summaryrefslogtreecommitdiffstats
path: root/postproc/yuv2rgb_template.c
blob: a2d906f63574ac5b86f1da6c4ddfa0dc11db7422 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520

/*
 * yuv2rgb_mmx.c, Software YUV to RGB coverter with Intel MMX "technology"
 *
 * Copyright (C) 2000, Silicon Integrated System Corp.
 * All Rights Reserved.
 *
 * Author: Olie Lho <ollie@sis.com.tw>
 *
 * This file is part of mpeg2dec, a free MPEG-2 video decoder
 *
 * mpeg2dec is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2, or (at your option)
 * any later version.
 *
 * mpeg2dec is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with GNU Make; see the file COPYING. If not, write to
 * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA.
 *
 * 15 and 24 bpp support from Michael Niedermayer (michaelni@gmx.at)
 */

#include <stdio.h>
#include <stdlib.h>

#include "../config.h"

//#include "libmpeg2/mpeg2.h"
//#include "libmpeg2/mpeg2_internal.h"
#include <inttypes.h>

#include "rgb2rgb.h"
#include "../mmx_defs.h"

/* hope these constant values are cache line aligned */
uint64_t __attribute__((aligned(8))) mmx_80w = 0x0080008000800080;
uint64_t __attribute__((aligned(8))) mmx_10w = 0x1010101010101010;
uint64_t __attribute__((aligned(8))) mmx_00ffw = 0x00ff00ff00ff00ff;
uint64_t __attribute__((aligned(8))) mmx_Y_coeff = 0x253f253f253f253f;

/* hope these constant values are cache line aligned */
uint64_t __attribute__((aligned(8))) mmx_U_green = 0xf37df37df37df37d;
uint64_t __attribute__((aligned(8))) mmx_U_blue = 0x4093409340934093;
uint64_t __attribute__((aligned(8))) mmx_V_red = 0x3312331233123312;
uint64_t __attribute__((aligned(8))) mmx_V_green = 0xe5fce5fce5fce5fc;

/* hope these constant values are cache line aligned */
uint64_t __attribute__((aligned(8))) mmx_redmask = 0xf8f8f8f8f8f8f8f8;
uint64_t __attribute__((aligned(8))) mmx_grnmask = 0xfcfcfcfcfcfcfcfc;

#define YUV2RGB \
		     /* Do the multiply part of the conversion for even and odd pixels,
			register usage:
			mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
			mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
			mm6 -> Y even, mm7 -> Y odd */\
		     /* convert the chroma part */\
		     "punpcklbw %%mm4, %%mm0;" /* scatter 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
		     "punpcklbw %%mm4, %%mm1;" /* scatter 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
\
		     "psubsw mmx_80w, %%mm0;" /* Cb -= 128 */ \
		     "psubsw mmx_80w, %%mm1;" /* Cr -= 128 */ \
\
		     "psllw $3, %%mm0;" /* Promote precision */ \
		     "psllw $3, %%mm1;" /* Promote precision */ \
\
		     "movq %%mm0, %%mm2;" /* Copy 4 Cb 00 u3 00 u2 00 u1 00 u0 */ \
		     "movq %%mm1, %%mm3;" /* Copy 4 Cr 00 v3 00 v2 00 v1 00 v0 */ \
\
		     "pmulhw mmx_U_green, %%mm2;" /* Mul Cb with green coeff -> Cb green */ \
		     "pmulhw mmx_V_green, %%mm3;" /* Mul Cr with green coeff -> Cr green */ \
\
		     "pmulhw mmx_U_blue, %%mm0;" /* Mul Cb -> Cblue 00 b3 00 b2 00 b1 00 b0 */\
		     "pmulhw mmx_V_red, %%mm1;" /* Mul Cr -> Cred 00 r3 00 r2 00 r1 00 r0 */\
\
		     "paddsw %%mm3, %%mm2;" /* Cb green + Cr green -> Cgreen */\
\
		     /* convert the luma part */\
		     "psubusb mmx_10w, %%mm6;" /* Y -= 16 */\
\
		     "movq %%mm6, %%mm7;" /* Copy 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */\
		     "pand mmx_00ffw, %%mm6;" /* get Y even 00 Y6 00 Y4 00 Y2 00 Y0 */\
\
		     "psrlw $8, %%mm7;" /* get Y odd 00 Y7 00 Y5 00 Y3 00 Y1 */\
\
		     "psllw $3, %%mm6;" /* Promote precision */\
		     "psllw $3, %%mm7;" /* Promote precision */\
\
		     "pmulhw mmx_Y_coeff, %%mm6;" /* Mul 4 Y even 00 y6 00 y4 00 y2 00 y0 */\
		     "pmulhw mmx_Y_coeff, %%mm7;" /* Mul 4 Y odd 00 y7 00 y5 00 y3 00 y1 */\
\
		     /* Do the addition part of the conversion for even and odd pixels,
			register usage:
			mm0 -> Cblue, mm1 -> Cred, mm2 -> Cgreen even pixels,
			mm3 -> Cblue, mm4 -> Cred, mm5 -> Cgreen odd pixels,
			mm6 -> Y even, mm7 -> Y odd */\
		     "movq %%mm0, %%mm3;" /* Copy Cblue */\
		     "movq %%mm1, %%mm4;" /* Copy Cred */\
		     "movq %%mm2, %%mm5;" /* Copy Cgreen */\
\
		     "paddsw %%mm6, %%mm0;" /* Y even + Cblue 00 B6 00 B4 00 B2 00 B0 */\
		     "paddsw %%mm7, %%mm3;" /* Y odd + Cblue 00 B7 00 B5 00 B3 00 B1 */\
\
		     "paddsw %%mm6, %%mm1;" /* Y even + Cred 00 R6 00 R4 00 R2 00 R0 */\
		     "paddsw %%mm7, %%mm4;" /* Y odd + Cred 00 R7 00 R5 00 R3 00 R1 */\
\
		     "paddsw %%mm6, %%mm2;" /* Y even + Cgreen 00 G6 00 G4 00 G2 00 G0 */\
		     "paddsw %%mm7, %%mm5;" /* Y odd + Cgreen 00 G7 00 G5 00 G3 00 G1 */\
\
		     /* Limit RGB even to 0..255 */\
		     "packuswb %%mm0, %%mm0;" /* B6 B4 B2 B0  B6 B4 B2 B0 */\
		     "packuswb %%mm1, %%mm1;" /* R6 R4 R2 R0  R6 R4 R2 R0 */\
		     "packuswb %%mm2, %%mm2;" /* G6 G4 G2 G0  G6 G4 G2 G0 */\
\
		     /* Limit RGB odd to 0..255 */\
		     "packuswb %%mm3, %%mm3;" /* B7 B5 B3 B1  B7 B5 B3 B1 */\
		     "packuswb %%mm4, %%mm4;" /* R7 R5 R3 R1  R7 R5 R3 R1 */\
		     "packuswb %%mm5, %%mm5;" /* G7 G5 G3 G1  G7 G5 G3 G1 */\
\
		     /* Interleave RGB even and odd */\
		     "punpcklbw %%mm3, %%mm0;" /* B7 B6 B5 B4 B3 B2 B1 B0 */\
		     "punpcklbw %%mm4, %%mm1;" /* R7 R6 R5 R4 R3 R2 R1 R0 */\
		     "punpcklbw %%mm5, %%mm2;" /* G7 G6 G5 G4 G3 G2 G1 G0 */\


static void yuv420_rgb16_mmx (uint8_t * image, uint8_t * py,
			      uint8_t * pu, uint8_t * pv,
			      int h_size, int v_size,
			      int rgb_stride, int y_stride, int uv_stride)
{
    int even = 1;
    int x, y;

    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );

    for (y = v_size; --y >= 0; ) {
	uint8_t *_image = image;
	uint8_t *_py = py;
	uint8_t *_pu = pu;
	uint8_t *_pv = pv;

	/* load data for start of next scan line */
	__asm__ __volatile__ (
		 "movd (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
		 "movd (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
		 "movq (%0), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */

		 : : "r" (_py), "r" (_pu), "r" (_pv));

	for (x = h_size >> 3; --x >= 0; ) {
	    /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
	       pixels in each iteration */

	    __asm__ __volatile__ (
/* no speed diference on my p3@500 with prefetch,
 * if it is faster for anyone with -benchmark then tell me
			PREFETCH" 64(%0) \n\t"
			PREFETCH" 64(%1) \n\t"
			PREFETCH" 64(%2) \n\t"
*/
YUV2RGB

		     /* mask unneeded bits off */
		     "pand mmx_redmask, %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
		     "pand mmx_grnmask, %%mm2;" /* g7g6g5g4 g3g2_0_0 g7g6g5g4 g3g2_0_0 */
		     "pand mmx_redmask, %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

		     "psrlw $3,%%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
		     "pxor %%mm4, %%mm4;" /* zero mm4 */

		     "movq %%mm0, %%mm5;" /* Copy B7-B0 */
		     "movq %%mm2, %%mm7;" /* Copy G7-G0 */

		     /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
		     "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
		     "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */

		     "psllw $3, %%mm2;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
		     "por %%mm2, %%mm0;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */

		     "movq 8 (%0), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
		     MOVNTQ " %%mm0, (%3);" /* store pixel 0-3 */

		     /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
		     "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3g2_0_0 */
		     "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */

		     "psllw $3, %%mm7;" /* 0_0_0_0 0_g7g6g5 g4g3g2_0 0_0_0_0 */
		     "movd 4 (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */

		     "por %%mm7, %%mm5;" /* r7r6r5r4 r3g7g6g5 g4g3g2b7 b6b5b4b3 */
		     "movd 4 (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */

		     MOVNTQ " %%mm5, 8 (%3);" /* store pixel 4-7 */
		     : : "r" (_py), "r" (_pu), "r" (_pv), "r" (_image));

	    _py += 8;
	    _pu += 4;
	    _pv += 4;
	    _image += 16;
	}

	if (!even) {
	    pu += uv_stride;
	    pv += uv_stride;
	}

	py += y_stride;
	image += rgb_stride;

	even = (!even);
    }

    __asm__ __volatile__ (EMMS);
}

static void yuv420_rgb15_mmx (uint8_t * image, uint8_t * py,
			      uint8_t * pu, uint8_t * pv,
			      int h_size, int v_size,
			      int rgb_stride, int y_stride, int uv_stride)
{
    int even = 1;
    int x, y;

    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );

    for (y = v_size; --y >= 0; ) {
	uint8_t *_image = image;
	uint8_t *_py = py;
	uint8_t *_pu = pu;
	uint8_t *_pv = pv;

	/* load data for start of next scan line */
	__asm__ __volatile__ (
		 "movd (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
		 "movd (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
		 "movq (%0), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */

		 : : "r" (_py), "r" (_pu), "r" (_pv));

	for (x = h_size >> 3; --x >= 0; ) {
	    /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
	       pixels in each iteration */

	    __asm__ __volatile__ (
YUV2RGB

		     /* mask unneeded bits off */
		     "pand mmx_redmask, %%mm0;" /* b7b6b5b4 b3_0_0_0 b7b6b5b4 b3_0_0_0 */
		     "pand mmx_redmask, %%mm2;" /* g7g6g5g4 g3_0_0_0 g7g6g5g4 g3_0_0_0 */
		     "pand mmx_redmask, %%mm1;" /* r7r6r5r4 r3_0_0_0 r7r6r5r4 r3_0_0_0 */

		     "psrlw $3,%%mm0;" /* 0_0_0_b7 b6b5b4b3 0_0_0_b7 b6b5b4b3 */
		     "psrlw $1,%%mm1;"            /* 0_r7r6r5  r4r3_0_0 0_r7r6r5 r4r3_0_0 */
		     "pxor %%mm4, %%mm4;" /* zero mm4 */

		     "movq %%mm0, %%mm5;" /* Copy B7-B0 */
		     "movq %%mm2, %%mm7;" /* Copy G7-G0 */

		     /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
		     "punpcklbw %%mm4, %%mm2;" /* 0_0_0_0 0_0_0_0 g7g6g5g4 g3_0_0_0 */
		     "punpcklbw %%mm1, %%mm0;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */

		     "psllw $2, %%mm2;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
		     "por %%mm2, %%mm0;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */

		     "movq 8 (%0), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */
		     MOVNTQ " %%mm0, (%3);" /* store pixel 0-3 */

		     /* convert rgb24 plane to rgb16 pack for pixel 0-3 */
		     "punpckhbw %%mm4, %%mm7;" /* 0_0_0_0 0_0_0_0 0_g7g6g5 g4g3_0_0 */
		     "punpckhbw %%mm1, %%mm5;" /* r7r6r5r4 r3_0_0_0 0_0_0_b7 b6b5b4b3 */

		     "psllw $2, %%mm7;" /* 0_0_0_0 0_0_g7g6 g5g4g3_0 0_0_0_0 */
		     "movd 4 (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */

		     "por %%mm7, %%mm5;" /* 0_r7r6r5 r4r3g7g6 g5g4g3b7 b6b5b4b3 */
		     "movd 4 (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */

		     MOVNTQ " %%mm5, 8 (%3);" /* store pixel 4-7 */
		     : : "r" (_py), "r" (_pu), "r" (_pv), "r" (_image));

	    _py += 8;
	    _pu += 4;
	    _pv += 4;
	    _image += 16;
	}

	if (!even) {
	    pu += uv_stride;
	    pv += uv_stride;
	}

	py += y_stride;
	image += rgb_stride;

	even = (!even);
    }

    __asm__ __volatile__ (EMMS);
}

static void yuv420_rgb24_mmx (uint8_t * image, uint8_t * py,
			      uint8_t * pu, uint8_t * pv,
			      int h_size, int v_size,
			      int rgb_stride, int y_stride, int uv_stride)
{
    int even = 1;
    int x, y;

    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );

    for (y = v_size; --y >= 0; ) {
	uint8_t *_image = image;
	uint8_t *_py = py;
	uint8_t *_pu = pu;
	uint8_t *_pv = pv;

	/* load data for start of next scan line */
	__asm__ __volatile__ (
		 "movd (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
		 "movd (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
		 "movq (%0), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */

		 : : "r" (_py), "r" (_pu), "r" (_pv));

	for (x = h_size >> 3; --x >= 0; ) {
	    /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
	       pixels in each iteration */

	    __asm__ __volatile__ (
YUV2RGB

	/* mm0=B, %%mm2=G, %%mm1=R */
			"pxor %%mm4, %%mm4		\n\t"
			"movq %%mm0, %%mm5		\n\t" /* B */
			"movq %%mm1, %%mm6		\n\t" /* R */
			"punpcklbw %%mm2, %%mm0		\n\t" /* GBGBGBGB 0 */
			"punpcklbw %%mm4, %%mm1		\n\t" /* 0R0R0R0R 0 */
			"punpckhbw %%mm2, %%mm5		\n\t" /* GBGBGBGB 2 */
			"punpckhbw %%mm4, %%mm6		\n\t" /* 0R0R0R0R 2 */
			"movq %%mm0, %%mm7		\n\t" /* GBGBGBGB 0 */
			"movq %%mm5, %%mm3		\n\t" /* GBGBGBGB 2 */
			"punpcklwd %%mm1, %%mm7		\n\t" /* 0RGB0RGB 0 */
			"punpckhwd %%mm1, %%mm0		\n\t" /* 0RGB0RGB 1 */
			"punpcklwd %%mm6, %%mm5		\n\t" /* 0RGB0RGB 2 */
			"punpckhwd %%mm6, %%mm3		\n\t" /* 0RGB0RGB 3 */

			"movq %%mm7, %%mm2		\n\t" /* 0RGB0RGB 0 */
			"movq %%mm0, %%mm6		\n\t" /* 0RGB0RGB 1 */
			"movq %%mm5, %%mm1		\n\t" /* 0RGB0RGB 2 */
			"movq %%mm3, %%mm4		\n\t" /* 0RGB0RGB 3 */

			"psllq $40, %%mm7		\n\t" /* RGB00000 0 */
			"psllq $40, %%mm0		\n\t" /* RGB00000 1 */
			"psllq $40, %%mm5		\n\t" /* RGB00000 2 */
			"psllq $40, %%mm3		\n\t" /* RGB00000 3 */

			"punpckhdq %%mm2, %%mm7		\n\t" /* 0RGBRGB0 0 */
			"punpckhdq %%mm6, %%mm0		\n\t" /* 0RGBRGB0 1 */
			"punpckhdq %%mm1, %%mm5		\n\t" /* 0RGBRGB0 2 */
			"punpckhdq %%mm4, %%mm3		\n\t" /* 0RGBRGB0 3 */

			"psrlq $8, %%mm7		\n\t" /* 00RGBRGB 0 */
			"movq %%mm0, %%mm6		\n\t" /* 0RGBRGB0 1 */
			"psllq $40, %%mm0		\n\t" /* GB000000 1 */
			"por %%mm0, %%mm7		\n\t" /* GBRGBRGB 0 */
			MOVNTQ" %%mm7, (%3)		\n\t"

			"movd 4 (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */

			"psrlq $24, %%mm6		\n\t" /* 0000RGBR 1 */
			"movq %%mm5, %%mm1		\n\t" /* 0RGBRGB0 2 */
			"psllq $24, %%mm5		\n\t" /* BRGB0000 2 */
			"por %%mm5, %%mm6		\n\t" /* BRGBRGBR 1 */
			MOVNTQ" %%mm6, 8(%3)		\n\t"

			"movq 8 (%0), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */

			"psrlq $40, %%mm1		\n\t" /* 000000RG 2 */
			"psllq $8, %%mm3		\n\t" /* RGBRGB00 3 */
			"por %%mm3, %%mm1		\n\t" /* RGBRGBRG 2 */
			MOVNTQ" %%mm1, 16(%3)		\n\t"

			"movd 4 (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */
			"pxor %%mm4, %%mm4		\n\t"


		     : : "r" (_py), "r" (_pu), "r" (_pv), "r" (_image));

	    _py += 8;
	    _pu += 4;
	    _pv += 4;
	    _image += 24;
	}

	if (!even) {
	    pu += uv_stride;
	    pv += uv_stride;
	}

	py += y_stride;
	image += rgb_stride;

	even = (!even);
    }

    __asm__ __volatile__ (EMMS);
}


static void yuv420_argb32_mmx (uint8_t * image, uint8_t * py,
			       uint8_t * pu, uint8_t * pv,
			       int h_size, int v_size,
			       int rgb_stride, int y_stride, int uv_stride)
{
    int even = 1;
    int x, y;

    __asm__ __volatile__ ("pxor %mm4, %mm4;" /* zero mm4 */ );

    for (y = v_size; --y >= 0; ) {
	uint8_t *_image = image;
	uint8_t *_py = py;
	uint8_t *_pu = pu;
	uint8_t *_pv = pv;

	/* load data for start of next scan line */
	__asm__ __volatile__ 
	    (
	     "movd (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */ 
	     "movd (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */ 
	     "movq (%0), %%mm6;" /* Load 8  Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */ 
	     : : "r" (_py), "r" (_pu), "r" (_pv)
	     );

	for (x = h_size >> 3; --x >= 0; ) {
	    /* this mmx assembly code deals with SINGLE scan line at a time, it convert 8
	       pixels in each iteration */
	    __asm__ __volatile__ (
YUV2RGB
		     /* convert RGB plane to RGB packed format,
			mm0 -> B, mm1 -> R, mm2 -> G, mm3 -> 0,
			mm4 -> GB, mm5 -> AR pixel 4-7,
			mm6 -> GB, mm7 -> AR pixel 0-3 */
		     "pxor %%mm3, %%mm3;" /* zero mm3 */

		     "movq %%mm0, %%mm6;" /* B7 B6 B5 B4 B3 B2 B1 B0 */
		     "movq %%mm1, %%mm7;" /* R7 R6 R5 R4 R3 R2 R1 R0 */

		     "movq %%mm0, %%mm4;" /* B7 B6 B5 B4 B3 B2 B1 B0 */
		     "movq %%mm1, %%mm5;" /* R7 R6 R5 R4 R3 R2 R1 R0 */

		     "punpcklbw %%mm2, %%mm6;" /* G3 B3 G2 B2 G1 B1 G0 B0 */
		     "punpcklbw %%mm3, %%mm7;" /* 00 R3 00 R2 00 R1 00 R0 */

		     "punpcklwd %%mm7, %%mm6;" /* 00 R1 B1 G1 00 R0 B0 G0 */
		     MOVNTQ " %%mm6, (%3);" /* Store ARGB1 ARGB0 */

		     "movq %%mm0, %%mm6;" /* B7 B6 B5 B4 B3 B2 B1 B0 */
		     "punpcklbw %%mm2, %%mm6;" /* G3 B3 G2 B2 G1 B1 G0 B0 */

		     "punpckhwd %%mm7, %%mm6;" /* 00 R3 G3 B3 00 R2 B3 G2 */
		     MOVNTQ " %%mm6, 8 (%3);" /* Store ARGB3 ARGB2 */

		     "punpckhbw %%mm2, %%mm4;" /* G7 B7 G6 B6 G5 B5 G4 B4 */
		     "punpckhbw %%mm3, %%mm5;" /* 00 R7 00 R6 00 R5 00 R4 */

		     "punpcklwd %%mm5, %%mm4;" /* 00 R5 B5 G5 00 R4 B4 G4 */
		     MOVNTQ " %%mm4, 16 (%3);" /* Store ARGB5 ARGB4 */

		     "movq %%mm0, %%mm4;" /* B7 B6 B5 B4 B3 B2 B1 B0 */
		     "punpckhbw %%mm2, %%mm4;" /* G7 B7 G6 B6 G5 B5 G4 B4 */

		     "punpckhwd %%mm5, %%mm4;" /* 00 R7 G7 B7 00 R6 B6 G6 */
		     MOVNTQ " %%mm4, 24 (%3);" /* Store ARGB7 ARGB6 */

		     "movd 4 (%1), %%mm0;" /* Load 4 Cb 00 00 00 00 u3 u2 u1 u0 */
		     "movd 4 (%2), %%mm1;" /* Load 4 Cr 00 00 00 00 v3 v2 v1 v0 */

		     "pxor %%mm4, %%mm4;" /* zero mm4 */
		     "movq 8 (%0), %%mm6;" /* Load 8 Y Y7 Y6 Y5 Y4 Y3 Y2 Y1 Y0 */

		     : : "r" (_py), "r" (_pu), "r" (_pv), "r" (_image));

	    _py += 8;
	    _pu += 4;
	    _pv += 4;
	    _image += 32;
	}

	if (!even) {
	    pu += uv_stride;
	    pv += uv_stride;
	}

	py += y_stride;
	image += rgb_stride;

	even = (!even);
    }

    __asm__ __volatile__ (EMMS);
}

yuv2rgb_fun yuv2rgb_init_mmx (int bpp, int mode)
{
    if (bpp == 15 && mode == MODE_RGB) return yuv420_rgb15_mmx;
    if (bpp == 16 && mode == MODE_RGB) return yuv420_rgb16_mmx;
    if (bpp == 24 && mode == MODE_RGB) return yuv420_rgb24_mmx;
    if (bpp == 32 && mode == MODE_RGB) return yuv420_argb32_mmx;
    return NULL; // Fallback to C.
}