = PARAM3;
653	r.L(0) = s->L(order & 3);
654	r.L(1) = s->L((order >> 2) & 3);
655	r.L(2) = s->L((order >> 4) & 3);
656	r.L(3) = s->L((order >> 6) & 3);
657	*d = r;
658	}
659
660	void OPPROTO glue(op_pshuflw, SUFFIX) (void)
661	{
662	Reg r, d, s;
663	int order;
664	d = (Reg )((char )env + PARAM1);
665	s = (Reg )((char )env + PARAM2);
666	order = PARAM3;
667	r.W(0) = s->W(order & 3);
668	r.W(1) = s->W((order >> 2) & 3);
669	r.W(2) = s->W((order >> 4) & 3);
670	r.W(3) = s->W((order >> 6) & 3);
671	r.Q(1) = s->Q(1);
672	*d = r;
673	}
674
675	void OPPROTO glue(op_pshufhw, SUFFIX) (void)
676	{
677	Reg r, d, s;
678	int order;
679	d = (Reg )((char )env + PARAM1);
680	s = (Reg )((char )env + PARAM2);
681	order = PARAM3;
682	r.Q(0) = s->Q(0);
683	r.W(4) = s->W(4 + (order & 3));
684	r.W(5) = s->W(4 + ((order >> 2) & 3));
685	r.W(6) = s->W(4 + ((order >> 4) & 3));
686	r.W(7) = s->W(4 + ((order >> 6) & 3));
687	*d = r;
688	}
689	#endif
690
691	#if SHIFT == 1
692	/* FPU ops */
693	/* XXX: not accurate */
694
695	#define SSE_OP_S(name, F)\
696	void OPPROTO op_ ## name ## ps (void)\
697	{\
698	Reg d, s;\
699	d = (Reg )((char )env + PARAM1);\
700	s = (Reg )((char )env + PARAM2);\
701	d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
702	d->XMM_S(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
703	d->XMM_S(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
704	d->XMM_S(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
705	}\
706	\
707	void OPPROTO op_ ## name ## ss (void)\
708	{\
709	Reg d, s;\
710	d = (Reg )((char )env + PARAM1);\
711	s = (Reg )((char )env + PARAM2);\
712	d->XMM_S(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
713	}\
714	void OPPROTO op_ ## name ## pd (void)\
715	{\
716	Reg d, s;\
717	d = (Reg )((char )env + PARAM1);\
718	s = (Reg )((char )env + PARAM2);\
719	d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
720	d->XMM_D(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
721	}\
722	\
723	void OPPROTO op_ ## name ## sd (void)\
724	{\
725	Reg d, s;\
726	d = (Reg )((char )env + PARAM1);\
727	s = (Reg )((char )env + PARAM2);\
728	d->XMM_D(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
729	}
730
731	#define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
732	#define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
733	#define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
734	#define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
735	#define FPU_MIN(size, a, b) (a) < (b) ? (a) : (b)
736	#define FPU_MAX(size, a, b) (a) > (b) ? (a) : (b)
737	#define FPU_SQRT(size, a, b) float ## size ## _sqrt(b, &env->sse_status)
738
739	SSE_OP_S(add, FPU_ADD)
740	SSE_OP_S(sub, FPU_SUB)
741	SSE_OP_S(mul, FPU_MUL)
742	SSE_OP_S(div, FPU_DIV)
743	SSE_OP_S(min, FPU_MIN)
744	SSE_OP_S(max, FPU_MAX)
745	SSE_OP_S(sqrt, FPU_SQRT)
746
747
748	/* float to float conversions */
749	void OPPROTO op_cvtps2pd(void)
750	{
751	float32 s0, s1;
752	Reg d, s;
753	d = (Reg )((char )env + PARAM1);
754	s = (Reg )((char )env + PARAM2);
755	s0 = s->XMM_S(0);
756	s1 = s->XMM_S(1);
757	d->XMM_D(0) = float32_to_float64(s0, &env->sse_status);
758	d->XMM_D(1) = float32_to_float64(s1, &env->sse_status);
759	}
760
761	void OPPROTO op_cvtpd2ps(void)
762	{
763	Reg d, s;
764	d = (Reg )((char )env + PARAM1);
765	s = (Reg )((char )env + PARAM2);
766	d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
767	d->XMM_S(1) = float64_to_float32(s->XMM_D(1), &env->sse_status);
768	d->Q(1) = 0;
769	}
770
771	void OPPROTO op_cvtss2sd(void)
772	{
773	Reg d, s;
774	d = (Reg )((char )env + PARAM1);
775	s = (Reg )((char )env + PARAM2);
776	d->XMM_D(0) = float32_to_float64(s->XMM_S(0), &env->sse_status);
777	}
778
779	void OPPROTO op_cvtsd2ss(void)
780	{
781	Reg d, s;
782	d = (Reg )((char )env + PARAM1);
783	s = (Reg )((char )env + PARAM2);
784	d->XMM_S(0) = float64_to_float32(s->XMM_D(0), &env->sse_status);
785	}
786
787	/* integer to float */
788	void OPPROTO op_cvtdq2ps(void)
789	{
790	XMMReg d = (XMMReg )((char *)env + PARAM1);
791	XMMReg s = (XMMReg )((char *)env + PARAM2);
792	d->XMM_S(0) = int32_to_float32(s->XMM_L(0), &env->sse_status);
793	d->XMM_S(1) = int32_to_float32(s->XMM_L(1), &env->sse_status);
794	d->XMM_S(2) = int32_to_float32(s->XMM_L(2), &env->sse_status);
795	d->XMM_S(3) = int32_to_float32(s->XMM_L(3), &env->sse_status);
796	}
797
798	void OPPROTO op_cvtdq2pd(void)
799	{
800	XMMReg d = (XMMReg )((char *)env + PARAM1);
801	XMMReg s = (XMMReg )((char *)env + PARAM2);
802	int32_t l0, l1;
803	l0 = (int32_t)s->XMM_L(0);
804	l1 = (int32_t)s->XMM_L(1);
805	d->XMM_D(0) = int32_to_float64(l0, &env->sse_status);
806	d->XMM_D(1) = int32_to_float64(l1, &env->sse_status);
807	}
808
809	void OPPROTO op_cvtpi2ps(void)
810	{
811	XMMReg d = (Reg )((char *)env + PARAM1);
812	MMXReg s = (MMXReg )((char *)env + PARAM2);
813	d->XMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
814	d->XMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
815	}
816
817	void OPPROTO op_cvtpi2pd(void)
818	{
819	XMMReg d = (Reg )((char *)env + PARAM1);
820	MMXReg s = (MMXReg )((char *)env + PARAM2);
821	d->XMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
822	d->XMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
823	}
824
825	void OPPROTO op_cvtsi2ss(void)
826	{
827	XMMReg d = (Reg )((char *)env + PARAM1);
828	d->XMM_S(0) = int32_to_float32(T0, &env->sse_status);
829	}
830
831	void OPPROTO op_cvtsi2sd(void)
832	{
833	XMMReg d = (Reg )((char *)env + PARAM1);
834	d->XMM_D(0) = int32_to_float64(T0, &env->sse_status);
835	}
836
837	#ifdef TARGET_X86_64
838	void OPPROTO op_cvtsq2ss(void)
839	{
840	XMMReg d = (Reg )((char *)env + PARAM1);
841	d->XMM_S(0) = int64_to_float32(T0, &env->sse_status);
842	}
843
844	void OPPROTO op_cvtsq2sd(void)
845	{
846	XMMReg d = (Reg )((char *)env + PARAM1);
847	d->XMM_D(0) = int64_to_float64(T0, &env->sse_status);
848	}
849	#endif
850
851	/* float to integer */
852	void OPPROTO op_cvtps2dq(void)
853	{
854	XMMReg d = (XMMReg )((char *)env + PARAM1);
855	XMMReg s = (XMMReg )((char *)env + PARAM2);
856	d->XMM_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
857	d->XMM_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
858	d->XMM_L(2) = float32_to_int32(s->XMM_S(2), &env->sse_status);
859	d->XMM_L(3) = float32_to_int32(s->XMM_S(3), &env->sse_status);
860	}
861
862	void OPPROTO op_cvtpd2dq(void)
863	{
864	XMMReg d = (XMMReg )((char *)env + PARAM1);
865	XMMReg s = (XMMReg )((char *)env + PARAM2);
866	d->XMM_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
867	d->XMM_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
868	d->XMM_Q(1) = 0;
869	}
870
871	void OPPROTO op_cvtps2pi(void)
872	{
873	MMXReg d = (MMXReg )((char *)env + PARAM1);
874	XMMReg s = (XMMReg )((char *)env + PARAM2);
875	d->MMX_L(0) = float32_to_int32(s->XMM_S(0), &env->sse_status);
876	d->MMX_L(1) = float32_to_int32(s->XMM_S(1), &env->sse_status);
877	}
878
879	void OPPROTO op_cvtpd2pi(void)
880	{
881	MMXReg d = (MMXReg )((char *)env + PARAM1);
882	XMMReg s = (XMMReg )((char *)env + PARAM2);
883	d->MMX_L(0) = float64_to_int32(s->XMM_D(0), &env->sse_status);
884	d->MMX_L(1) = float64_to_int32(s->XMM_D(1), &env->sse_status);
885	}
886
887	void OPPROTO op_cvtss2si(void)
888	{
889	XMMReg s = (XMMReg )((char *)env + PARAM1);
890	T0 = float32_to_int32(s->XMM_S(0), &env->sse_status);
891	}
892
893	void OPPROTO op_cvtsd2si(void)
894	{
895	XMMReg s = (XMMReg )((char *)env + PARAM1);
896	T0 = float64_to_int32(s->XMM_D(0), &env->sse_status);
897	}
898
899	#ifdef TARGET_X86_64
900	void OPPROTO op_cvtss2sq(void)
901	{
902	XMMReg s = (XMMReg )((char *)env + PARAM1);
903	T0 = float32_to_int64(s->XMM_S(0), &env->sse_status);
904	}
905
906	void OPPROTO op_cvtsd2sq(void)
907	{
908	XMMReg s = (XMMReg )((char *)env + PARAM1);
909	T0 = float64_to_int64(s->XMM_D(0), &env->sse_status);
910	}
911	#endif
912
913	/* float to integer truncated */
914	void OPPROTO op_cvttps2dq(void)
915	{
916	XMMReg d = (XMMReg )((char *)env + PARAM1);
917	XMMReg s = (XMMReg )((char *)env + PARAM2);
918	d->XMM_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
919	d->XMM_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
920	d->XMM_L(2) = float32_to_int32_round_to_zero(s->XMM_S(2), &env->sse_status);
921	d->XMM_L(3) = float32_to_int32_round_to_zero(s->XMM_S(3), &env->sse_status);
922	}
923
924	void OPPROTO op_cvttpd2dq(void)
925	{
926	XMMReg d = (XMMReg )((char *)env + PARAM1);
927	XMMReg s = (XMMReg )((char *)env + PARAM2);
928	d->XMM_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
929	d->XMM_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
930	d->XMM_Q(1) = 0;
931	}
932
933	void OPPROTO op_cvttps2pi(void)
934	{
935	MMXReg d = (MMXReg )((char *)env + PARAM1);
936	XMMReg s = (XMMReg )((char *)env + PARAM2);
937	d->MMX_L(0) = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
938	d->MMX_L(1) = float32_to_int32_round_to_zero(s->XMM_S(1), &env->sse_status);
939	}
940
941	void OPPROTO op_cvttpd2pi(void)
942	{
943	MMXReg d = (MMXReg )((char *)env + PARAM1);
944	XMMReg s = (XMMReg )((char *)env + PARAM2);
945	d->MMX_L(0) = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
946	d->MMX_L(1) = float64_to_int32_round_to_zero(s->XMM_D(1), &env->sse_status);
947	}
948
949	void OPPROTO op_cvttss2si(void)
950	{
951	XMMReg s = (XMMReg )((char *)env + PARAM1);
952	T0 = float32_to_int32_round_to_zero(s->XMM_S(0), &env->sse_status);
953	}
954
955	void OPPROTO op_cvttsd2si(void)
956	{
957	XMMReg s = (XMMReg )((char *)env + PARAM1);
958	T0 = float64_to_int32_round_to_zero(s->XMM_D(0), &env->sse_status);
959	}
960
961	#ifdef TARGET_X86_64
962	void OPPROTO op_cvttss2sq(void)
963	{
964	XMMReg s = (XMMReg )((char *)env + PARAM1);
965	T0 = float32_to_int64_round_to_zero(s->XMM_S(0), &env->sse_status);
966	}
967
968	void OPPROTO op_cvttsd2sq(void)
969	{
970	XMMReg s = (XMMReg )((char *)env + PARAM1);
971	T0 = float64_to_int64_round_to_zero(s->XMM_D(0), &env->sse_status);
972	}
973	#endif
974
975	void OPPROTO op_rsqrtps(void)
976	{
977	XMMReg d = (XMMReg )((char *)env + PARAM1);
978	XMMReg s = (XMMReg )((char *)env + PARAM2);
979	d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
980	d->XMM_S(1) = approx_rsqrt(s->XMM_S(1));
981	d->XMM_S(2) = approx_rsqrt(s->XMM_S(2));
982	d->XMM_S(3) = approx_rsqrt(s->XMM_S(3));
983	}
984
985	void OPPROTO op_rsqrtss(void)
986	{
987	XMMReg d = (XMMReg )((char *)env + PARAM1);
988	XMMReg s = (XMMReg )((char *)env + PARAM2);
989	d->XMM_S(0) = approx_rsqrt(s->XMM_S(0));
990	}
991
992	void OPPROTO op_rcpps(void)
993	{
994	XMMReg d = (XMMReg )((char *)env + PARAM1);
995	XMMReg s = (XMMReg )((char *)env + PARAM2);
996	d->XMM_S(0) = approx_rcp(s->XMM_S(0));
997	d->XMM_S(1) = approx_rcp(s->XMM_S(1));
998	d->XMM_S(2) = approx_rcp(s->XMM_S(2));
999	d->XMM_S(3) = approx_rcp(s->XMM_S(3));
1000	}
1001
1002	void OPPROTO op_rcpss(void)
1003	{
1004	XMMReg d = (XMMReg )((char *)env + PARAM1);
1005	XMMReg s = (XMMReg )((char *)env + PARAM2);
1006	d->XMM_S(0) = approx_rcp(s->XMM_S(0));
1007	}
1008
1009	void OPPROTO op_haddps(void)
1010	{
1011	XMMReg d = (XMMReg )((char *)env + PARAM1);
1012	XMMReg s = (XMMReg )((char *)env + PARAM2);
1013	XMMReg r;
1014	r.XMM_S(0) = d->XMM_S(0) + d->XMM_S(1);
1015	r.XMM_S(1) = d->XMM_S(2) + d->XMM_S(3);
1016	r.XMM_S(2) = s->XMM_S(0) + s->XMM_S(1);
1017	r.XMM_S(3) = s->XMM_S(2) + s->XMM_S(3);
1018	*d = r;
1019	}
1020
1021	void OPPROTO op_haddpd(void)
1022	{
1023	XMMReg d = (XMMReg )((char *)env + PARAM1);
1024	XMMReg s = (XMMReg )((char *)env + PARAM2);
1025	XMMReg r;
1026	r.XMM_D(0) = d->XMM_D(0) + d->XMM_D(1);
1027	r.XMM_D(1) = s->XMM_D(0) + s->XMM_D(1);
1028	*d = r;
1029	}
1030
1031	void OPPROTO op_hsubps(void)
1032	{
1033	XMMReg d = (XMMReg )((char *)env + PARAM1);
1034	XMMReg s = (XMMReg )((char *)env + PARAM2);
1035	XMMReg r;
1036	r.XMM_S(0) = d->XMM_S(0) - d->XMM_S(1);
1037	r.XMM_S(1) = d->XMM_S(2) - d->XMM_S(3);
1038	r.XMM_S(2) = s->XMM_S(0) - s->XMM_S(1);
1039	r.XMM_S(3) = s->XMM_S(2) - s->XMM_S(3);
1040	*d = r;
1041	}
1042
1043	void OPPROTO op_hsubpd(void)
1044	{
1045	XMMReg d = (XMMReg )((char *)env + PARAM1);
1046	XMMReg s = (XMMReg )((char *)env + PARAM2);
1047	XMMReg r;
1048	r.XMM_D(0) = d->XMM_D(0) - d->XMM_D(1);
1049	r.XMM_D(1) = s->XMM_D(0) - s->XMM_D(1);
1050	*d = r;
1051	}
1052
1053	void OPPROTO op_addsubps(void)
1054	{
1055	XMMReg d = (XMMReg )((char *)env + PARAM1);
1056	XMMReg s = (XMMReg )((char *)env + PARAM2);
1057	d->XMM_S(0) = d->XMM_S(0) - s->XMM_S(0);
1058	d->XMM_S(1) = d->XMM_S(1) + s->XMM_S(1);
1059	d->XMM_S(2) = d->XMM_S(2) - s->XMM_S(2);
1060	d->XMM_S(3) = d->XMM_S(3) + s->XMM_S(3);
1061	}
1062
1063	void OPPROTO op_addsubpd(void)
1064	{
1065	XMMReg d = (XMMReg )((char *)env + PARAM1);
1066	XMMReg s = (XMMReg )((char *)env + PARAM2);
1067	d->XMM_D(0) = d->XMM_D(0) - s->XMM_D(0);
1068	d->XMM_D(1) = d->XMM_D(1) + s->XMM_D(1);
1069	}
1070
1071	/* XXX: unordered */
1072	#define SSE_OP_CMP(name, F)\
1073	void OPPROTO op_ ## name ## ps (void)\
1074	{\
1075	Reg d, s;\
1076	d = (Reg )((char )env + PARAM1);\
1077	s = (Reg )((char )env + PARAM2);\
1078	d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
1079	d->XMM_L(1) = F(32, d->XMM_S(1), s->XMM_S(1));\
1080	d->XMM_L(2) = F(32, d->XMM_S(2), s->XMM_S(2));\
1081	d->XMM_L(3) = F(32, d->XMM_S(3), s->XMM_S(3));\
1082	}\
1083	\
1084	void OPPROTO op_ ## name ## ss (void)\
1085	{\
1086	Reg d, s;\
1087	d = (Reg )((char )env + PARAM1);\
1088	s = (Reg )((char )env + PARAM2);\
1089	d->XMM_L(0) = F(32, d->XMM_S(0), s->XMM_S(0));\
1090	}\
1091	void OPPROTO op_ ## name ## pd (void)\
1092	{\
1093	Reg d, s;\
1094	d = (Reg )((char )env + PARAM1);\
1095	s = (Reg )((char )env + PARAM2);\
1096	d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
1097	d->XMM_Q(1) = F(64, d->XMM_D(1), s->XMM_D(1));\
1098	}\
1099	\
1100	void OPPROTO op_ ## name ## sd (void)\
1101	{\
1102	Reg d, s;\
1103	d = (Reg )((char )env + PARAM1);\
1104	s = (Reg )((char )env + PARAM2);\
1105	d->XMM_Q(0) = F(64, d->XMM_D(0), s->XMM_D(0));\
1106	}
1107
1108	#define FPU_CMPEQ(size, a, b) float ## size ## _eq(a, b, &env->sse_status) ? -1 : 0
1109	#define FPU_CMPLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? -1 : 0
1110	#define FPU_CMPLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? -1 : 0
1111	#define FPU_CMPUNORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? - 1 : 0
1112	#define FPU_CMPNEQ(size, a, b) float ## size ## _eq(a, b, &env->sse_status) ? 0 : -1
1113	#define FPU_CMPNLT(size, a, b) float ## size ## _lt(a, b, &env->sse_status) ? 0 : -1
1114	#define FPU_CMPNLE(size, a, b) float ## size ## _le(a, b, &env->sse_status) ? 0 : -1
1115	#define FPU_CMPORD(size, a, b) float ## size ## _unordered(a, b, &env->sse_status) ? 0 : -1
1116
1117	SSE_OP_CMP(cmpeq, FPU_CMPEQ)
1118	SSE_OP_CMP(cmplt, FPU_CMPLT)
1119	SSE_OP_CMP(cmple, FPU_CMPLE)
1120	SSE_OP_CMP(cmpunord, FPU_CMPUNORD)
1121	SSE_OP_CMP(cmpneq, FPU_CMPNEQ)
1122	SSE_OP_CMP(cmpnlt, FPU_CMPNLT)
1123	SSE_OP_CMP(cmpnle, FPU_CMPNLE)
1124	SSE_OP_CMP(cmpord, FPU_CMPORD)
1125
1126	const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z \| CC_P \| CC_C};
1127
1128	void OPPROTO op_ucomiss(void)
1129	{
1130	int ret;
1131	float32 s0, s1;
1132	Reg d, s;
1133	d = (Reg )((char )env + PARAM1);
1134	s = (Reg )((char )env + PARAM2);
1135
1136	s0 = d->XMM_S(0);
1137	s1 = s->XMM_S(0);
1138	ret = float32_compare_quiet(s0, s1, &env->sse_status);
1139	CC_SRC = comis_eflags[ret + 1];
1140	FORCE_RET();
1141	}
1142
1143	void OPPROTO op_comiss(void)
1144	{
1145	int ret;
1146	float32 s0, s1;
1147	Reg d, s;
1148	d = (Reg )((char )env + PARAM1);
1149	s = (Reg )((char )env + PARAM2);
1150
1151	s0 = d->XMM_S(0);
1152	s1 = s->XMM_S(0);
1153	ret = float32_compare(s0, s1, &env->sse_status);
1154	CC_SRC = comis_eflags[ret + 1];
1155	FORCE_RET();
1156	}
1157
1158	void OPPROTO op_ucomisd(void)
1159	{
1160	int ret;
1161	float64 d0, d1;
1162	Reg d, s;
1163	d = (Reg )((char )env + PARAM1);
1164	s = (Reg )((char )env + PARAM2);
1165
1166	d0 = d->XMM_D(0);
1167	d1 = s->XMM_D(0);
1168	ret = float64_compare_quiet(d0, d1, &env->sse_status);
1169	CC_SRC = comis_eflags[ret + 1];
1170	FORCE_RET();
1171	}
1172
1173	void OPPROTO op_comisd(void)
1174	{
1175	int ret;
1176	float64 d0, d1;
1177	Reg d, s;
1178	d = (Reg )((char )env + PARAM1);
1179	s = (Reg )((char )env + PARAM2);
1180
1181	d0 = d->XMM_D(0);
1182	d1 = s->XMM_D(0);
1183	ret = float64_compare(d0, d1, &env->sse_status);
1184	CC_SRC = comis_eflags[ret + 1];
1185	FORCE_RET();
1186	}
1187
1188	void OPPROTO op_movmskps(void)
1189	{
1190	int b0, b1, b2, b3;
1191	Reg *s;
1192	s = (Reg )((char )env + PARAM1);
1193	b0 = s->XMM_L(0) >> 31;
1194	b1 = s->XMM_L(1) >> 31;
1195	b2 = s->XMM_L(2) >> 31;
1196	b3 = s->XMM_L(3) >> 31;
1197	T0 = b0 \| (b1 << 1) \| (b2 << 2) \| (b3 << 3);
1198	}
1199
1200	void OPPROTO op_movmskpd(void)
1201	{
1202	int b0, b1;
1203	Reg *s;
1204	s = (Reg )((char )env + PARAM1);
1205	b0 = s->XMM_L(1) >> 31;
1206	b1 = s->XMM_L(3) >> 31;
1207	T0 = b0 \| (b1 << 1);
1208	}
1209
1210	#endif
1211
1212	void OPPROTO glue(op_pmovmskb, SUFFIX)(void)
1213	{
1214	Reg *s;
1215	s = (Reg )((char )env + PARAM1);
1216	T0 = 0;
1217	T0 \|= (s->XMM_B(0) >> 7);
1218	T0 \|= (s->XMM_B(1) >> 6) & 0x02;
1219	T0 \|= (s->XMM_B(2) >> 5) & 0x04;
1220	T0 \|= (s->XMM_B(3) >> 4) & 0x08;
1221	T0 \|= (s->XMM_B(4) >> 3) & 0x10;
1222	T0 \|= (s->XMM_B(5) >> 2) & 0x20;
1223	T0 \|= (s->XMM_B(6) >> 1) & 0x40;
1224	T0 \|= (s->XMM_B(7)) & 0x80;
1225	#if SHIFT == 1
1226	T0 \|= (s->XMM_B(8) << 1) & 0x0100;
1227	T0 \|= (s->XMM_B(9) << 2) & 0x0200;
1228	T0 \|= (s->XMM_B(10) << 3) & 0x0400;
1229	T0 \|= (s->XMM_B(11) << 4) & 0x0800;
1230	T0 \|= (s->XMM_B(12) << 5) & 0x1000;
1231	T0 \|= (s->XMM_B(13) << 6) & 0x2000;
1232	T0 \|= (s->XMM_B(14) << 7) & 0x4000;
1233	T0 \|= (s->XMM_B(15) << 8) & 0x8000;
1234	#endif
1235	}
1236
1237	void OPPROTO glue(op_pinsrw, SUFFIX) (void)
1238	{
1239	Reg d = (Reg )((char *)env + PARAM1);
1240	int pos = PARAM2;
1241
1242	d->W(pos) = T0;
1243	}
1244
1245	void OPPROTO glue(op_pextrw, SUFFIX) (void)
1246	{
1247	Reg s = (Reg )((char *)env + PARAM1);
1248	int pos = PARAM2;
1249
1250	T0 = s->W(pos);
1251	}
1252
1253	void OPPROTO glue(op_packsswb, SUFFIX) (void)
1254	{
1255	Reg r, d, s;
1256	d = (Reg )((char )env + PARAM1);
1257	s = (Reg )((char )env + PARAM2);
1258
1259	r.B(0) = satsb((int16_t)d->W(0));
1260	r.B(1) = satsb((int16_t)d->W(1));
1261	r.B(2) = satsb((int16_t)d->W(2));
1262	r.B(3) = satsb((int16_t)d->W(3));
1263	#if SHIFT == 1
1264	r.B(4) = satsb((int16_t)d->W(4));
1265	r.B(5) = satsb((int16_t)d->W(5));
1266	r.B(6) = satsb((int16_t)d->W(6));
1267	r.B(7) = satsb((int16_t)d->W(7));
1268	#endif
1269	r.B((4 << SHIFT) + 0) = satsb((int16_t)s->W(0));
1270	r.B((4 << SHIFT) + 1) = satsb((int16_t)s->W(1));
1271	r.B((4 << SHIFT) + 2) = satsb((int16_t)s->W(2));
1272	r.B((4 << SHIFT) + 3) = satsb((int16_t)s->W(3));
1273	#if SHIFT == 1
1274	r.B(12) = satsb((int16_t)s->W(4));
1275	r.B(13) = satsb((int16_t)s->W(5));
1276	r.B(14) = satsb((int16_t)s->W(6));
1277	r.B(15) = satsb((int16_t)s->W(7));
1278	#endif
1279	*d = r;
1280	}
1281
1282	void OPPROTO glue(op_packuswb, SUFFIX) (void)
1283	{
1284	Reg r, d, s;
1285	d = (Reg )((char )env + PARAM1);
1286	s = (Reg )((char )env + PARAM2);
1287
1288	r.B(0) = satub((int16_t)d->W(0));
1289	r.B(1) = satub((int16_t)d->W(1));
1290	r.B(2) = satub((int16_t)d->W(2));
1291	r.B(3) = satub((int16_t)d->W(3));
1292	#if SHIFT == 1
1293	r.B(4) = satub((int16_t)d->W(4));
1294	r.B(5) = satub((int16_t)d->W(5));
1295	r.B(6) = satub((int16_t)d->W(6));
1296	r.B(7) = satub((int16_t)d->W(7));
1297	#endif
1298	r.B((4 << SHIFT) + 0) = satub((int16_t)s->W(0));
1299	r.B((4 << SHIFT) + 1) = satub((int16_t)s->W(1));
1300	r.B((4 << SHIFT) + 2) = satub((int16_t)s->W(2));
1301	r.B((4 << SHIFT) + 3) = satub((int16_t)s->W(3));
1302	#if SHIFT == 1
1303	r.B(12) = satub((int16_t)s->W(4));
1304	r.B(13) = satub((int16_t)s->W(5));
1305	r.B(14) = satub((int16_t)s->W(6));
1306	r.B(15) = satub((int16_t)s->W(7));
1307	#endif
1308	*d = r;
1309	}
1310
1311	void OPPROTO glue(op_packssdw, SUFFIX) (void)
1312	{
1313	Reg r, d, s;
1314	d = (Reg )((char )env + PARAM1);
1315	s = (Reg )((char )env + PARAM2);
1316
1317	r.W(0) = satsw(d->L(0));
1318	r.W(1) = satsw(d->L(1));
1319	#if SHIFT == 1
1320	r.W(2) = satsw(d->L(2));
1321	r.W(3) = satsw(d->L(3));
1322	#endif
1323	r.W((2 << SHIFT) + 0) = satsw(s->L(0));
1324	r.W((2 << SHIFT) + 1) = satsw(s->L(1));
1325	#if SHIFT == 1
1326	r.W(6) = satsw(s->L(2));
1327	r.W(7) = satsw(s->L(3));
1328	#endif
1329	*d = r;
1330	}
1331
1332	#define UNPCK_OP(base_name, base) \
1333	\
1334	void OPPROTO glue(op_punpck ## base_name ## bw, SUFFIX) (void) \
1335	{ \
1336	Reg r, d, s; \
1337	d = (Reg )((char )env + PARAM1); \
1338	s = (Reg )((char )env + PARAM2); \
1339	\
1340	r.B(0) = d->B((base << (SHIFT + 2)) + 0); \
1341	r.B(1) = s->B((base << (SHIFT + 2)) + 0); \
1342	r.B(2) = d->B((base << (SHIFT + 2)) + 1); \
1343	r.B(3) = s->B((base << (SHIFT + 2)) + 1); \
1344	r.B(4) = d->B((base << (SHIFT + 2)) + 2); \
1345	r.B(5) = s->B((base << (SHIFT + 2)) + 2); \
1346	r.B(6) = d->B((base << (SHIFT + 2)) + 3); \
1347	r.B(7) = s->B((base << (SHIFT + 2)) + 3); \
1348	XMM_ONLY( \
1349	r.B(8) = d->B((base << (SHIFT + 2)) + 4); \
1350	r.B(9) = s->B((base << (SHIFT + 2)) + 4); \
1351	r.B(10) = d->B((base << (SHIFT + 2)) + 5); \
1352	r.B(11) = s->B((base << (SHIFT + 2)) + 5); \
1353	r.B(12) = d->B((base << (SHIFT + 2)) + 6); \
1354	r.B(13) = s->B((base << (SHIFT + 2)) + 6); \
1355	r.B(14) = d->B((base << (SHIFT + 2)) + 7); \
1356	r.B(15) = s->B((base << (SHIFT + 2)) + 7); \
1357	) \
1358	*d = r; \
1359	} \
1360	\
1361	void OPPROTO glue(op_punpck ## base_name ## wd, SUFFIX) (void) \
1362	{ \
1363	Reg r, d, s; \
1364	d = (Reg )((char )env + PARAM1); \
1365	s = (Reg )((char )env + PARAM2); \
1366	\
1367	r.W(0) = d->W((base << (SHIFT + 1)) + 0); \
1368	r.W(1) = s->W((base << (SHIFT + 1)) + 0); \
1369	r.W(2) = d->W((base << (SHIFT + 1)) + 1); \
1370	r.W(3) = s->W((base << (SHIFT + 1)) + 1); \
1371	XMM_ONLY( \
1372	r.W(4) = d->W((base << (SHIFT + 1)) + 2); \
1373	r.W(5) = s->W((base << (SHIFT + 1)) + 2); \
1374	r.W(6) = d->W((base << (SHIFT + 1)) + 3); \
1375	r.W(7) = s->W((base << (SHIFT + 1)) + 3); \
1376	) \
1377	*d = r; \
1378	} \
1379	\
1380	void OPPROTO glue(op_punpck ## base_name ## dq, SUFFIX) (void) \
1381	{ \
1382	Reg r, d, s; \
1383	d = (Reg )((char )env + PARAM1); \
1384	s = (Reg )((char )env + PARAM2); \
1385	\
1386	r.L(0) = d->L((base << SHIFT) + 0); \
1387	r.L(1) = s->L((base << SHIFT) + 0); \
1388	XMM_ONLY( \
1389	r.L(2) = d->L((base << SHIFT) + 1); \
1390	r.L(3) = s->L((base << SHIFT) + 1); \
1391	) \
1392	*d = r; \
1393	} \
1394	\
1395	XMM_ONLY( \
1396	void OPPROTO glue(op_punpck ## base_name ## qdq, SUFFIX) (void) \
1397	{ \
1398	Reg r, d, s; \
1399	d = (Reg )((char )env + PARAM1); \
1400	s = (Reg )((char )env + PARAM2); \
1401	\
1402	r.Q(0) = d->Q(base); \
1403	r.Q(1) = s->Q(base); \
1404	*d = r; \
1405	} \
1406	)
1407
1408	UNPCK_OP(l, 0)
1409	UNPCK_OP(h, 1)
1410
1411	#undef SHIFT
1412	#undef XMM_ONLY
1413	#undef Reg
1414	#undef B
1415	#undef W
1416	#undef L
1417	#undef Q
1418	#undef SUFFIX

注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

source: vbox/trunk/src/recompiler/target-i386/ops_sse.h@ 13480

以其他格式下載: