Attachment #51031 for bug #80685

Lines 3-15 Link Here

(-)xc-orig/programs/Xserver/fb/Imakefile (-2 / +11 lines)
3	XCOMM	3	XCOMM
4	XCOMM Id: Imakefile,v 1.1 1999/11/02 03:54:44 keithp Exp $	4	XCOMM Id: Imakefile,v 1.1 1999/11/02 03:54:44 keithp Exp $
5		5
6	#if defined(i386Architecture) && defined(HasGcc34) && HasGcc34	6	#if defined(HasGcc34) && HasGcc34
7	MMXOPTIONS= -mmmx -Winline --param inline-unit-growth=10000 \	7	MMXOPTIONS= -mmmx -Winline --param inline-unit-growth=10000 \
8	--param large-function-growth=10000 -DUSE_GCC34_MMX	8	--param large-function-growth=10000 -DUSE_MMX
		9	SSEOPTIONS= $(MMXOPTIONS) -msse -DUSE_SSE
9		10
		11	#if defined(i386Architecture)
10	SpecialCObjectRule(fbmmx,fbmmx.c,$(MMXOPTIONS))	12	SpecialCObjectRule(fbmmx,fbmmx.c,$(MMXOPTIONS))
		13	#elif defined(AMD64Architecture)
		14	SpecialCObjectRule(fbmmx,fbmmx.c,$(SSEOPTIONS))
		15	#endif
		16
		17	#if defined(i386Architecture) \|\| defined(AMD64Architecture)
11	SpecialCObjectRule(fbpict,fbpict.c,$(MMXOPTIONS))	18	SpecialCObjectRule(fbpict,fbpict.c,$(MMXOPTIONS))
12	SpecialCObjectRule(fbfill,fbfill.c,$(MMXOPTIONS))	19	SpecialCObjectRule(fbfill,fbfill.c,$(MMXOPTIONS))
		20	SpecialCObjectRule(fbcopy,fbcopy.c,$(MMXOPTIONS))
		21	#endif
13		22
14	#endif	23	#endif
15		24




/*
 * $XdotOrg: xc/programs/Xserver/fb/fbcompose.c,v 1.5 2005/01/13 20:49:21 sandmann Exp $
 * $XFree86: xc/programs/Xserver/fb/fbcompose.c,v 1.17tsi Exp $
 *
 * Copyright Â© 2000 Keith Packard, member of The XFree86 Project, Inc.
 *
 * Permission to use, copy, modify, distribute, and sell this software and its
 * documentation for any purpose is hereby granted without fee, provided that

    op->u.transform.y = y - op->u.transform.top_y;
}


Bool
fbBuildCompositeOperand (PicturePtr	    pPict,
			 FbCompositeOperand op[4],

	
	op->u.transform.top_y = pPict->pDrawable->y;
	op->u.transform.left_x = pPict->pDrawable->x;
	
	op->u.transform.start_x = x - op->u.transform.left_x;
	op->u.transform.x = op->u.transform.start_x;
	op->u.transform.y = y - op->u.transform.top_y;

    FbCombineFunc	f;
    int			w;

#if 0
    ErrorF ("op:		%d\n"
	    "src format:	%lx\n"
	    "msk format		%lx\n"
	    "dst format		%lx\n"
	    "width:		%d\n"
	    "height		%d\n",
	    op,
	    pSrc? pSrc->format : 0,
	    pMask? pMask->format : 0,
	    pDst? pDst->format : 0,
	    width, height);
    ErrorF ("PICT_x8r8g8b8: %lx\n", PICT_x8r8g8b8);
#endif
    
    if (!fbBuildCompositeOperand (pSrc, src, xSrc, ySrc, TRUE, TRUE))
	return;
    if (!fbBuildCompositeOperand (pDst, dst, xDst, yDst, FALSE, TRUE))




/*
 * Id: fbcopy.c,v 1.1 1999/11/02 03:54:45 keithp Exp $
 *
 * Copyright Â© 1998 Keith Packard
 *
 * Permission to use, copy, modify, distribute, and sell this software and its
 * documentation for any purpose is hereby granted without fee, provided that

#ifdef IN_MODULE
#include "xf86_ansic.h"
#endif
#include "fbmmx.h"

void
fbCopyNtoN (DrawablePtr	pSrcDrawable,

    
    fbGetDrawable (pSrcDrawable, src, srcStride, srcBpp, srcXoff, srcYoff);
    fbGetDrawable (pDstDrawable, dst, dstStride, dstBpp, dstXoff, dstYoff);

    while (nbox--)
    {
#ifdef USE_MMX
	if (!reverse && !upsidedown && fbHaveMMX())
	{
	    if (!fbCopyAreammx (pSrcDrawable,
				pDstDrawable,
				
				(pbox->x1 + dx + srcXoff),
				(pbox->y1 + dy + srcYoff),
				
				(pbox->x1 + dstXoff),
				(pbox->y1 + dstYoff),
				
				(pbox->x2 - pbox->x1),
				(pbox->y2 - pbox->y1)))
		goto fallback;
	    else
		goto next;
	}
    fallback:
#endif
	fbBlt (src + (pbox->y1 + dy + srcYoff) * srcStride,
	       srcStride,
	       (pbox->x1 + dx + srcXoff) * srcBpp,
	       
	       dst + (pbox->y1 + dstYoff) * dstStride,
	       dstStride,
	       (pbox->x1 + dstXoff) * dstBpp,
	       
	       (pbox->x2 - pbox->x1) * dstBpp,
	       (pbox->y2 - pbox->y1),
	       
	       alu,
	       pm,
	       dstBpp,
	       
	       reverse,
	       upsidedown);
#ifdef USE_MMX
    next:
#endif
	pbox++;
    }    
}

void

	    int		yOut)
{
    fbCopyProc	copy;

#ifdef FB_24_32BIT
    if (pSrcDrawable->bitsPerPixel != pDstDrawable->bitsPerPixel)
	copy = fb24_32CopyMtoN;

Lines 1-7 Link Here

(-)xc-orig/programs/Xserver/fb/fbfill.c (-2 / +2 lines)
1	/*	1	/*
2	* Id: fbfill.c,v 1.1 1999/11/02 03:54:45 keithp Exp $	2	* Id: fbfill.c,v 1.1 1999/11/02 03:54:45 keithp Exp $
3	*	3	*
4	* Copyright © 1998 Keith Packard	4	* Copyright Â© 1998 Keith Packard
5	*	5	*
6	* Permission to use, copy, modify, distribute, and sell this software and its	6	* Permission to use, copy, modify, distribute, and sell this software and its
7	* documentation for any purpose is hereby granted without fee, provided that	7	* documentation for any purpose is hereby granted without fee, provided that
Lines 44-50 Link Here
44		44
45	switch (pGC->fillStyle) {	45	switch (pGC->fillStyle) {
46	case FillSolid:	46	case FillSolid:
47	#ifdef USE_GCC34_MMX	47	#ifdef USE_MMX
48	if (!pPriv->and && fbHaveMMX())	48	if (!pPriv->and && fbHaveMMX())
49	if (fbSolidFillmmx (pDrawable, x, y, width, height, pPriv->xor))	49	if (fbSolidFillmmx (pDrawable, x, y, width, height, pPriv->xor))
50	return;	50	return;

Lines 1-5 Link Here

(-)xc-orig/programs/Xserver/fb/fbmmx.c (-353 / +587 lines)
1	/*	1	/*
2	* Copyright © 2004 Red Hat, Inc.	2	* Copyright Â© 2004 Red Hat, Inc.
		3	* Copyright Â© 2004 Nicholas Miell
3	*	4	*
4	* Permission to use, copy, modify, distribute, and sell this software and its	5	* Permission to use, copy, modify, distribute, and sell this software and its
5	* documentation for any purpose is hereby granted without fee, provided that	6	* documentation for any purpose is hereby granted without fee, provided that
Lines 18-31 Link Here
18	* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN	19	* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
19	* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.	20	* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
20	*	21	*
21	* Author: Søren Sandmann (sandmann@redhat.com)	22	* Author: SÃ¸ren Sandmann (sandmann@redhat.com)
22	*	23	* Minor Improvements: Nicholas Miell (nmiell@gmail.com)
		24	*
23	* Based on work by Owen Taylor	25	* Based on work by Owen Taylor
24	*/	26	*/
25		27
		28
		29	#ifdef USE_MMX
		30
26	#include "fb.h"	31	#include "fb.h"
		32	#include "fbmmx.h"
		33
		34	#include <mmintrin.h>
27		35
28	#ifdef USE_GCC34_MMX	36	#ifdef USE_SSE
		37	#include <xmmintrin.h> /* for _mm_shuffle_pi16 and _MM_SHUFFLE */
		38	#endif
29		39
30	#ifdef RENDER	40	#ifdef RENDER
31		41
Lines 33-43 Link Here
33	#include "mipict.h"	43	#include "mipict.h"
34	#include "fbpict.h"	44	#include "fbpict.h"
35		45
36	typedef int Vector1x64 __attribute__ ((mode(DI)));
37	typedef int Vector2x32 __attribute__ ((mode(V2SI)));
38	typedef int Vector4x16 __attribute__ ((mode(V4HI)));
39	typedef int Vector8x8 __attribute__ ((mode(V8QI)));
40
41	typedef unsigned long long ullong;	46	typedef unsigned long long ullong;
42		47
43	#define noVERBOSE	48	#define noVERBOSE
Lines 50-56 Link Here
50		55
51	typedef struct	56	typedef struct
52	{	57	{
53	ullong mmx_zero;
54	ullong mmx_4x00ff;	58	ullong mmx_4x00ff;
55	ullong mmx_4x0080;	59	ullong mmx_4x0080;
56	ullong mmx_565_rgb;	60	ullong mmx_565_rgb;
Lines 70-76 Link Here
70		74
71	static const MMXData c =	75	static const MMXData c =
72	{	76	{
73	.mmx_zero = 0x0000000000000000ULL,
74	.mmx_4x00ff = 0x00ff00ff00ff00ffULL,	77	.mmx_4x00ff = 0x00ff00ff00ff00ffULL,
75	.mmx_4x0080 = 0x0080008000800080ULL,	78	.mmx_4x0080 = 0x0080008000800080ULL,
76	.mmx_565_rgb = 0x000001f0003f001fULL,	79	.mmx_565_rgb = 0x000001f0003f001fULL,
Lines 88-208 Link Here
88	.mmx_000000000000ffff = 0x000000000000ffffULL,	91	.mmx_000000000000ffff = 0x000000000000ffffULL,
89	};	92	};
90		93
91	static __inline__ Vector1x64	94	#define MC(x) ((__m64) c.mmx_##x)
92	shift (Vector1x64 v, int s)	95
		96	static __inline__ __m64
		97	shift (__m64 v, int s)
93	{	98	{
94	if (s > 0)	99	if (s > 0)
95	return __builtin_ia32_psllq (v, s);	100	return _mm_slli_si64 (v, s);
96	else if (s < 0)	101	else if (s < 0)
97	return __builtin_ia32_psrlq (v, -s);	102	return _mm_srli_si64 (v, -s);
98	else	103	else
99	return v;	104	return v;
100	}	105	}
101		106
102	static __inline__ Vector4x16	107	static __inline__ __m64
103	negate (Vector4x16 mask)	108	negate (__m64 mask)
104	{	109	{
105	return (Vector4x16)__builtin_ia32_pxor (	110	return _mm_xor_si64 (mask, MC(4x00ff));
106	(Vector1x64)mask,
107	(Vector1x64)c.mmx_4x00ff);
108	}	111	}
109		112
110	static __inline__ Vector4x16	113	static __inline__ __m64
111	pix_multiply (Vector4x16 a, Vector4x16 b)	114	pix_multiply (__m64 a, __m64 b)
112	{	115	{
113	Vector4x16 res;	116	__m64 res;
114		117
115	res = __builtin_ia32_pmullw (a, b);	118	res = _mm_mullo_pi16 (a, b);
116	res = __builtin_ia32_paddw (res, (Vector4x16)c.mmx_4x0080);	119	res = _mm_add_pi16 (res, MC(4x0080));
117	res = __builtin_ia32_psrlw (res, 8);	120	res = _mm_srli_pi16 (res, 8);
118		121
119	return res;	122	return res;
120	}	123	}
121		124
122	#if 0	125	#ifdef USE_SSE
123	#define HAVE_PSHUFW	126	#define HAVE_PSHUFW
124	#endif	127	#endif
125		128
126	#ifdef HAVE_PSHUFW	129	#ifdef HAVE_PSHUFW
127		130
128	static __inline__ Vector4x16	131	static __inline__ __m64
129	expand_alpha (Vector4x16 pixel)	132	expand_alpha (__m64 pixel)
130	{	133	{
131	Vector4x16 result;	134	return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 3, 3, 3));
132	__asm__ ("pshufw $0xFF, %1, %0\n\t" : "=y" (result) : "y" (pixel));
133	return result;
134	}	135	}
135		136
136	static __inline__ Vector4x16	137	static __inline__ __m64
137	expand_alpha_rev (Vector4x16 pixel)	138	expand_alpha_rev (__m64 pixel)
138	{	139	{
139	Vector4x16 result;	140	return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(0, 0, 0, 0));
140	__asm__ ("pshufw $0x00, %1, %0\n\t" : "=y" (result) : "y" (pixel));
141	return result;
142	}	141	}
143		142
144	static __inline__ Vector4x16	143	static __inline__ __m64
145	invert_colors (Vector4x16 pixel)	144	invert_colors (__m64 pixel)
146	{	145	{
147	Vector4x16 result;	146	return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 0, 1, 2));
148
149	/* 0xC6 = 11000110 */
150	/* 3 0 1 2 */
151
152	__asm__ ("pshufw $0xC6, %1, %0\n\t" : "=y" (result) : "y" (pixel));
153
154	return result;
155	}	147	}
156		148
157	#else	149	#else
158		150
159	static __inline__ Vector4x16	151	static __inline__ __m64
160	expand_alpha (Vector4x16 pixel)	152	expand_alpha (__m64 pixel)
161	{	153	{
162	Vector1x64 t1, t2;	154	__m64 t1, t2;
163		155
164	t1 = shift ((Vector1x64)pixel, -48);	156	t1 = shift (pixel, -48);
165	t2 = shift (t1, 16);	157	t2 = shift (t1, 16);
166	t1 = __builtin_ia32_por (t1, t2);	158	t1 = _mm_or_si64 (t1, t2);
167	t2 = shift (t1, 32);	159	t2 = shift (t1, 32);
168	t1 = __builtin_ia32_por (t1, t2);	160	t1 = _mm_or_si64 (t1, t2);
169		161
170	return (Vector4x16)t1;	162	return t1;
171	}	163	}
172		164
173	static __inline__ Vector4x16	165	static __inline__ __m64
174	expand_alpha_rev (Vector4x16 pixel)	166	expand_alpha_rev (__m64 pixel)
175	{	167	{
176	Vector1x64 t1, t2;	168	__m64 t1, t2;
177		169
178	t1 = shift ((Vector1x64)pixel, 48);	170	/* move alpha to low 16 bits and zero the rest */
		171	t1 = shift (pixel, 48);
179	t1 = shift (t1, -48);	172	t1 = shift (t1, -48);
		173
180	t2 = shift (t1, 16);	174	t2 = shift (t1, 16);
181	t1 = __builtin_ia32_por (t1, t2);	175	t1 = _mm_or_si64 (t1, t2);
182	t2 = shift (t1, 32);	176	t2 = shift (t1, 32);
183	t1 = __builtin_ia32_por (t1, t2);	177	t1 = _mm_or_si64 (t1, t2);
184		178
185	return (Vector4x16)t1;	179	return t1;
186	}	180	}
187		181
188	static __inline__ Vector4x16	182	static __inline__ __m64
189	invert_colors (Vector4x16 pixel)	183	invert_colors (__m64 pixel)
190	{	184	{
191	Vector1x64 x, y, z;	185	__m64 x, y, z;
192		186
193	x = y = z = (Vector1x64)pixel;	187	x = y = z = pixel;
194		188
195	x = __builtin_ia32_pand (x, (Vector1x64)c.mmx_ffff0000ffff0000);	189	x = _mm_and_si64 (x, MC(ffff0000ffff0000));
196	y = __builtin_ia32_pand (y, (Vector1x64)c.mmx_000000000000ffff);	190	y = _mm_and_si64 (y, MC(000000000000ffff));
197	z = __builtin_ia32_pand (z, (Vector1x64)c.mmx_0000ffff00000000);	191	z = _mm_and_si64 (z, MC(0000ffff00000000));
198		192
199	y = shift (y, 32);	193	y = shift (y, 32);
200	z = shift (z, -32);	194	z = shift (z, -32);
201		195
202	x = __builtin_ia32_por (x, y);	196	x = _mm_or_si64 (x, y);
203	x = __builtin_ia32_por (x, z);	197	x = _mm_or_si64 (x, z);
204		198
205	return (Vector4x16)x;	199	return x;
206	}	200	}
207		201
208	#endif	202	#endif
Lines 210-356 Link Here
210	/* Notes about writing mmx code	204	/* Notes about writing mmx code
211	*	205	*
212	* give memory operands as the second operand. If you give it as the	206	* give memory operands as the second operand. If you give it as the
213	* first, gcc will first load it into a register, then use that register	207	* first, gcc will first load it into a register, then use that
		208	* register
214	*	209	*
215	* ie. use	210	* ie. use
216	*	211	*
217	* __builtin_pmullw (x, mmx_constant[8]);	212	* _mm_mullo_pi16 (x, mmx_constant);
218	*	213	*
219	* not	214	* not
220	*	215	*
221	* __builtin_pmullw (mmx_constant[8], x);	216	* _mm_mullo_pi16 (mmx_constant, x);
222	*	217	*
223	* Also try to minimize dependencies. Ie. when you need a value, try to calculate	218	* Also try to minimize dependencies. i.e. when you need a value, try
224	* it from a value that was calculated as early as possible.	219	* to calculate it from a value that was calculated as early as
		220	* possible.
225	*/	221	*/
226		222
227	static __inline__ Vector4x16	223	static __inline__ __m64
228	over (Vector4x16 src, Vector4x16 srca, Vector4x16 dest)	224	over (__m64 src, __m64 srca, __m64 dest)
229	{	225	{
230	return (Vector4x16)__builtin_ia32_paddusb ((Vector8x8)src, (Vector8x8)pix_multiply(dest, negate(srca)));	226	return _mm_adds_pu8 (src, pix_multiply(dest, negate(srca)));
231	}	227	}
232		228
233	static __inline__ Vector4x16	229	static __inline__ __m64
234	over_rev_non_pre (Vector4x16 src, Vector4x16 dest)	230	over_rev_non_pre (__m64 src, __m64 dest)
235	{	231	{
236	Vector4x16 srca = expand_alpha (src);	232	__m64 srca = expand_alpha (src);
237	Vector4x16 srcfaaa = (Vector4x16)__builtin_ia32_por((Vector1x64)srca, (Vector1x64)c.mmx_full_alpha);	233	__m64 srcfaaa = _mm_or_si64 (srca, MC(full_alpha));
238		234
239	return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest);	235	return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest);
240	}	236	}
241		237
242	static __inline__ Vector4x16	238	static __inline__ __m64
243	in (Vector4x16 src,	239	in (__m64 src,
244	Vector4x16 mask)	240	__m64 mask)
245	{	241	{
246	return pix_multiply (src, mask);	242	return pix_multiply (src, mask);
247	}	243	}
248		244
249	static __inline__ Vector4x16	245	static __inline__ __m64
250	in_over (Vector4x16 src,	246	in_over (__m64 src,
251	Vector4x16 srca,	247	__m64 srca,
252	Vector4x16 mask,	248	__m64 mask,
253	Vector4x16 dest)	249	__m64 dest)
254	{	250	{
255	return over(in(src, mask), pix_multiply(srca, mask), dest);	251	return over(in(src, mask), pix_multiply(srca, mask), dest);
256	}	252	}
257		253
258	static __inline__ Vector8x8	254	static __inline__ __m64
259	cvt32to64 (CARD32 v)
260	{
261	ullong r = v;
262	return (Vector8x8)r;
263	}
264
265	static __inline__ Vector4x16
266	load8888 (CARD32 v)	255	load8888 (CARD32 v)
267	{	256	{
268	return (Vector4x16)__builtin_ia32_punpcklbw (cvt32to64 (v),	257	return _mm_unpacklo_pi8 (_mm_cvtsi32_si64 (v), _mm_setzero_si64());
269	(Vector8x8)c.mmx_zero);
270	}	258	}
271		259
272	static __inline__ Vector8x8	260	static __inline__ __m64
273	pack8888 (Vector4x16 lo, Vector4x16 hi)	261	pack8888 (__m64 lo, __m64 hi)
274	{	262	{
275	Vector8x8 r;	263	__m64 r;
276	r = __builtin_ia32_packuswb ((Vector4x16)lo, (Vector4x16)hi);	264	r = _mm_packs_pu16 (lo, hi);
277	return r;	265	return r;
278	}	266	}
279		267
280	/* Expand 16 bits positioned at @pos (0-3) of a mmx register into 00RR00GG00BB	268	/* Expand 16 bits positioned at @pos (0-3) of a mmx register into
281		269	*
282	--- Expanding 565 in the low word ---	270	* 00RR00GG00BB
283		271	*
284	m = (m << (32 - 3)) \| (m << (16 - 5)) \| m;	272	* --- Expanding 565 in the low word ---
285	m = m & (01f0003f001f);	273	*
286	m = m * (008404100840);	274	* m = (m << (32 - 3)) \| (m << (16 - 5)) \| m;
287	m = m >> 8;	275	* m = m & (01f0003f001f);
288		276	* m = m * (008404100840);
289	Note the trick here - the top word is shifted by another nibble to avoid	277	* m = m >> 8;
290	it bumping into the middle word	278	*
291	*/	279	* Note the trick here - the top word is shifted by another nibble to
292	static __inline__ Vector4x16	280	* avoid it bumping into the middle word
293	expand565 (Vector4x16 pixel, int pos)	281	*/
		282	static __inline__ __m64
		283	expand565 (__m64 pixel, int pos)
294	{	284	{
295	Vector1x64 p = (Vector1x64)pixel;	285	__m64 p = pixel;
		286	__m64 t1, t2;
296		287
297	/* move pixel to low 16 bit and zero the rest */	288	/* move pixel to low 16 bit and zero the rest */
298	p = shift (shift (p, (3 - pos) * 16), -48);	289	p = shift (shift (p, (3 - pos) * 16), -48);
299		290
300	Vector1x64 t1 = shift (p, 36 - 11);	291	t1 = shift (p, 36 - 11);
301	Vector1x64 t2 = shift (p, 16 - 5);	292	t2 = shift (p, 16 - 5);
302		293
303	p = __builtin_ia32_por (t1, p);	294	p = _mm_or_si64 (t1, p);
304	p = __builtin_ia32_por (t2, p);	295	p = _mm_or_si64 (t2, p);
305	p = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_rgb);	296	p = _mm_and_si64 (p, MC(565_rgb));
306		297
307	pixel = __builtin_ia32_pmullw ((Vector4x16)p, (Vector4x16)c.mmx_565_unpack_multiplier);	298	pixel = _mm_mullo_pi16 (p, MC(565_unpack_multiplier));
308	return __builtin_ia32_psrlw (pixel, 8);	299	return _mm_srli_pi16 (pixel, 8);
309	}	300	}
310		301
311	static __inline__ Vector4x16	302	static __inline__ __m64
312	expand8888 (Vector4x16 in, int pos)	303	expand8888 (__m64 in, int pos)
313	{	304	{
314	if (pos == 0)	305	if (pos == 0)
315	return (Vector4x16)__builtin_ia32_punpcklbw ((Vector8x8)in, (Vector8x8)c.mmx_zero);	306	return _mm_unpacklo_pi8 (in, _mm_setzero_si64());
316	else	307	else
317	return (Vector4x16)__builtin_ia32_punpckhbw ((Vector8x8)in, (Vector8x8)c.mmx_zero);	308	return _mm_unpackhi_pi8 (in, _mm_setzero_si64());
318	}	309	}
319		310
320	static __inline__ Vector4x16	311	static __inline__ __m64
321	pack565 (Vector4x16 pixel, Vector4x16 target, int pos)	312	pack565 (__m64 pixel, __m64 target, int pos)
322	{	313	{
323	Vector1x64 p = (Vector1x64)pixel;	314	__m64 p = pixel;
324	Vector1x64 t = (Vector1x64)target;	315	__m64 t = target;
325	Vector1x64 r, g, b;	316	__m64 r, g, b;
326		317
327	r = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_r);	318	r = _mm_and_si64 (p, MC(565_r));
328	g = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_g);	319	g = _mm_and_si64 (p, MC(565_g));
329	b = __builtin_ia32_pand (p, (Vector1x64)c.mmx_565_b);	320	b = _mm_and_si64 (p, MC(565_b));
330		321
331	r = shift (r, - (32 - 8) + pos * 16);	322	r = shift (r, - (32 - 8) + pos * 16);
332	g = shift (g, - (16 - 3) + pos * 16);	323	g = shift (g, - (16 - 3) + pos * 16);
333	b = shift (b, - (0 + 3) + pos * 16);	324	b = shift (b, - (0 + 3) + pos * 16);
334		325
335	if (pos == 0)	326	if (pos == 0)
336	t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_0);	327	t = _mm_and_si64 (t, MC(mask_0));
337	else if (pos == 1)	328	else if (pos == 1)
338	t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_1);	329	t = _mm_and_si64 (t, MC(mask_1));
339	else if (pos == 2)	330	else if (pos == 2)
340	t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_2);	331	t = _mm_and_si64 (t, MC(mask_2));
341	else if (pos == 3)	332	else if (pos == 3)
342	t = __builtin_ia32_pand (t, (Vector1x64)c.mmx_mask_3);	333	t = _mm_and_si64 (t, MC(mask_3));
343		334
344	p = __builtin_ia32_por (r, t);	335	p = _mm_or_si64 (r, t);
345	p = __builtin_ia32_por (g, p);	336	p = _mm_or_si64 (g, p);
346		337
347	return (Vector4x16)__builtin_ia32_por (b, p);	338	return _mm_or_si64 (b, p);
348	}
349
350	static __inline__ void
351	emms (void)
352	{
353	__asm__ __volatile__ ("emms");
354	}	339	}
355		340
356	void	341	void
Lines 371-378 Link Here
371	CARD32 dstLine, dst;	356	CARD32 dstLine, dst;
372	CARD16 w;	357	CARD16 w;
373	FbStride dstStride;	358	FbStride dstStride;
374	Vector4x16 vsrc, vsrca;	359	__m64 vsrc, vsrca;
375		360
376	CHECKPOINT();	361	CHECKPOINT();
377		362
378	fbComposeGetSolid(pSrc, src, pDst->format);	363	fbComposeGetSolid(pSrc, src, pDst->format);
Lines 384-434 Link Here
384		369
385	vsrc = load8888 (src);	370	vsrc = load8888 (src);
386	vsrca = expand_alpha (vsrc);	371	vsrca = expand_alpha (vsrc);
387		372
388	while (height--)	373	while (height--)
389	{	374	{
390	dst = dstLine;	375	dst = dstLine;
391	dstLine += dstStride;	376	dstLine += dstStride;
392	w = width;	377	w = width;
393		378
394	CHECKPOINT();	379	CHECKPOINT();
395		380
396	while (w && (unsigned long)dst & 7)	381	while (w && (unsigned long)dst & 7)
397	{	382	{
398	dst = (ullong) pack8888(over(vsrc, vsrca, load8888(dst)), (Vector4x16)c.mmx_zero);	383	dst = (ullong) pack8888(over(vsrc, vsrca, load8888(dst)),
		384	_mm_setzero_si64());
399		385
400	w--;	386	w--;
401	dst++;	387	dst++;
402	}	388	}
403		389
404	while (w >= 2)	390	while (w >= 2)
405	{	391	{
406	Vector4x16 vdest;	392	__m64 vdest;
407	Vector4x16 dest0, dest1;	393	__m64 dest0, dest1;
408		394
409	vdest = (Vector4x16 )dst;	395	vdest = (__m64 )dst;
410		396
411	dest0 = over(vsrc, vsrca, expand8888(vdest, 0));	397	dest0 = over(vsrc, vsrca, expand8888(vdest, 0));
412	dest1 = over(vsrc, vsrca, expand8888(vdest, 1));	398	dest1 = over(vsrc, vsrca, expand8888(vdest, 1));
413		399
414	(Vector8x8 )dst = (Vector8x8)pack8888(dest0, dest1);	400	(__m64 )dst = pack8888(dest0, dest1);
415		401
416	dst += 2;	402	dst += 2;
417	w -= 2;	403	w -= 2;
418	}	404	}
419		405
420	CHECKPOINT();	406	CHECKPOINT();
421		407
422	while (w)	408	while (w)
423	{	409	{
424	dst = (ullong) pack8888(over(vsrc, vsrca, load8888(dst)), (Vector4x16)c.mmx_zero);	410	dst = (ullong) pack8888(over(vsrc, vsrca, load8888(dst)), _mm_setzero_si64());
425		411
426	w--;	412	w--;
427	dst++;	413	dst++;
428	}	414	}
429	}	415	}
430		416
431	emms();	417	_mm_empty();
432	}	418	}
433		419
434	void	420	void
Lines 449-456 Link Here
449	CARD16 dstLine, dst;	435	CARD16 dstLine, dst;
450	CARD16 w;	436	CARD16 w;
451	FbStride dstStride;	437	FbStride dstStride;
452	Vector4x16 vsrc, vsrca;	438	__m64 vsrc, vsrca;
453		439
454	CHECKPOINT();	440	CHECKPOINT();
455		441
456	fbComposeGetSolid(pSrc, src, pDst->format);	442	fbComposeGetSolid(pSrc, src, pDst->format);
Lines 462-510 Link Here
462		448
463	vsrc = load8888 (src);	449	vsrc = load8888 (src);
464	vsrca = expand_alpha (vsrc);	450	vsrca = expand_alpha (vsrc);
465		451
466	while (height--)	452	while (height--)
467	{	453	{
468	dst = dstLine;	454	dst = dstLine;
469	dstLine += dstStride;	455	dstLine += dstStride;
470	w = width;	456	w = width;
471		457
472	CHECKPOINT();	458	CHECKPOINT();
473		459
474	while (w && (unsigned long)dst & 7)	460	while (w && (unsigned long)dst & 7)
475	{	461	{
476	ullong d = *dst;	462	ullong d = *dst;
477	Vector4x16 vdest = expand565 ((Vector4x16)d, 0);	463	__m64 vdest = expand565 ((__m64)d, 0);
478	vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);	464	vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);
479	*dst = (ullong)vdest;	465	*dst = (ullong)vdest;
480		466
481	w--;	467	w--;
482	dst++;	468	dst++;
483	}	469	}
484		470
485	while (w >= 4)	471	while (w >= 4)
486	{	472	{
487	Vector4x16 vdest;	473	__m64 vdest;
488		474
489	vdest = (Vector4x16 )dst;	475	vdest = (__m64 )dst;
490		476
491	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 0)), vdest, 0);	477	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 0)), vdest, 0);
492	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 1)), vdest, 1);	478	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 1)), vdest, 1);
493	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 2)), vdest, 2);	479	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 2)), vdest, 2);
494	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 3)), vdest, 3);	480	vdest = pack565 (over(vsrc, vsrca, expand565(vdest, 3)), vdest, 3);
495		481
496	(Vector8x8 )dst = (Vector8x8)vdest;	482	(__m64 )dst = vdest;
497		483
498	dst += 4;	484	dst += 4;
499	w -= 4;	485	w -= 4;
500	}	486	}
501		487
502	CHECKPOINT();	488	CHECKPOINT();
503		489
504	while (w)	490	while (w)
505	{	491	{
506	ullong d = *dst;	492	ullong d = *dst;
507	Vector4x16 vdest = expand565 ((Vector4x16)d, 0);	493	__m64 vdest = expand565 ((__m64)d, 0);
508	vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);	494	vdest = pack565(over(vsrc, vsrca, vdest), vdest, 0);
509	*dst = (ullong)vdest;	495	*dst = (ullong)vdest;
510		496
Lines 513-519 Link Here
513	}	499	}
514	}	500	}
515		501
516	emms();	502	_mm_empty();
517	}	503	}
518		504
519	void	505	void
Lines 534-541 Link Here
534	CARD32 *dstLine;	520	CARD32 *dstLine;
535	CARD32 *maskLine;	521	CARD32 *maskLine;
536	FbStride dstStride, maskStride;	522	FbStride dstStride, maskStride;
537	Vector4x16 vsrc, vsrca;	523	__m64 vsrc, vsrca;
538		524
539	CHECKPOINT();	525	CHECKPOINT();
540		526
541	fbComposeGetSolid(pSrc, src, pDst->format);	527	fbComposeGetSolid(pSrc, src, pDst->format);
Lines 562-570 Link Here
562		548
563	if (m)	549	if (m)
564	{	550	{
565	Vector4x16 vdest = load8888(*q);	551	__m64 vdest = load8888(*q);
566	vdest = in_over(vsrc, vsrca, load8888(m), vdest);	552	vdest = in_over(vsrc, vsrca, load8888(m), vdest);
567	*q = (ullong)pack8888(vdest, (Vector4x16)c.mmx_zero);	553	*q = (ullong)pack8888(vdest, _mm_setzero_si64());
568	}	554	}
569		555
570	twidth--;	556	twidth--;
Lines 580-594 Link Here
580		566
581	if (m0 \| m1)	567	if (m0 \| m1)
582	{	568	{
583	Vector4x16 dest0, dest1;	569	__m64 dest0, dest1;
584	Vector4x16 vdest = (Vector4x16 )q;	570	__m64 vdest = (__m64 )q;
585		571
586	dest0 = in_over(vsrc, vsrca, load8888(m0),	572	dest0 = in_over(vsrc, vsrca, load8888(m0),
587	expand8888 (vdest, 0));	573	expand8888 (vdest, 0));
588	dest1 = in_over(vsrc, vsrca, load8888(m1),	574	dest1 = in_over(vsrc, vsrca, load8888(m1),
589	expand8888 (vdest, 1));	575	expand8888 (vdest, 1));
590		576
591	(Vector8x8 )q = (Vector8x8)pack8888(dest0, dest1);	577	(__m64 )q = pack8888(dest0, dest1);
592	}	578	}
593		579
594	p += 2;	580	p += 2;
Lines 602-610 Link Here
602		588
603	if (m)	589	if (m)
604	{	590	{
605	Vector4x16 vdest = load8888(*q);	591	__m64 vdest = load8888(*q);
606	vdest = in_over(vsrc, vsrca, load8888(m), vdest);	592	vdest = in_over(vsrc, vsrca, load8888(m), vdest);
607	*q = (ullong)pack8888(vdest, (Vector4x16)c.mmx_zero);	593	*q = (ullong)pack8888(vdest, _mm_setzero_si64());
608	}	594	}
609		595
610	twidth--;	596	twidth--;
Lines 616-622 Link Here
616	maskLine += maskStride;	602	maskLine += maskStride;
617	}	603	}
618		604
619	emms();	605	_mm_empty();
		606	}
		607
		608	void
		609	fbCompositeSrc_8888x8x8888mmx (CARD8 op,
		610	PicturePtr pSrc,
		611	PicturePtr pMask,
		612	PicturePtr pDst,
		613	INT16 xSrc,
		614	INT16 ySrc,
		615	INT16 xMask,
		616	INT16 yMask,
		617	INT16 xDst,
		618	INT16 yDst,
		619	CARD16 width,
		620	CARD16 height)
		621	{
		622	CARD32 dstLine, dst;
		623	CARD32 srcLine, src;
		624	CARD8 *maskLine;
		625	CARD32 mask;
		626	__m64 vmask;
		627	FbStride dstStride, srcStride, maskStride;
		628	CARD16 w;
		629	__m64 srca;
		630
		631	CHECKPOINT();
		632
		633	fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
		634	fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
		635	fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
		636
		637	mask = maskLine << 24 \| maskLine << 16 \| maskLine << 8 \| maskLine;
		638	vmask = load8888 (mask);
		639	srca = MC(4x00ff);
		640
		641	while (height--)
		642	{
		643	dst = dstLine;
		644	dstLine += dstStride;
		645	src = srcLine;
		646	srcLine += srcStride;
		647	w = width;
		648
		649	while (w && (unsigned long)dst & 7)
		650	{
		651	__m64 s = load8888 (*src);
		652	__m64 d = load8888 (*dst);
		653
		654	*dst = (ullong)pack8888 (in_over (s, srca, vmask, d), (__m64)_mm_setzero_si64());
		655
		656	w--;
		657	dst++;
		658	src++;
		659	}
		660
		661	while (w >= 16)
		662	{
		663	__m64 vd0 = (__m64 )(dst + 0);
		664	__m64 vd1 = (__m64 )(dst + 2);
		665	__m64 vd2 = (__m64 )(dst + 4);
		666	__m64 vd3 = (__m64 )(dst + 6);
		667	__m64 vd4 = (__m64 )(dst + 8);
		668	__m64 vd5 = (__m64 )(dst + 10);
		669	__m64 vd6 = (__m64 )(dst + 12);
670	__m64 vd7 = (__m64 )(dst + 14);
671
672	__m64 vs0 = (__m64 )(src + 0);
673	__m64 vs1 = (__m64 )(src + 2);
674	__m64 vs2 = (__m64 )(src + 4);
675	__m64 vs3 = (__m64 )(src + 6);
676	__m64 vs4 = (__m64 )(src + 8);
677	__m64 vs5 = (__m64 )(src + 10);
678	__m64 vs6 = (__m64 )(src + 12);
679	__m64 vs7 = (__m64 )(dst + 14);
680
681	vd0 = (__m64)pack8888 (
682	in_over (expand8888 (vs0, 0), srca, vmask, expand8888 (vd0, 0)),
683	in_over (expand8888 (vs0, 1), srca, vmask, expand8888 (vd0, 1)));
684
685	vd1 = (__m64)pack8888 (
686	in_over (expand8888 (vs1, 0), srca, vmask, expand8888 (vd1, 0)),
687	in_over (expand8888 (vs1, 1), srca, vmask, expand8888 (vd1, 1)));
688
689	vd2 = (__m64)pack8888 (
690	in_over (expand8888 (vs2, 0), srca, vmask, expand8888 (vd2, 0)),
691	in_over (expand8888 (vs2, 1), srca, vmask, expand8888 (vd2, 1)));
692
693	vd3 = (__m64)pack8888 (
694	in_over (expand8888 (vs3, 0), srca, vmask, expand8888 (vd3, 0)),
695	in_over (expand8888 (vs3, 1), srca, vmask, expand8888 (vd3, 1)));
696
697	vd4 = (__m64)pack8888 (
698	in_over (expand8888 (vs4, 0), srca, vmask, expand8888 (vd4, 0)),
699	in_over (expand8888 (vs4, 1), srca, vmask, expand8888 (vd4, 1)));
700
701	vd5 = (__m64)pack8888 (
702	in_over (expand8888 (vs5, 0), srca, vmask, expand8888 (vd5, 0)),
703	in_over (expand8888 (vs5, 1), srca, vmask, expand8888 (vd5, 1)));
704
705	vd6 = (__m64)pack8888 (
706	in_over (expand8888 (vs6, 0), srca, vmask, expand8888 (vd6, 0)),
707	in_over (expand8888 (vs6, 1), srca, vmask, expand8888 (vd6, 1)));
708
709	vd7 = (__m64)pack8888 (
710	in_over (expand8888 (vs7, 0), srca, vmask, expand8888 (vd7, 0)),
711	in_over (expand8888 (vs7, 1), srca, vmask, expand8888 (vd7, 1)));
712
713	w -= 16;
714	dst += 16;
715	src += 16;
716	}
717
718	while (w)
719	{
720	__m64 s = load8888 (*src);
721	__m64 d = load8888 (*dst);
722
723	*dst = (ullong)pack8888 (in_over (s, srca, vmask, d), (__m64)_mm_setzero_si64());
724
725	w--;
726	dst++;
727	src++;
728	}
729	}
730
731	_mm_empty();
620	}	732	}
621		733
622	void	734	void
Lines 638-644 Link Here
638	CARD8 maskLine, mask;	750	CARD8 maskLine, mask;
639	FbStride dstStride, maskStride;	751	FbStride dstStride, maskStride;
640	CARD16 w;	752	CARD16 w;
641	Vector4x16 vsrc, vsrca;	753	__m64 vsrc, vsrca;
642	ullong srcsrc;	754	ullong srcsrc;
643		755
644	CHECKPOINT();	756	CHECKPOINT();
Lines 648-654 Link Here
648	srca = src >> 24;	760	srca = src >> 24;
649	if (srca == 0)	761	if (srca == 0)
650	return;	762	return;
651		763
652	srcsrc = (unsigned long long)src << 32 \| src;	764	srcsrc = (unsigned long long)src << 32 \| src;
653		765
654	fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);	766	fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
Lines 664-670 Link Here
664	mask = maskLine;	776	mask = maskLine;
665	maskLine += maskStride;	777	maskLine += maskStride;
666	w = width;	778	w = width;
667		779
668	CHECKPOINT();	780	CHECKPOINT();
669		781
670	while (w && (unsigned long)dst & 7)	782	while (w && (unsigned long)dst & 7)
Lines 673-687 Link Here
673		785
674	if (m)	786	if (m)
675	{	787	{
676	Vector4x16 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m), load8888(*dst));	788	__m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), load8888(*dst));
677	*dst = (ullong)pack8888(vdest, (Vector4x16)c.mmx_zero);	789	*dst = (ullong)pack8888(vdest, _mm_setzero_si64());
678	}	790	}
679		791
680	w--;	792	w--;
681	mask++;	793	mask++;
682	dst++;	794	dst++;
683	}	795	}
684		796
685	CHECKPOINT();	797	CHECKPOINT();
686		798
687	while (w >= 2)	799	while (w >= 2)
Lines 689-717 Link Here
689	ullong m0, m1;	801	ullong m0, m1;
690	m0 = *mask;	802	m0 = *mask;
691	m1 = *(mask + 1);	803	m1 = *(mask + 1);
692		804
693	if (srca == 0xff && (m0 & m1) == 0xff)	805	if (srca == 0xff && (m0 & m1) == 0xff)
694	{	806	{
695	(unsigned long long )dst = srcsrc;	807	(unsigned long long )dst = srcsrc;
696	}	808	}
697	else if (m0 \| m1)	809	else if (m0 \| m1)
698	{	810	{
699	Vector4x16 vdest;	811	__m64 vdest;
700	Vector4x16 dest0, dest1;	812	__m64 dest0, dest1;
701		813
702	vdest = (Vector4x16 )dst;	814	vdest = (__m64 )dst;
703		815
704	dest0 = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m0), expand8888(vdest, 0));	816	dest0 = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m0), expand8888(vdest, 0));
705	dest1 = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m1), expand8888(vdest, 1));	817	dest1 = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m1), expand8888(vdest, 1));
706		818
707	(Vector8x8 )dst = (Vector8x8)pack8888(dest0, dest1);	819	(__m64 )dst = pack8888(dest0, dest1);
708	}	820	}
709		821
710	mask += 2;	822	mask += 2;
711	dst += 2;	823	dst += 2;
712	w -= 2;	824	w -= 2;
713	}	825	}
714		826
715	CHECKPOINT();	827	CHECKPOINT();
716		828
717	while (w)	829	while (w)
Lines 720-728 Link Here
720		832
721	if (m)	833	if (m)
722	{	834	{
723	Vector4x16 vdest = load8888(*dst);	835	__m64 vdest = load8888(*dst);
724	vdest = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m), vdest);	836	vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), vdest);
725	*dst = (ullong)pack8888(vdest, (Vector4x16)c.mmx_zero);	837	*dst = (ullong)pack8888(vdest, _mm_setzero_si64());
726	}	838	}
727		839
728	w--;	840	w--;
Lines 731-737 Link Here
731	}	843	}
732	}	844	}
733		845
734	emms();	846	_mm_empty();
735	}	847	}
736		848
737		849
Lines 754-760 Link Here
754	CARD8 maskLine, mask;	866	CARD8 maskLine, mask;
755	FbStride dstStride, maskStride;	867	FbStride dstStride, maskStride;
756	CARD16 w;	868	CARD16 w;
757	Vector4x16 vsrc, vsrca;	869	__m64 vsrc, vsrca;
758	unsigned long long srcsrcsrcsrc, src16;	870	unsigned long long srcsrcsrcsrc, src16;
759		871
760	CHECKPOINT();	872	CHECKPOINT();
Lines 770-778 Link Here
770		882
771	vsrc = load8888 (src);	883	vsrc = load8888 (src);
772	vsrca = expand_alpha (vsrc);	884	vsrca = expand_alpha (vsrc);
773		885
774	src16 = (ullong)pack565(vsrc, (Vector4x16)c.mmx_zero, 0);	886	src16 = (ullong)pack565(vsrc, _mm_setzero_si64(), 0);
775		887
776	srcsrcsrcsrc = (ullong)src16 << 48 \| (ullong)src16 << 32 \|	888	srcsrcsrcsrc = (ullong)src16 << 48 \| (ullong)src16 << 32 \|
777	(ullong)src16 << 16 \| (ullong)src16;	889	(ullong)src16 << 16 \| (ullong)src16;
778		890
Lines 783-789 Link Here
783	mask = maskLine;	895	mask = maskLine;
784	maskLine += maskStride;	896	maskLine += maskStride;
785	w = width;	897	w = width;
786		898
787	CHECKPOINT();	899	CHECKPOINT();
788		900
789	while (w && (unsigned long)dst & 7)	901	while (w && (unsigned long)dst & 7)
Lines 793-808 Link Here
793	if (m)	905	if (m)
794	{	906	{
795	ullong d = *dst;	907	ullong d = *dst;
796	Vector4x16 vd = (Vector4x16)d;	908	__m64 vd = (__m64)d;
797	Vector4x16 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m), expand565(vd, 0));	909	__m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), expand565(vd, 0));
798	*dst = (ullong)pack565(vdest, (Vector4x16)c.mmx_zero, 0);	910	*dst = (ullong)pack565(vdest, _mm_setzero_si64(), 0);
799	}	911	}
800		912
801	w--;	913	w--;
802	mask++;	914	mask++;
803	dst++;	915	dst++;
804	}	916	}
805		917
806	CHECKPOINT();	918	CHECKPOINT();
807		919
808	while (w >= 4)	920	while (w >= 4)
Lines 812-846 Link Here
812	m1 = *(mask + 1);	924	m1 = *(mask + 1);
813	m2 = *(mask + 2);	925	m2 = *(mask + 2);
814	m3 = *(mask + 3);	926	m3 = *(mask + 3);
815		927
816	if (srca == 0xff && (m0 & m1 & m2 & m3) == 0xff)	928	if (srca == 0xff && (m0 & m1 & m2 & m3) == 0xff)
817	{	929	{
818	(unsigned long long )dst = srcsrcsrcsrc;	930	(unsigned long long )dst = srcsrcsrcsrc;
819	}	931	}
820	else if (m0 \| m1 \| m2 \| m3)	932	else if (m0 \| m1 \| m2 \| m3)
821	{	933	{
822	Vector4x16 vdest;	934	__m64 vdest;
823	Vector4x16 vm0, vm1, vm2, vm3;	935	__m64 vm0, vm1, vm2, vm3;
824		936
825	vdest = (Vector4x16 )dst;	937	vdest = (__m64 )dst;
826		938
827	vm0 = (Vector4x16)m0;	939	vm0 = (__m64)m0;
828	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm0), expand565(vdest, 0)), vdest, 0);	940	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm0), expand565(vdest, 0)), vdest, 0);
829	vm1 = (Vector4x16)m1;	941	vm1 = (__m64)m1;
830	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm1), expand565(vdest, 1)), vdest, 1);	942	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm1), expand565(vdest, 1)), vdest, 1);
831	vm2 = (Vector4x16)m2;	943	vm2 = (__m64)m2;
832	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm2), expand565(vdest, 2)), vdest, 2);	944	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm2), expand565(vdest, 2)), vdest, 2);
833	vm3 = (Vector4x16)m3;	945	vm3 = (__m64)m3;
834	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm3), expand565(vdest, 3)), vdest, 3);	946	vdest = pack565(in_over(vsrc, vsrca, expand_alpha_rev(vm3), expand565(vdest, 3)), vdest, 3);
835		947
836	(Vector4x16 )dst = vdest;	948	(__m64 )dst = vdest;
837	}	949	}
838		950
839	w -= 4;	951	w -= 4;
840	mask += 4;	952	mask += 4;
841	dst += 4;	953	dst += 4;
842	}	954	}
843		955
844	CHECKPOINT();	956	CHECKPOINT();
845		957
846	while (w)	958	while (w)
Lines 850-858 Link Here
850	if (m)	962	if (m)
851	{	963	{
852	ullong d = *dst;	964	ullong d = *dst;
853	Vector4x16 vd = (Vector4x16)d;	965	__m64 vd = (__m64)d;
854	Vector4x16 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((Vector4x16)m), expand565(vd, 0));	966	__m64 vdest = in_over(vsrc, vsrca, expand_alpha_rev ((__m64)m), expand565(vd, 0));
855	*dst = (ullong)pack565(vdest, (Vector4x16)c.mmx_zero, 0);	967	*dst = (ullong)pack565(vdest, _mm_setzero_si64(), 0);
856	}	968	}
857		969
858	w--;	970	w--;
Lines 861-867 Link Here
861	}	973	}
862	}	974	}
863		975
864	emms();	976	_mm_empty();
865	}	977	}
866		978
867	void	979	void
Lines 887-895 Link Here
887		999
888	fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);	1000	fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
889	fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);	1001	fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
890		1002
891	assert (pSrc->pDrawable == pMask->pDrawable);	1003	assert (pSrc->pDrawable == pMask->pDrawable);
892		1004
893	while (height--)	1005	while (height--)
894	{	1006	{
895	dst = dstLine;	1007	dst = dstLine;
Lines 897-910 Link Here
897	src = srcLine;	1009	src = srcLine;
898	srcLine += srcStride;	1010	srcLine += srcStride;
899	w = width;	1011	w = width;
900		1012
901	CHECKPOINT();	1013	CHECKPOINT();
902		1014
903	while (w && (unsigned long)dst & 7)	1015	while (w && (unsigned long)dst & 7)
904	{	1016	{
905	Vector4x16 vsrc = load8888 (*src);	1017	__m64 vsrc = load8888 (*src);
906	ullong d = *dst;	1018	ullong d = *dst;
907	Vector4x16 vdest = expand565 ((Vector4x16)d, 0);	1019	__m64 vdest = expand565 ((__m64)d, 0);
908		1020
909	vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);	1021	vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);
910		1022
Lines 914-932 Link Here
914	dst++;	1026	dst++;
915	src++;	1027	src++;
916	}	1028	}
917		1029
918	CHECKPOINT();	1030	CHECKPOINT();
919		1031
920	while (w >= 4)	1032	while (w >= 4)
921	{	1033	{
922	CARD32 s0, s1, s2, s3;	1034	CARD32 s0, s1, s2, s3;
923	unsigned char a0, a1, a2, a3;	1035	unsigned char a0, a1, a2, a3;
924		1036
925	s0 = *src;	1037	s0 = *src;
926	s1 = *(src + 1);	1038	s1 = *(src + 1);
927	s2 = *(src + 2);	1039	s2 = *(src + 2);
928	s3 = *(src + 3);	1040	s3 = *(src + 3);
929		1041
930	a0 = (s0 >> 24);	1042	a0 = (s0 >> 24);
931	a1 = (s1 >> 24);	1043	a1 = (s1 >> 24);
932	a2 = (s2 >> 24);	1044	a2 = (s2 >> 24);
Lines 934-971 Link Here
934		1046
935	if ((a0 & a1 & a2 & a3) == 0xFF)	1047	if ((a0 & a1 & a2 & a3) == 0xFF)
936	{	1048	{
937	Vector4x16 vdest;	1049	__m64 vdest;
938	vdest = pack565(invert_colors(load8888(s0)), (Vector4x16)c.mmx_zero, 0);	1050	vdest = pack565(invert_colors(load8888(s0)), _mm_setzero_si64(), 0);
939	vdest = pack565(invert_colors(load8888(s1)), vdest, 1);	1051	vdest = pack565(invert_colors(load8888(s1)), vdest, 1);
940	vdest = pack565(invert_colors(load8888(s2)), vdest, 2);	1052	vdest = pack565(invert_colors(load8888(s2)), vdest, 2);
941	vdest = pack565(invert_colors(load8888(s3)), vdest, 3);	1053	vdest = pack565(invert_colors(load8888(s3)), vdest, 3);
942		1054
943	(Vector4x16 )dst = vdest;	1055	(__m64 )dst = vdest;
944	}	1056	}
945	else if (a0 \| a1 \| a2 \| a3)	1057	else if (a0 \| a1 \| a2 \| a3)
946	{	1058	{
947	Vector4x16 vdest = (Vector4x16 )dst;	1059	__m64 vdest = (__m64 )dst;
948		1060
949	vdest = pack565(over_rev_non_pre(load8888(s0), expand565(vdest, 0)), vdest, 0);	1061	vdest = pack565(over_rev_non_pre(load8888(s0), expand565(vdest, 0)), vdest, 0);
950	vdest = pack565(over_rev_non_pre(load8888(s1), expand565(vdest, 1)), vdest, 1);	1062	vdest = pack565(over_rev_non_pre(load8888(s1), expand565(vdest, 1)), vdest, 1);
951	vdest = pack565(over_rev_non_pre(load8888(s2), expand565(vdest, 2)), vdest, 2);	1063	vdest = pack565(over_rev_non_pre(load8888(s2), expand565(vdest, 2)), vdest, 2);
952	vdest = pack565(over_rev_non_pre(load8888(s3), expand565(vdest, 3)), vdest, 3);	1064	vdest = pack565(over_rev_non_pre(load8888(s3), expand565(vdest, 3)), vdest, 3);
953		1065
954	(Vector4x16 )dst = vdest;	1066	(__m64 )dst = vdest;
955	}	1067	}
956		1068
957	w -= 4;	1069	w -= 4;
958	dst += 4;	1070	dst += 4;
959	src += 4;	1071	src += 4;
960	}	1072	}
961		1073
962	CHECKPOINT();	1074	CHECKPOINT();
963		1075
964	while (w)	1076	while (w)
965	{	1077	{
966	Vector4x16 vsrc = load8888 (*src);	1078	__m64 vsrc = load8888 (*src);
967	ullong d = *dst;	1079	ullong d = *dst;
968	Vector4x16 vdest = expand565 ((Vector4x16)d, 0);	1080	__m64 vdest = expand565 ((__m64)d, 0);
969		1081
970	vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);	1082	vdest = pack565(over_rev_non_pre(vsrc, vdest), vdest, 0);
971		1083
Lines 976-986 Link Here
976	src++;	1088	src++;
977	}	1089	}
978	}	1090	}
979		1091
980	emms();	1092	_mm_empty();
981	}	1093	}
982		1094
983	/* "888RevNP" is GdkPixbuf's format: ABGR, non premultiplied */	1095	/* "8888RevNP" is GdkPixbuf's format: ABGR, non premultiplied */
984		1096
985	void	1097	void
986	fbCompositeSrc_8888RevNPx8888mmx (CARD8 op,	1098	fbCompositeSrc_8888RevNPx8888mmx (CARD8 op,
Lines 1005-1013 Link Here
1005		1117
1006	fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);	1118	fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
1007	fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);	1119	fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
1008		1120
1009	assert (pSrc->pDrawable == pMask->pDrawable);	1121	assert (pSrc->pDrawable == pMask->pDrawable);
1010		1122
1011	while (height--)	1123	while (height--)
1012	{	1124	{
1013	dst = dstLine;	1125	dst = dstLine;
Lines 1015-1042 Link Here
1015	src = srcLine;	1127	src = srcLine;
1016	srcLine += srcStride;	1128	srcLine += srcStride;
1017	w = width;	1129	w = width;
1018		1130
1019	while (w && (unsigned long)dst & 7)	1131	while (w && (unsigned long)dst & 7)
1020	{	1132	{
1021	Vector4x16 s = load8888 (*src);	1133	__m64 s = load8888 (*src);
1022	Vector4x16 d = load8888 (*dst);	1134	__m64 d = load8888 (*dst);
1023		1135
1024	*dst = (ullong)pack8888 (over_rev_non_pre (s, d), (Vector4x16)c.mmx_zero);	1136	*dst = (ullong)pack8888 (over_rev_non_pre (s, d), _mm_setzero_si64());
1025		1137
1026	w--;	1138	w--;
1027	dst++;	1139	dst++;
1028	src++;	1140	src++;
1029	}	1141	}
1030		1142
1031	while (w >= 2)	1143	while (w >= 2)
1032	{	1144	{
1033	ullong s0, s1;	1145	ullong s0, s1;
1034	unsigned char a0, a1;	1146	unsigned char a0, a1;
1035	Vector4x16 d0, d1;	1147	__m64 d0, d1;
1036		1148
1037	s0 = *src;	1149	s0 = *src;
1038	s1 = *(src + 1);	1150	s1 = *(src + 1);
1039		1151
1040	a0 = (s0 >> 24);	1152	a0 = (s0 >> 24);
1041	a1 = (s1 >> 24);	1153	a1 = (s1 >> 24);
1042		1154
Lines 1044-1060 Link Here
1044	{	1156	{
1045	d0 = invert_colors(load8888(s0));	1157	d0 = invert_colors(load8888(s0));
1046	d1 = invert_colors(load8888(s1));	1158	d1 = invert_colors(load8888(s1));
1047		1159
1048	(Vector8x8 )dst = pack8888 (d0, d1);	1160	(__m64 )dst = pack8888 (d0, d1);
1049	}	1161	}
1050	else if (a0 \| a1)	1162	else if (a0 \| a1)
1051	{	1163	{
1052	Vector4x16 vdest = (Vector4x16 )dst;	1164	__m64 vdest = (__m64 )dst;
1053		1165
1054	d0 = over_rev_non_pre (load8888(s0), expand8888 (vdest, 0));	1166	d0 = over_rev_non_pre (load8888(s0), expand8888 (vdest, 0));
1055	d1 = over_rev_non_pre (load8888(s1), expand8888 (vdest, 1));	1167	d1 = over_rev_non_pre (load8888(s1), expand8888 (vdest, 1));
1056		1168
1057	(Vector8x8 )dst = pack8888 (d0, d1);	1169	(__m64 )dst = pack8888 (d0, d1);
1058	}	1170	}
1059		1171
1060	w -= 2;	1172	w -= 2;
Lines 1064-1081 Link Here
1064		1176
1065	while (w)	1177	while (w)
1066	{	1178	{
1067	Vector4x16 s = load8888 (*src);	1179	__m64 s = load8888 (*src);
1068	Vector4x16 d = load8888 (*dst);	1180	__m64 d = load8888 (*dst);
1069		1181
1070	*dst = (ullong)pack8888 (over_rev_non_pre (s, d), (Vector4x16)c.mmx_zero);	1182	*dst = (ullong)pack8888 (over_rev_non_pre (s, d), _mm_setzero_si64());
1071		1183
1072	w--;	1184	w--;
1073	dst++;	1185	dst++;
1074	src++;	1186	src++;
1075	}	1187	}
1076	}	1188	}
1077		1189
1078	emms();	1190	_mm_empty();
1079	}	1191	}
1080		1192
1081	void	1193	void
Lines 1096-1102 Link Here
1096	CARD16 *dstLine;	1208	CARD16 *dstLine;
1097	CARD32 *maskLine;	1209	CARD32 *maskLine;
1098	FbStride dstStride, maskStride;	1210	FbStride dstStride, maskStride;
1099	Vector4x16 vsrc, vsrca;	1211	__m64 vsrc, vsrca;
1100		1212
1101	CHECKPOINT();	1213	CHECKPOINT();
1102		1214
Lines 1125-1131 Link Here
1125	if (m)	1237	if (m)
1126	{	1238	{
1127	ullong d = *q;	1239	ullong d = *q;
1128	Vector4x16 vdest = expand565 ((Vector4x16)d, 0);	1240	__m64 vdest = expand565 ((__m64)d, 0);
1129	vdest = pack565 (in_over (vsrc, vsrca, load8888 (m), vdest), vdest, 0);	1241	vdest = pack565 (in_over (vsrc, vsrca, load8888 (m), vdest), vdest, 0);
1130	*q = (ullong)vdest;	1242	*q = (ullong)vdest;
1131	}	1243	}
Lines 1146-1159 Link Here
1146		1258
1147	if ((m0 \| m1 \| m2 \| m3))	1259	if ((m0 \| m1 \| m2 \| m3))
1148	{	1260	{
1149	Vector4x16 vdest = (Vector4x16 )q;	1261	__m64 vdest = (__m64 )q;
1150		1262
1151	vdest = pack565(in_over(vsrc, vsrca, load8888(m0), expand565(vdest, 0)), vdest, 0);	1263	vdest = pack565(in_over(vsrc, vsrca, load8888(m0), expand565(vdest, 0)), vdest, 0);
1152	vdest = pack565(in_over(vsrc, vsrca, load8888(m1), expand565(vdest, 1)), vdest, 1);	1264	vdest = pack565(in_over(vsrc, vsrca, load8888(m1), expand565(vdest, 1)), vdest, 1);
1153	vdest = pack565(in_over(vsrc, vsrca, load8888(m2), expand565(vdest, 2)), vdest, 2);	1265	vdest = pack565(in_over(vsrc, vsrca, load8888(m2), expand565(vdest, 2)), vdest, 2);
1154	vdest = pack565(in_over(vsrc, vsrca, load8888(m3), expand565(vdest, 3)), vdest, 3);	1266	vdest = pack565(in_over(vsrc, vsrca, load8888(m3), expand565(vdest, 3)), vdest, 3);
1155		1267
1156	(Vector4x16 )q = vdest;	1268	(__m64 )q = vdest;
1157	}	1269	}
1158	twidth -= 4;	1270	twidth -= 4;
1159	p += 4;	1271	p += 4;
Lines 1168-1174 Link Here
1168	if (m)	1280	if (m)
1169	{	1281	{
1170	ullong d = *q;	1282	ullong d = *q;
1171	Vector4x16 vdest = expand565((Vector4x16)d, 0);	1283	__m64 vdest = expand565((__m64)d, 0);
1172	vdest = pack565 (in_over(vsrc, vsrca, load8888(m), vdest), vdest, 0);	1284	vdest = pack565 (in_over(vsrc, vsrca, load8888(m), vdest), vdest, 0);
1173	*q = (ullong)vdest;	1285	*q = (ullong)vdest;
1174	}	1286	}
Lines 1182-1188 Link Here
1182	dstLine += dstStride;	1294	dstLine += dstStride;
1183	}	1295	}
1184		1296
1185	emms ();	1297	_mm_empty ();
1186	}	1298	}
1187		1299
1188	void	1300	void
Lines 1210-1216 Link Here
1210		1322
1211	fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1);	1323	fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1);
1212	fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);	1324	fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
1213		1325
1214	while (height--)	1326	while (height--)
1215	{	1327	{
1216	dst = dstLine;	1328	dst = dstLine;
Lines 1218-1224 Link Here
1218	src = srcLine;	1330	src = srcLine;
1219	srcLine += srcStride;	1331	srcLine += srcStride;
1220	w = width;	1332	w = width;
1221		1333
1222	while (w && (unsigned long)dst & 7)	1334	while (w && (unsigned long)dst & 7)
1223	{	1335	{
1224	s = *src;	1336	s = *src;
Lines 1234-1246 Link Here
1234		1346
1235	while (w >= 8)	1347	while (w >= 8)
1236	{	1348	{
1237	__asm__ __volatile__ (	1349	(__m64)dst = _mm_adds_pu8((__m64)src, (__m64)dst);
1238	"movq (%0), %%mm2\n\t"
1239	"movq (%1), %%mm3\n\t"
1240	"paddusb %%mm2, %%mm3\n\t"
1241	"movq %%mm3, (%1)\n\t"
1242	: /* no output */ : "r" (src), "r" (dst));
1243
1244	dst += 8;	1350	dst += 8;
1245	src += 8;	1351	src += 8;
1246	w -= 8;	1352	w -= 8;
Lines 1259-1266 Link Here
1259	w--;	1365	w--;
1260	}	1366	}
1261	}	1367	}
1262		1368
1263	emms();	1369	_mm_empty();
1264	}	1370	}
1265		1371
1266	void	1372	void
Lines 1297-1309 Link Here
1297		1403
1298	while (w && (unsigned long)dst & 7)	1404	while (w && (unsigned long)dst & 7)
1299	{	1405	{
1300	__asm__ __volatile__ (	1406	dst = _mm_cvtsi64_si32(_mm_adds_pu8(_mm_cvtsi32_si64(src),
1301	"movd %0, %%mm2\n\t"	1407	_mm_cvtsi32_si64(*dst)));
1302	"movd %1, %%mm3\n\t"
1303	"paddusb %%mm2, %%mm3\n\t"
1304	"movd %%mm3, %1\n\t"
1305	: /* no output / : "m" (src), "m" (*dst));
1306
1307	dst++;	1408	dst++;
1308	src++;	1409	src++;
1309	w--;	1410	w--;
Lines 1311-1323 Link Here
1311		1412
1312	while (w >= 2)	1413	while (w >= 2)
1313	{	1414	{
1314	__asm__ __volatile__ (	1415	(ullong)dst = (ullong) _mm_adds_pu8((__m64)src, (__m64)dst);
1315	"movq (%0), %%mm2\n\t"
1316	"movq (%1), %%mm3\n\t"
1317	"paddusb %%mm2, %%mm3\n\t"
1318	"movq %%mm3, (%1)\n\t"
1319	: /* no output */ : "r" (src), "r" (dst));
1320
1321	dst += 2;	1416	dst += 2;
1322	src += 2;	1417	src += 2;
1323	w -= 2;	1418	w -= 2;
Lines 1325-1340 Link Here
1325		1420
1326	if (w)	1421	if (w)
1327	{	1422	{
1328	__asm__ __volatile__ (	1423	dst = _mm_cvtsi64_si32(_mm_adds_pu8(_mm_cvtsi32_si64(src),
1329	"movd %0, %%mm2\n\t"	1424	_mm_cvtsi32_si64(*dst)));
1330	"movd %1, %%mm3\n\t"	1425
1331	"paddusb %%mm2, %%mm3\n\t"
1332	"movd %%mm3, %1\n\t"
1333	: /* no output / : "m" (src), "m" (*dst));
1334	}	1426	}
1335	}	1427	}
1336		1428
1337	emms();	1429	_mm_empty();
1338	}	1430	}
1339		1431
1340	#define GetStart(drw,x,y,type,stride,line,bpp) {\	1432	#define GetStart(drw,x,y,type,stride,line,bpp) {\
Lines 1358-1376 Link Here
1358	FbStride stride;	1450	FbStride stride;
1359	int bpp;	1451	int bpp;
1360	ullong fill;	1452	ullong fill;
1361	Vector8x8 vfill;	1453	__m64 vfill;
1362	CARD32 byte_width;	1454	CARD32 byte_width;
1363	CARD8 *byte_line;	1455	CARD8 *byte_line;
1364	FbBits *bits;	1456	FbBits *bits;
1365	int xoff, yoff;	1457	int xoff, yoff;
1366		1458
1367	CHECKPOINT();	1459	CHECKPOINT();
1368		1460
1369	fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff);	1461	fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff);
1370		1462
1371	if (bpp == 16 && (xor >> 16 != (xor & 0xffff)))	1463	if (bpp == 16 && (xor >> 16 != (xor & 0xffff)))
1372	return FALSE;	1464	return FALSE;
1373		1465
1374	if (bpp != 16 && bpp != 32)	1466	if (bpp != 16 && bpp != 32)
1375	return FALSE;	1467	return FALSE;
1376		1468
Lines 1388-1396 Link Here
1388	byte_width = 4 * width;	1480	byte_width = 4 * width;
1389	stride *= 4;	1481	stride *= 4;
1390	}	1482	}
1391		1483
1392	fill = ((ullong)xor << 32) \| xor;	1484	fill = ((ullong)xor << 32) \| xor;
1393	vfill = (Vector8x8)fill;	1485	vfill = (__m64)fill;
1394		1486
1395	while (height--)	1487	while (height--)
1396	{	1488	{
Lines 1398-1404 Link Here
1398	CARD8 *d = byte_line;	1490	CARD8 *d = byte_line;
1399	byte_line += stride;	1491	byte_line += stride;
1400	w = byte_width;	1492	w = byte_width;
1401		1493
1402	while (w >= 2 && ((unsigned long)d & 3))	1494	while (w >= 2 && ((unsigned long)d & 3))
1403	{	1495	{
1404	(CARD16 )d = xor;	1496	(CARD16 )d = xor;
Lines 1406-1440 Link Here
1406	d += 2;	1498	d += 2;
1407	}	1499	}
1408		1500
1409	while (w >= 4 && ((unsigned int)d & 7))	1501	while (w >= 4 && ((unsigned long)d & 7))
1410	{	1502	{
1411	(CARD32 )d = xor;	1503	(CARD32 )d = xor;
1412		1504
1413	w -= 4;	1505	w -= 4;
1414	d += 4;	1506	d += 4;
1415	}	1507	}
1416		1508
1417	while (w >= 64)	1509	while (w >= 64)
1418	{	1510	{
1419	__asm__ __volatile (	1511	(__m64) (d + 0) = vfill;
1420	"movq %0, (%1)\n\t"	1512	(__m64) (d + 8) = vfill;
1421	"movq %0, 8(%1)\n\t"	1513	(__m64) (d + 16) = vfill;
1422	"movq %0, 16(%1)\n\t"	1514	(__m64) (d + 24) = vfill;
1423	"movq %0, 24(%1)\n\t"	1515	(__m64) (d + 32) = vfill;
1424	"movq %0, 32(%1)\n\t"	1516	(__m64) (d + 40) = vfill;
1425	"movq %0, 40(%1)\n\t"	1517	(__m64) (d + 48) = vfill;
1426	"movq %0, 48(%1)\n\t"	1518	(__m64) (d + 56) = vfill;
1427	"movq %0, 56(%1)\n\t"	1519
1428	: /* no output */
1429	: "y" (vfill), "r" (d)
1430	: "memory");
1431	w -= 64;	1520	w -= 64;
1432	d += 64;	1521	d += 64;
1433	}	1522	}
1434	while (w >= 4)	1523	while (w >= 4)
1435	{	1524	{
1436	(CARD32 )d = xor;	1525	(CARD32 )d = xor;
1437		1526
1438	w -= 4;	1527	w -= 4;
1439	d += 4;	1528	d += 4;
1440	}	1529	}
Lines 1446-1461 Link Here
1446	}	1535	}
1447	}	1536	}
1448		1537
1449	emms();	1538	_mm_empty();
		1539	return TRUE;
		1540	}
		1541
		1542	Bool
		1543	fbCopyAreammx (DrawablePtr pSrc,
		1544	DrawablePtr pDst,
		1545	int src_x,
		1546	int src_y,
		1547	int dst_x,
		1548	int dst_y,
		1549	int width,
		1550	int height)
		1551	{
		1552	FbBits * src_bits;
		1553	FbStride src_stride;
		1554	int src_bpp;
		1555	int src_xoff;
		1556	int src_yoff;
		1557
		1558	FbBits * dst_bits;
		1559	FbStride dst_stride;
		1560	int dst_bpp;
		1561	int dst_xoff;
		1562	int dst_yoff;
		1563
		1564	CARD8 * src_bytes;
		1565	CARD8 * dst_bytes;
		1566	int byte_width;
		1567
		1568	fbGetDrawable(pSrc, src_bits, src_stride, src_bpp, src_xoff, src_yoff);
		1569	fbGetDrawable(pDst, dst_bits, dst_stride, dst_bpp, dst_xoff, dst_yoff);
		1570
		1571	if (src_bpp != 16 && src_bpp != 32)
		1572	return FALSE;
		1573
		1574	if (dst_bpp != 16 && dst_bpp != 32)
		1575	return FALSE;
		1576
		1577	if (src_bpp != dst_bpp)
		1578	{
		1579	return FALSE;
		1580	}
		1581
		1582	if (src_bpp == 16)
		1583	{
		1584	src_stride = src_stride * sizeof (FbBits) / 2;
		1585	dst_stride = dst_stride * sizeof (FbBits) / 2;
		1586	src_bytes = (CARD8 )(((CARD16 )src_bits) + src_stride * (src_y - src_yoff) + (src_x - src_xoff));
		1587	dst_bytes = (CARD8 )(((CARD16 )dst_bits) + dst_stride * (dst_y - dst_yoff) + (dst_x - dst_xoff));
		1588	byte_width = 2 * width;
		1589	src_stride *= 2;
		1590	dst_stride *= 2;
		1591	}
		1592	else
		1593	{
		1594	src_stride = src_stride * sizeof (FbBits) / 4;
		1595	dst_stride = dst_stride * sizeof (FbBits) / 4;
		1596	src_bytes = (CARD8 )(((CARD32 )src_bits) + src_stride * (src_y - src_yoff) + (src_x - src_xoff));
		1597	dst_bytes = (CARD8 )(((CARD32 )dst_bits) + dst_stride * (dst_y - dst_yoff) + (dst_x - dst_xoff));
		1598	byte_width = 4 * width;
		1599	src_stride *= 4;
		1600	dst_stride *= 4;
		1601	}
		1602
1603	while (height--)
1604	{
1605	int w;
1606	CARD8 *s = src_bytes;
1607	CARD8 *d = dst_bytes;
1608	src_bytes += src_stride;
1609	dst_bytes += dst_stride;
1610	w = byte_width;
1611
1612	while (w >= 2 && ((unsigned long)d & 3))
1613	{
1614	(CARD16 )d = (CARD16 )s;
1615	w -= 2;
1616	s += 2;
1617	d += 2;
1618	}
1619
1620	while (w >= 4 && ((unsigned int)d & 7))
1621	{
1622	(CARD32 )d = (CARD32 )s;
1623
1624	w -= 4;
1625	s += 4;
1626	d += 4;
1627	}
1628
1629	while (w >= 64)
1630	{
1631	(__m64 )(d + 0) = (__m64 )(s + 0);
1632	(__m64 )(d + 8) = (__m64 )(s + 8);
1633	(__m64 )(d + 16) = (__m64 )(s + 16);
1634	(__m64 )(d + 24) = (__m64 )(s + 24);
1635	(__m64 )(d + 32) = (__m64 )(s + 32);
1636	(__m64 )(d + 40) = (__m64 )(s + 40);
1637	(__m64 )(d + 48) = (__m64 )(s + 48);
1638	(__m64 )(d + 56) = (__m64 )(s + 56);
1639	w -= 64;
1640	s += 64;
1641	d += 64;
1642	}
1643	while (w >= 4)
1644	{
1645	(CARD32 )d = (CARD32 )s;
1646
1647	w -= 4;
1648	s += 4;
1649	d += 4;
1650	}
1651	if (w >= 2)
1652	{
1653	(CARD16 )d = (CARD16 )s;
1654	w -= 2;
1655	s += 2;
1656	d += 2;
1657	}
1658	}
1659
1660	_mm_empty();
1450	return TRUE;	1661	return TRUE;
1451	}	1662	}
1452		1663
		1664	void
		1665	fbCompositeCopyAreammx (CARD8 op,
		1666	PicturePtr pSrc,
		1667	PicturePtr pMask,
		1668	PicturePtr pDst,
		1669	INT16 xSrc,
		1670	INT16 ySrc,
		1671	INT16 xMask,
		1672	INT16 yMask,
		1673	INT16 xDst,
		1674	INT16 yDst,
		1675	CARD16 width,
		1676	CARD16 height)
		1677	{
		1678	fbCopyAreammx (pSrc->pDrawable,
		1679	pDst->pDrawable,
		1680	xSrc, ySrc,
		1681	xDst, yDst,
		1682	width, height);
		1683	}
		1684
		1685	#ifndef __amd64__
1453	Bool	1686	Bool
1454	fbHaveMMX (void)	1687	fbHaveMMX (void)
1455	{	1688	{
1456	static Bool initialized = FALSE;	1689	static Bool initialized = FALSE;
1457	static Bool mmx_present;	1690	static Bool mmx_present;
1458		1691
1459	if (!initialized)	1692	if (!initialized)
1460	{	1693	{
1461	int tmp; /* static variables are accessed through %ebx,	1694	int tmp; /* static variables are accessed through %ebx,
Lines 1466-1472 Link Here
1466		1699
1467	__asm__ __volatile__ (	1700	__asm__ __volatile__ (
1468	/* Check if bit 21 in flags word is writeable */	1701	/* Check if bit 21 in flags word is writeable */
1469		1702
1470	"pusha \n\t"	1703	"pusha \n\t"
1471	"pushfl \n\t"	1704	"pushfl \n\t"
1472	"popl %%eax \n\t"	1705	"popl %%eax \n\t"
Lines 1502-1514 Link Here
1502	: /* no input */);	1735	: /* no input */);
1503		1736
1504	initialized = TRUE;	1737	initialized = TRUE;
1505		1738
1506	mmx_present = tmp;	1739	mmx_present = tmp;
1507	}	1740	}
1508		1741
1509	return mmx_present;	1742	return mmx_present;
1510	}	1743	}
		1744	#endif /* __amd64__ */
1511		1745
1512		1746
1513	#endif /* RENDER */	1747	#endif /* RENDER */
1514	#endif /* USE_GCC34_MMX */	1748	#endif /* USE_MMX */

Lines 1-5 Link Here

(-)xc-orig/programs/Xserver/fb/fbmmx.h (-6 / +44 lines)
1	/*	1	/*
2	* Copyright © 2004 Red Hat, Inc.	2	* Copyright Â© 2004 Red Hat, Inc.
3	*	3	*
4	* Permission to use, copy, modify, distribute, and sell this software and its	4	* Permission to use, copy, modify, distribute, and sell this software and its
5	* documentation for any purpose is hereby granted without fee, provided that	5	* documentation for any purpose is hereby granted without fee, provided that
Lines 18-34 Link Here
18	* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN	18	* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
19	* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.	19	* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
20	*	20	*
21	* Author: Søren Sandmann (sandmann@redhat.com)	21	* Author: SÃ¸ren Sandmann (sandmann@redhat.com)
22	*	22	*
23	* Based on work by Owen Taylor	23	* Based on work by Owen Taylor
24	*/	24	*/
25	#ifdef USE_GCC34_MMX	25	#ifdef USE_MMX
		26
		27	#ifndef __amd64__
26	Bool fbHaveMMX(void);	28	Bool fbHaveMMX(void);
27	#else	29	#else
28	#define fbHaveMMX FALSE	30	#define fbHaveMMX() TRUE
		31	#endif
		32
		33	#else
		34	#define fbHaveMMX() FALSE
29	#endif	35	#endif
30		36
31	#ifdef USE_GCC34_MMX	37	#ifdef USE_MMX
32		38
33	void fbCompositeSolidMask_nx8888x0565Cmmx (CARD8 op,	39	void fbCompositeSolidMask_nx8888x0565Cmmx (CARD8 op,
34	PicturePtr pSrc,	40	PicturePtr pSrc,
Lines 150-155 Link Here
150	INT16 yDst,	156	INT16 yDst,
151	CARD16 width,	157	CARD16 width,
152	CARD16 height);	158	CARD16 height);
		159	void fbCompositeSrc_8888x8x8888mmx (CARD8 op,
		160	PicturePtr pSrc,
		161	PicturePtr pMask,
		162	PicturePtr pDst,
		163	INT16 xSrc,
		164	INT16 ySrc,
		165	INT16 xMask,
		166	INT16 yMask,
		167	INT16 xDst,
		168	INT16 yDst,
		169	CARD16 width,
		170	CARD16 height);
		171	Bool fbCopyAreammx (DrawablePtr pSrc,
		172	DrawablePtr pDst,
		173	int src_x,
		174	int src_y,
		175	int dst_x,
		176	int dst_y,
		177	int width,
		178	int height);
		179	void fbCompositeCopyAreammx (CARD8 op,
		180	PicturePtr pSrc,
		181	PicturePtr pMask,
		182	PicturePtr pDst,
		183	INT16 xSrc,
		184	INT16 ySrc,
		185	INT16 xMask,
		186	INT16 yMask,
		187	INT16 xDst,
		188	INT16 yDst,
		189	CARD16 width,
		190	CARD16 height);
153	Bool fbSolidFillmmx (DrawablePtr pDraw,	191	Bool fbSolidFillmmx (DrawablePtr pDraw,
154	int x,	192	int x,
155	int y,	193	int y,
Lines 157-160 Link Here
157	int height,	195	int height,
158	FbBits xor);	196	FbBits xor);
159		197
160	#endif /* USE_GCC34_MMX */	198	#endif /* USE_MMX */

Lines 1-7 Link Here

(-)xc-orig/programs/Xserver/fb/fbpict.c (-17 / +66 lines)
1	/*	1	/*
2	* $XFree86: xc/programs/Xserver/fb/fbpict.c,v 1.15 2002/09/26 02:56:48 keithp Exp $	2	* $XFree86: xc/programs/Xserver/fb/fbpict.c,v 1.15 2002/09/26 02:56:48 keithp Exp $
3	*	3	*
4	* Copyright © 2000 SuSE, Inc.	4	* Copyright Â© 2000 SuSE, Inc.
5	*	5	*
6	* Permission to use, copy, modify, distribute, and sell this software and its	6	* Permission to use, copy, modify, distribute, and sell this software and its
7	* documentation for any purpose is hereby granted without fee, provided that	7	* documentation for any purpose is hereby granted without fee, provided that
Lines 863-868 Link Here
863	if (!pSrc->transform && !(pMask && pMask->transform))	863	if (!pSrc->transform && !(pMask && pMask->transform))
864	if (!maskAlphaMap && !srcAlphaMap && !dstAlphaMap)	864	if (!maskAlphaMap && !srcAlphaMap && !dstAlphaMap)
865	switch (op) {	865	switch (op) {
		866	case PictOpSrc:
		867	#ifdef USE_MMX
		868	if (!pMask && pSrc->format == pDst->format &&
		869	pSrc->pDrawable != pDst->pDrawable)
		870	{
		871	func = fbCompositeCopyAreammx;
		872	}
		873	#endif
		874	break;
866	case PictOpOver:	875	case PictOpOver:
867	if (pMask)	876	if (pMask)
868	{	877	{
Lines 877-883 Link Here
877	switch (pDst->format) {	886	switch (pDst->format) {
878	case PICT_r5g6b5:	887	case PICT_r5g6b5:
879	case PICT_b5g6r5:	888	case PICT_b5g6r5:
880	#ifdef USE_GCC34_MMX	889	#ifdef USE_MMX
881	if (fbHaveMMX())	890	if (fbHaveMMX())
882	func = fbCompositeSolidMask_nx8x0565mmx;	891	func = fbCompositeSolidMask_nx8x0565mmx;
883	else	892	else
Lines 892-898 Link Here
892	case PICT_x8r8g8b8:	901	case PICT_x8r8g8b8:
893	case PICT_a8b8g8r8:	902	case PICT_a8b8g8r8:
894	case PICT_x8b8g8r8:	903	case PICT_x8b8g8r8:
895	#ifdef USE_GCC34_MMX	904	#ifdef USE_MMX
896	if (fbHaveMMX())	905	if (fbHaveMMX())
897	func = fbCompositeSolidMask_nx8x8888mmx;	906	func = fbCompositeSolidMask_nx8x8888mmx;
898	else	907	else
Lines 906-912 Link Here
906	switch (pDst->format) {	915	switch (pDst->format) {
907	case PICT_a8r8g8b8:	916	case PICT_a8r8g8b8:
908	case PICT_x8r8g8b8:	917	case PICT_x8r8g8b8:
909	#ifdef USE_GCC34_MMX	918	#ifdef USE_MMX
910	if (fbHaveMMX())	919	if (fbHaveMMX())
911	func = fbCompositeSolidMask_nx8888x8888Cmmx;	920	func = fbCompositeSolidMask_nx8888x8888Cmmx;
912	else	921	else
Lines 914-920 Link Here
914	func = fbCompositeSolidMask_nx8888x8888C;	923	func = fbCompositeSolidMask_nx8888x8888C;
915	break;	924	break;
916	case PICT_r5g6b5:	925	case PICT_r5g6b5:
917	#ifdef USE_GCC34_MMX	926	#ifdef USE_MMX
918	if (fbHaveMMX())	927	if (fbHaveMMX())
919	func = fbCompositeSolidMask_nx8888x0565Cmmx;	928	func = fbCompositeSolidMask_nx8888x0565Cmmx;
920	else	929	else
Lines 929-935 Link Here
929	switch (pDst->format) {	938	switch (pDst->format) {
930	case PICT_a8b8g8r8:	939	case PICT_a8b8g8r8:
931	case PICT_x8b8g8r8:	940	case PICT_x8b8g8r8:
932	#ifdef USE_GCC34_MMX	941	#ifdef USE_MMX
933	if (fbHaveMMX())	942	if (fbHaveMMX())
934	func = fbCompositeSolidMask_nx8888x8888Cmmx;	943	func = fbCompositeSolidMask_nx8888x8888Cmmx;
935	else	944	else
Lines 937-943 Link Here
937	func = fbCompositeSolidMask_nx8888x8888C;	946	func = fbCompositeSolidMask_nx8888x8888C;
938	break;	947	break;
939	case PICT_b5g6r5:	948	case PICT_b5g6r5:
940	#ifdef USE_GCC34_MMX	949	#ifdef USE_MMX
941	if (fbHaveMMX())	950	if (fbHaveMMX())
942	func = fbCompositeSolidMask_nx8888x0565Cmmx;	951	func = fbCompositeSolidMask_nx8888x0565Cmmx;
943	else	952	else
Lines 970-975 Link Here
970	xSrc == xMask && ySrc == yMask &&	979	xSrc == xMask && ySrc == yMask &&
971	!pMask->componentAlpha)	980	!pMask->componentAlpha)
972	{	981	{
		982	/* source == mask: non-premultiplied data */
973	switch (pSrc->format) {	983	switch (pSrc->format) {
974	case PICT_x8b8g8r8:	984	case PICT_x8b8g8r8:
975	switch (pMask->format) {	985	switch (pMask->format) {
Lines 978-990 Link Here
978	switch (pDst->format) {	988	switch (pDst->format) {
979	case PICT_a8r8g8b8:	989	case PICT_a8r8g8b8:
980	case PICT_x8r8g8b8:	990	case PICT_x8r8g8b8:
981	#ifdef USE_GCC34_MMX	991	#ifdef USE_MMX
982	if (fbHaveMMX())	992	if (fbHaveMMX())
983	func = fbCompositeSrc_8888RevNPx8888mmx;	993	func = fbCompositeSrc_8888RevNPx8888mmx;
984	#endif	994	#endif
985	break;	995	break;
986	case PICT_r5g6b5:	996	case PICT_r5g6b5:
987	#ifdef USE_GCC34_MMX	997	#ifdef USE_MMX
988	if (fbHaveMMX())	998	if (fbHaveMMX())
989	func = fbCompositeSrc_8888RevNPx0565mmx;	999	func = fbCompositeSrc_8888RevNPx0565mmx;
990	#endif	1000	#endif
Lines 1000-1012 Link Here
1000	switch (pDst->format) {	1010	switch (pDst->format) {
1001	case PICT_a8b8g8r8:	1011	case PICT_a8b8g8r8:
1002	case PICT_x8b8g8r8:	1012	case PICT_x8b8g8r8:
1003	#ifdef USE_GCC34_MMX	1013	#ifdef USE_MMX
1004	if (fbHaveMMX())	1014	if (fbHaveMMX())
1005	func = fbCompositeSrc_8888RevNPx8888mmx;	1015	func = fbCompositeSrc_8888RevNPx8888mmx;
1006	#endif	1016	#endif
1007	break;	1017	break;
1008	case PICT_r5g6b5:	1018	case PICT_r5g6b5:
1009	#ifdef USE_GCC34_MMX	1019	#ifdef USE_MMX
1010	if (fbHaveMMX())	1020	if (fbHaveMMX())
1011	func = fbCompositeSrc_8888RevNPx0565mmx;	1021	func = fbCompositeSrc_8888RevNPx0565mmx;
1012	#endif	1022	#endif
Lines 1018-1026 Link Here
1018	}	1028	}
1019	break;	1029	break;
1020	}	1030	}
		1031	else
		1032	{
		1033	/* non-repeating source, repeating mask => translucent window */
		1034	if (maskRepeat &&
		1035	pMask->pDrawable->width == 1 &&
		1036	pMask->pDrawable->height == 1)
		1037	{
		1038	if (pSrc->format == PICT_x8r8g8b8 &&
		1039	pDst->format == PICT_x8r8g8b8 &&
		1040	pMask->format == PICT_a8)
		1041	{
		1042	#ifdef USE_MMX
		1043	if (fbHaveMMX())
		1044	func = fbCompositeSrc_8888x8x8888mmx;
		1045	#endif
		1046	}
		1047	}
		1048	}
1021	}	1049	}
1022	}	1050	}
1023	else	1051	else /* no mask */
1024	{	1052	{
1025	if (srcRepeat &&	1053	if (srcRepeat &&
1026	pSrc->pDrawable->width == 1 &&	1054	pSrc->pDrawable->width == 1 &&
Lines 1032-1038 Link Here
1032	switch (pDst->format) {	1060	switch (pDst->format) {
1033	case PICT_a8r8g8b8:	1061	case PICT_a8r8g8b8:
1034	case PICT_x8r8g8b8:	1062	case PICT_x8r8g8b8:
1035	#ifdef USE_GCC34_MMX	1063	#ifdef USE_MMX
1036	if (fbHaveMMX())	1064	if (fbHaveMMX())
1037	{	1065	{
1038	srcRepeat = FALSE;	1066	srcRepeat = FALSE;
Lines 1041-1047 Link Here
1041	#endif	1069	#endif
1042	break;	1070	break;
1043	case PICT_r5g6b5:	1071	case PICT_r5g6b5:
1044	#ifdef USE_GCC34_MMX	1072	#ifdef USE_MMX
1045	if (fbHaveMMX())	1073	if (fbHaveMMX())
1046	{	1074	{
1047	srcRepeat = FALSE;	1075	srcRepeat = FALSE;
Lines 1070-1075 Link Here
1070	break;	1098	break;
1071	}	1099	}
1072	break;	1100	break;
		1101	case PICT_x8r8g8b8:
		1102	switch (pDst->format) {
		1103	case PICT_a8r8g8b8:
		1104	case PICT_x8r8g8b8:
		1105	#ifdef USE_MMX
		1106	if (fbHaveMMX())
		1107	func = fbCompositeCopyAreammx;
		1108	#endif
		1109	break;
		1110	}
		1111	case PICT_x8b8g8r8:
		1112	switch (pDst->format) {
		1113	case PICT_a8b8g8r8:
		1114	case PICT_x8b8g8r8:
		1115	#ifdef USE_MMX
		1116	if (fbHaveMMX())
		1117	func = fbCompositeCopyAreammx;
		1118	#endif
		1119	break;
		1120	}
		1121	break;
1073	case PICT_a8b8g8r8:	1122	case PICT_a8b8g8r8:
1074	switch (pDst->format) {	1123	switch (pDst->format) {
1075	case PICT_a8b8g8r8:	1124	case PICT_a8b8g8r8:
Lines 1109-1115 Link Here
1109	case PICT_a8r8g8b8:	1158	case PICT_a8r8g8b8:
1110	switch (pDst->format) {	1159	switch (pDst->format) {
1111	case PICT_a8r8g8b8:	1160	case PICT_a8r8g8b8:
1112	#ifdef USE_GCC34_MMX	1161	#ifdef USE_MMX
1113	if (fbHaveMMX())	1162	if (fbHaveMMX())
1114	func = fbCompositeSrcAdd_8888x8888mmx;	1163	func = fbCompositeSrcAdd_8888x8888mmx;
1115	else	1164	else
Lines 1121-1127 Link Here
1121	case PICT_a8b8g8r8:	1170	case PICT_a8b8g8r8:
1122	switch (pDst->format) {	1171	switch (pDst->format) {
1123	case PICT_a8b8g8r8:	1172	case PICT_a8b8g8r8:
1124	#ifdef USE_GCC34_MMX	1173	#ifdef USE_MMX
1125	if (fbHaveMMX())	1174	if (fbHaveMMX())
1126	func = fbCompositeSrcAdd_8888x8888mmx;	1175	func = fbCompositeSrcAdd_8888x8888mmx;
1127	else	1176	else
Lines 1133-1139 Link Here
1133	case PICT_a8:	1182	case PICT_a8:
1134	switch (pDst->format) {	1183	switch (pDst->format) {
1135	case PICT_a8:	1184	case PICT_a8:
1136	#ifdef USE_GCC34_MMX	1185	#ifdef USE_MMX
1137	if (fbHaveMMX())	1186	if (fbHaveMMX())
1138	func = fbCompositeSrcAdd_8000x8000mmx;	1187	func = fbCompositeSrcAdd_8000x8000mmx;
1139	else	1188	else

Return to bug 80685

Lines 1-8 Link Here

(-)xc-orig/programs/Xserver/fb/fbcompose.c (-4 / +17 lines)
1	/*	1	/*
2	* $XdotOrg: xc/programs/Xserver/fb/fbcompose.c,v 1.3 2004/05/12 01:49:46 anholt Exp $	2	* $XdotOrg: xc/programs/Xserver/fb/fbcompose.c,v 1.5 2005/01/13 20:49:21 sandmann Exp $
3	* $XFree86: xc/programs/Xserver/fb/fbcompose.c,v 1.17tsi Exp $	3	* $XFree86: xc/programs/Xserver/fb/fbcompose.c,v 1.17tsi Exp $
4	*	4	*
5	* Copyright © 2000 Keith Packard, member of The XFree86 Project, Inc.	5	* Copyright Â© 2000 Keith Packard, member of The XFree86 Project, Inc.
6	*	6	*
7	* Permission to use, copy, modify, distribute, and sell this software and its	7	* Permission to use, copy, modify, distribute, and sell this software and its
8	* documentation for any purpose is hereby granted without fee, provided that	8	* documentation for any purpose is hereby granted without fee, provided that
Lines 2693-2699 Link Here
2693	op->u.transform.y = y - op->u.transform.top_y;	2693	op->u.transform.y = y - op->u.transform.top_y;
2694	}	2694	}
2695		2695
2696
2697	Bool	2696	Bool
2698	fbBuildCompositeOperand (PicturePtr pPict,	2697	fbBuildCompositeOperand (PicturePtr pPict,
2699	FbCompositeOperand op[4],	2698	FbCompositeOperand op[4],
Lines 2710-2716 Link Here
2710		2709
2711	op->u.transform.top_y = pPict->pDrawable->y;	2710	op->u.transform.top_y = pPict->pDrawable->y;
2712	op->u.transform.left_x = pPict->pDrawable->x;	2711	op->u.transform.left_x = pPict->pDrawable->x;
2713
2714	op->u.transform.start_x = x - op->u.transform.left_x;	2712	op->u.transform.start_x = x - op->u.transform.left_x;
2715	op->u.transform.x = op->u.transform.start_x;	2713	op->u.transform.x = op->u.transform.start_x;
2716	op->u.transform.y = y - op->u.transform.top_y;	2714	op->u.transform.y = y - op->u.transform.top_y;
Lines 2822-2827 Link Here
2822	FbCombineFunc f;	2820	FbCombineFunc f;
2823	int w;	2821	int w;
2824		2822
		2823	#if 0
		2824	ErrorF ("op: %d\n"
		2825	"src format: %lx\n"
		2826	"msk format %lx\n"
		2827	"dst format %lx\n"
		2828	"width: %d\n"
		2829	"height %d\n",
		2830	op,
		2831	pSrc? pSrc->format : 0,
		2832	pMask? pMask->format : 0,
		2833	pDst? pDst->format : 0,
		2834	width, height);
		2835	ErrorF ("PICT_x8r8g8b8: %lx\n", PICT_x8r8g8b8);
		2836	#endif
		2837
2825	if (!fbBuildCompositeOperand (pSrc, src, xSrc, ySrc, TRUE, TRUE))	2838	if (!fbBuildCompositeOperand (pSrc, src, xSrc, ySrc, TRUE, TRUE))
2826	return;	2839	return;
2827	if (!fbBuildCompositeOperand (pDst, dst, xDst, yDst, FALSE, TRUE))	2840	if (!fbBuildCompositeOperand (pDst, dst, xDst, yDst, FALSE, TRUE))

Lines 1-7 Link Here

(-)xc-orig/programs/Xserver/fb/fbcopy.c (-8 / +32 lines)
1	/*	1	/*
2	* Id: fbcopy.c,v 1.1 1999/11/02 03:54:45 keithp Exp $	2	* Id: fbcopy.c,v 1.1 1999/11/02 03:54:45 keithp Exp $
3	*	3	*
4	* Copyright © 1998 Keith Packard	4	* Copyright Â© 1998 Keith Packard
5	*	5	*
6	* Permission to use, copy, modify, distribute, and sell this software and its	6	* Permission to use, copy, modify, distribute, and sell this software and its
7	* documentation for any purpose is hereby granted without fee, provided that	7	* documentation for any purpose is hereby granted without fee, provided that
Lines 27-32 Link Here
27	#ifdef IN_MODULE	27	#ifdef IN_MODULE
28	#include "xf86_ansic.h"	28	#include "xf86_ansic.h"
29	#endif	29	#endif
		30	#include "fbmmx.h"
30		31
31	void	32	void
32	fbCopyNtoN (DrawablePtr pSrcDrawable,	33	fbCopyNtoN (DrawablePtr pSrcDrawable,
Lines 54-81 Link Here
54		55
55	fbGetDrawable (pSrcDrawable, src, srcStride, srcBpp, srcXoff, srcYoff);	56	fbGetDrawable (pSrcDrawable, src, srcStride, srcBpp, srcXoff, srcYoff);
56	fbGetDrawable (pDstDrawable, dst, dstStride, dstBpp, dstXoff, dstYoff);	57	fbGetDrawable (pDstDrawable, dst, dstStride, dstBpp, dstXoff, dstYoff);
57		58
58	while (nbox--)	59	while (nbox--)
59	{	60	{
		61	#ifdef USE_MMX
		62	if (!reverse && !upsidedown && fbHaveMMX())
		63	{
		64	if (!fbCopyAreammx (pSrcDrawable,
		65	pDstDrawable,
		66
		67	(pbox->x1 + dx + srcXoff),
		68	(pbox->y1 + dy + srcYoff),
		69
		70	(pbox->x1 + dstXoff),
		71	(pbox->y1 + dstYoff),
		72
		73	(pbox->x2 - pbox->x1),
		74	(pbox->y2 - pbox->y1)))
		75	goto fallback;
		76	else
		77	goto next;
		78	}
		79	fallback:
		80	#endif
60	fbBlt (src + (pbox->y1 + dy + srcYoff) * srcStride,	81	fbBlt (src + (pbox->y1 + dy + srcYoff) * srcStride,
61	srcStride,	82	srcStride,
62	(pbox->x1 + dx + srcXoff) * srcBpp,	83	(pbox->x1 + dx + srcXoff) * srcBpp,
63		84
64	dst + (pbox->y1 + dstYoff) * dstStride,	85	dst + (pbox->y1 + dstYoff) * dstStride,
65	dstStride,	86	dstStride,
66	(pbox->x1 + dstXoff) * dstBpp,	87	(pbox->x1 + dstXoff) * dstBpp,
67		88
68	(pbox->x2 - pbox->x1) * dstBpp,	89	(pbox->x2 - pbox->x1) * dstBpp,
69	(pbox->y2 - pbox->y1),	90	(pbox->y2 - pbox->y1),
70		91
71	alu,	92	alu,
72	pm,	93	pm,
73	dstBpp,	94	dstBpp,
74		95
75	reverse,	96	reverse,
76	upsidedown);	97	upsidedown);
		98	#ifdef USE_MMX
		99	next:
		100	#endif
77	pbox++;	101	pbox++;
78	}	102	}
79	}	103	}
80		104
81	void	105	void
Lines 594-600 Link Here
594	int yOut)	618	int yOut)
595	{	619	{
596	fbCopyProc copy;	620	fbCopyProc copy;
597		621
598	#ifdef FB_24_32BIT	622	#ifdef FB_24_32BIT
599	if (pSrcDrawable->bitsPerPixel != pDstDrawable->bitsPerPixel)	623	if (pSrcDrawable->bitsPerPixel != pDstDrawable->bitsPerPixel)
600	copy = fb24_32CopyMtoN;	624	copy = fb24_32CopyMtoN;