Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 30c6fefd

History | View | Annotate | Download (74.3 KB)

1
/*
2
 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3
 *
4
 * This file is part of FFmpeg.
5
 *
6
 * FFmpeg is free software; you can redistribute it and/or
7
 * modify it under the terms of the GNU Lesser General Public
8
 * License as published by the Free Software Foundation; either
9
 * version 2.1 of the License, or (at your option) any later version.
10
 *
11
 * FFmpeg is distributed in the hope that it will be useful,
12
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
 * Lesser General Public License for more details.
15
 *
16
 * You should have received a copy of the GNU Lesser General Public
17
 * License along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
 */
20

    
21
/*
22
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
23
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
24
  {BGR,RGB}{1,4,8,15,16} support dithering
25

26
  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
27
  YV12 -> {BGR,RGB}{1,4,8,12,15,16,24,32}
28
  x -> x
29
  YUV9 -> YV12
30
  YUV9/YV12 -> Y800
31
  Y800 -> YUV9/YV12
32
  BGR24 -> BGR32 & RGB24 -> RGB32
33
  BGR32 -> BGR24 & RGB32 -> RGB24
34
  BGR15 -> BGR16
35
*/
36

    
37
/*
38
tested special converters (most are tested actually, but I did not write it down ...)
39
 YV12 -> BGR12/BGR16
40
 YV12 -> YV12
41
 BGR15 -> BGR16
42
 BGR16 -> BGR16
43
 YVU9 -> YV12
44

45
untested special converters
46
  YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
47
  YV12/I420 -> YV12/I420
48
  YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
49
  BGR24 -> BGR32 & RGB24 -> RGB32
50
  BGR32 -> BGR24 & RGB32 -> RGB24
51
  BGR24 -> YV12
52
*/
53

    
54
#include <inttypes.h>
55
#include <string.h>
56
#include <math.h>
57
#include <stdio.h>
58
#include "config.h"
59
#include <assert.h>
60
#include "swscale.h"
61
#include "swscale_internal.h"
62
#include "rgb2rgb.h"
63
#include "libavutil/intreadwrite.h"
64
#include "libavutil/x86_cpu.h"
65
#include "libavutil/cpu.h"
66
#include "libavutil/avutil.h"
67
#include "libavutil/mathematics.h"
68
#include "libavutil/bswap.h"
69
#include "libavutil/pixdesc.h"
70

    
71
#undef MOVNTQ
72
#undef PAVGB
73

    
74
//#undef HAVE_MMX2
75
//#define HAVE_AMD3DNOW
76
//#undef HAVE_MMX
77
//#undef ARCH_X86
78
#define DITHER1XBPP
79

    
80
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
81

    
82
#define isPacked(x)         (       \
83
           (x)==PIX_FMT_PAL8        \
84
        || (x)==PIX_FMT_YUYV422     \
85
        || (x)==PIX_FMT_UYVY422     \
86
        || (x)==PIX_FMT_Y400A       \
87
        || isAnyRGB(x)              \
88
    )
89

    
90
#define RGB2YUV_SHIFT 15
91
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
92
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
93
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
94
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
95
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
96
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
97
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
98
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
99
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
100

    
101
static const double rgb2yuv_table[8][9]={
102
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
103
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
104
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
105
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
106
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
107
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
108
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
109
    {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
110
};
111

    
112
/*
113
NOTES
114
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
115

116
TODO
117
more intelligent misalignment avoidance for the horizontal scaler
118
write special vertical cubic upscale version
119
optimize C code (YV12 / minmax)
120
add support for packed pixel YUV input & output
121
add support for Y8 output
122
optimize BGR24 & BGR32
123
add BGR4 output support
124
write special BGR->BGR scaler
125
*/
126

    
127
#if ARCH_X86
128
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
129
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
130
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
131
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
132
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
133
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
134
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
135
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
136

    
137
const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
138
        0x0103010301030103LL,
139
        0x0200020002000200LL,};
140

    
141
const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
142
        0x0602060206020602LL,
143
        0x0004000400040004LL,};
144

    
145
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
146
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
147
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
148
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
149
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
150
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
151

    
152
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
153
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
154
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
155

    
156
#ifdef FAST_BGR2YV12
157
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
158
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
159
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
160
#else
161
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
162
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
163
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
164
#endif /* FAST_BGR2YV12 */
165
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
166
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
167
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
168

    
169
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
170
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
171
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
172
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
173
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
174

    
175
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
176
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
177
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
178
};
179

    
180
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
181

    
182
#endif /* ARCH_X86 */
183

    
184
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
185
{  1,   3,   1,   3,   1,   3,   1,   3, },
186
{  2,   0,   2,   0,   2,   0,   2,   0, },
187
};
188

    
189
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
190
{  6,   2,   6,   2,   6,   2,   6,   2, },
191
{  0,   4,   0,   4,   0,   4,   0,   4, },
192
};
193

    
194
DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
195
{  8,   4,  11,   7,   8,   4,  11,   7, },
196
{  2,  14,   1,  13,   2,  14,   1,  13, },
197
{ 10,   6,   9,   5,  10,   6,   9,   5, },
198
{  0,  12,   3,  15,   0,  12,   3,  15, },
199
};
200

    
201
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
202
{ 17,   9,  23,  15,  16,   8,  22,  14, },
203
{  5,  29,   3,  27,   4,  28,   2,  26, },
204
{ 21,  13,  19,  11,  20,  12,  18,  10, },
205
{  0,  24,   6,  30,   1,  25,   7,  31, },
206
{ 16,   8,  22,  14,  17,   9,  23,  15, },
207
{  4,  28,   2,  26,   5,  29,   3,  27, },
208
{ 20,  12,  18,  10,  21,  13,  19,  11, },
209
{  1,  25,   7,  31,   0,  24,   6,  30, },
210
};
211

    
212
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
213
{  0,  55,  14,  68,   3,  58,  17,  72, },
214
{ 37,  18,  50,  32,  40,  22,  54,  35, },
215
{  9,  64,   5,  59,  13,  67,   8,  63, },
216
{ 46,  27,  41,  23,  49,  31,  44,  26, },
217
{  2,  57,  16,  71,   1,  56,  15,  70, },
218
{ 39,  21,  52,  34,  38,  19,  51,  33, },
219
{ 11,  66,   7,  62,  10,  65,   6,  60, },
220
{ 48,  30,  43,  25,  47,  29,  42,  24, },
221
};
222

    
223
#if 1
224
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
225
{117,  62, 158, 103, 113,  58, 155, 100, },
226
{ 34, 199,  21, 186,  31, 196,  17, 182, },
227
{144,  89, 131,  76, 141,  86, 127,  72, },
228
{  0, 165,  41, 206,  10, 175,  52, 217, },
229
{110,  55, 151,  96, 120,  65, 162, 107, },
230
{ 28, 193,  14, 179,  38, 203,  24, 189, },
231
{138,  83, 124,  69, 148,  93, 134,  79, },
232
{  7, 172,  48, 213,   3, 168,  45, 210, },
233
};
234
#elif 1
235
// tries to correct a gamma of 1.5
236
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
237
{  0, 143,  18, 200,   2, 156,  25, 215, },
238
{ 78,  28, 125,  64,  89,  36, 138,  74, },
239
{ 10, 180,   3, 161,  16, 195,   8, 175, },
240
{109,  51,  93,  38, 121,  60, 105,  47, },
241
{  1, 152,  23, 210,   0, 147,  20, 205, },
242
{ 85,  33, 134,  71,  81,  30, 130,  67, },
243
{ 14, 190,   6, 171,  12, 185,   5, 166, },
244
{117,  57, 101,  44, 113,  54,  97,  41, },
245
};
246
#elif 1
247
// tries to correct a gamma of 2.0
248
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
249
{  0, 124,   8, 193,   0, 140,  12, 213, },
250
{ 55,  14, 104,  42,  66,  19, 119,  52, },
251
{  3, 168,   1, 145,   6, 187,   3, 162, },
252
{ 86,  31,  70,  21,  99,  39,  82,  28, },
253
{  0, 134,  11, 206,   0, 129,   9, 200, },
254
{ 62,  17, 114,  48,  58,  16, 109,  45, },
255
{  5, 181,   2, 157,   4, 175,   1, 151, },
256
{ 95,  36,  78,  26,  90,  34,  74,  24, },
257
};
258
#else
259
// tries to correct a gamma of 2.5
260
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
261
{  0, 107,   3, 187,   0, 125,   6, 212, },
262
{ 39,   7,  86,  28,  49,  11, 102,  36, },
263
{  1, 158,   0, 131,   3, 180,   1, 151, },
264
{ 68,  19,  52,  12,  81,  25,  64,  17, },
265
{  0, 119,   5, 203,   0, 113,   4, 195, },
266
{ 45,   9,  96,  33,  42,   8,  91,  30, },
267
{  2, 172,   1, 144,   2, 165,   0, 137, },
268
{ 77,  23,  60,  15,  72,  21,  56,  14, },
269
};
270
#endif
271

    
272
static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
273
                                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
274
                                                    const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
275
                                                    int dstW, int chrDstW, int big_endian)
276
{
277
    //FIXME Optimize (just quickly written not optimized..)
278
    int i;
279

    
280
    for (i = 0; i < dstW; i++) {
281
        int val = 1 << 10;
282
        int j;
283

    
284
        for (j = 0; j < lumFilterSize; j++)
285
            val += lumSrc[j][i] * lumFilter[j];
286

    
287
        if (big_endian) {
288
            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
289
        } else {
290
            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
291
        }
292
    }
293

    
294
    if (uDest) {
295
        for (i = 0; i < chrDstW; i++) {
296
            int u = 1 << 10;
297
            int v = 1 << 10;
298
            int j;
299

    
300
            for (j = 0; j < chrFilterSize; j++) {
301
                u += chrSrc[j][i       ] * chrFilter[j];
302
                v += chrSrc[j][i + VOFW] * chrFilter[j];
303
            }
304

    
305
            if (big_endian) {
306
                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
307
                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
308
            } else {
309
                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
310
                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
311
            }
312
        }
313
    }
314

    
315
    if (CONFIG_SWSCALE_ALPHA && aDest) {
316
        for (i = 0; i < dstW; i++) {
317
            int val = 1 << 10;
318
            int j;
319

    
320
            for (j = 0; j < lumFilterSize; j++)
321
                val += alpSrc[j][i] * lumFilter[j];
322

    
323
            if (big_endian) {
324
                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
325
            } else {
326
                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
327
            }
328
        }
329
    }
330
}
331

    
332
static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
333
                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
334
                                 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
335
                                 enum PixelFormat dstFormat)
336
{
337
    if (isBE(dstFormat)) {
338
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
339
                               chrFilter, chrSrc, chrFilterSize,
340
                               alpSrc,
341
                               dest, uDest, vDest, aDest,
342
                               dstW, chrDstW, 1);
343
    } else {
344
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
345
                               chrFilter, chrSrc, chrFilterSize,
346
                               alpSrc,
347
                               dest, uDest, vDest, aDest,
348
                               dstW, chrDstW, 0);
349
    }
350
}
351

    
352
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
353
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
354
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
355
{
356
    //FIXME Optimize (just quickly written not optimized..)
357
    int i;
358
    for (i=0; i<dstW; i++) {
359
        int val=1<<18;
360
        int j;
361
        for (j=0; j<lumFilterSize; j++)
362
            val += lumSrc[j][i] * lumFilter[j];
363

    
364
        dest[i]= av_clip_uint8(val>>19);
365
    }
366

    
367
    if (uDest)
368
        for (i=0; i<chrDstW; i++) {
369
            int u=1<<18;
370
            int v=1<<18;
371
            int j;
372
            for (j=0; j<chrFilterSize; j++) {
373
                u += chrSrc[j][i] * chrFilter[j];
374
                v += chrSrc[j][i + VOFW] * chrFilter[j];
375
            }
376

    
377
            uDest[i]= av_clip_uint8(u>>19);
378
            vDest[i]= av_clip_uint8(v>>19);
379
        }
380

    
381
    if (CONFIG_SWSCALE_ALPHA && aDest)
382
        for (i=0; i<dstW; i++) {
383
            int val=1<<18;
384
            int j;
385
            for (j=0; j<lumFilterSize; j++)
386
                val += alpSrc[j][i] * lumFilter[j];
387

    
388
            aDest[i]= av_clip_uint8(val>>19);
389
        }
390

    
391
}
392

    
393
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
394
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
395
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
396
{
397
    //FIXME Optimize (just quickly written not optimized..)
398
    int i;
399
    for (i=0; i<dstW; i++) {
400
        int val=1<<18;
401
        int j;
402
        for (j=0; j<lumFilterSize; j++)
403
            val += lumSrc[j][i] * lumFilter[j];
404

    
405
        dest[i]= av_clip_uint8(val>>19);
406
    }
407

    
408
    if (!uDest)
409
        return;
410

    
411
    if (dstFormat == PIX_FMT_NV12)
412
        for (i=0; i<chrDstW; i++) {
413
            int u=1<<18;
414
            int v=1<<18;
415
            int j;
416
            for (j=0; j<chrFilterSize; j++) {
417
                u += chrSrc[j][i] * chrFilter[j];
418
                v += chrSrc[j][i + VOFW] * chrFilter[j];
419
            }
420

    
421
            uDest[2*i]= av_clip_uint8(u>>19);
422
            uDest[2*i+1]= av_clip_uint8(v>>19);
423
        }
424
    else
425
        for (i=0; i<chrDstW; i++) {
426
            int u=1<<18;
427
            int v=1<<18;
428
            int j;
429
            for (j=0; j<chrFilterSize; j++) {
430
                u += chrSrc[j][i] * chrFilter[j];
431
                v += chrSrc[j][i + VOFW] * chrFilter[j];
432
            }
433

    
434
            uDest[2*i]= av_clip_uint8(v>>19);
435
            uDest[2*i+1]= av_clip_uint8(u>>19);
436
        }
437
}
438

    
439
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
440
    for (i=0; i<(dstW>>1); i++) {\
441
        int j;\
442
        int Y1 = 1<<18;\
443
        int Y2 = 1<<18;\
444
        int U  = 1<<18;\
445
        int V  = 1<<18;\
446
        int av_unused A1, A2;\
447
        type av_unused *r, *b, *g;\
448
        const int i2= 2*i;\
449
        \
450
        for (j=0; j<lumFilterSize; j++) {\
451
            Y1 += lumSrc[j][i2] * lumFilter[j];\
452
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
453
        }\
454
        for (j=0; j<chrFilterSize; j++) {\
455
            U += chrSrc[j][i] * chrFilter[j];\
456
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
457
        }\
458
        Y1>>=19;\
459
        Y2>>=19;\
460
        U >>=19;\
461
        V >>=19;\
462
        if (alpha) {\
463
            A1 = 1<<18;\
464
            A2 = 1<<18;\
465
            for (j=0; j<lumFilterSize; j++) {\
466
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
467
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
468
            }\
469
            A1>>=19;\
470
            A2>>=19;\
471
        }
472

    
473
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
474
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
475
        if ((Y1|Y2|U|V)&256) {\
476
            if (Y1>255)   Y1=255; \
477
            else if (Y1<0)Y1=0;   \
478
            if (Y2>255)   Y2=255; \
479
            else if (Y2<0)Y2=0;   \
480
            if (U>255)    U=255;  \
481
            else if (U<0) U=0;    \
482
            if (V>255)    V=255;  \
483
            else if (V<0) V=0;    \
484
        }\
485
        if (alpha && ((A1|A2)&256)) {\
486
            A1=av_clip_uint8(A1);\
487
            A2=av_clip_uint8(A2);\
488
        }
489

    
490
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
491
    for (i=0; i<dstW; i++) {\
492
        int j;\
493
        int Y = 0;\
494
        int U = -128<<19;\
495
        int V = -128<<19;\
496
        int av_unused A;\
497
        int R,G,B;\
498
        \
499
        for (j=0; j<lumFilterSize; j++) {\
500
            Y += lumSrc[j][i     ] * lumFilter[j];\
501
        }\
502
        for (j=0; j<chrFilterSize; j++) {\
503
            U += chrSrc[j][i     ] * chrFilter[j];\
504
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
505
        }\
506
        Y >>=10;\
507
        U >>=10;\
508
        V >>=10;\
509
        if (alpha) {\
510
            A = rnd;\
511
            for (j=0; j<lumFilterSize; j++)\
512
                A += alpSrc[j][i     ] * lumFilter[j];\
513
            A >>=19;\
514
            if (A&256)\
515
                A = av_clip_uint8(A);\
516
        }
517

    
518
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
519
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
520
        Y-= c->yuv2rgb_y_offset;\
521
        Y*= c->yuv2rgb_y_coeff;\
522
        Y+= rnd;\
523
        R= Y + V*c->yuv2rgb_v2r_coeff;\
524
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
525
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
526
        if ((R|G|B)&(0xC0000000)) {\
527
            if (R>=(256<<22))   R=(256<<22)-1; \
528
            else if (R<0)R=0;   \
529
            if (G>=(256<<22))   G=(256<<22)-1; \
530
            else if (G<0)G=0;   \
531
            if (B>=(256<<22))   B=(256<<22)-1; \
532
            else if (B<0)B=0;   \
533
        }
534

    
535
#define YSCALE_YUV_2_GRAY16_C \
536
    for (i=0; i<(dstW>>1); i++) {\
537
        int j;\
538
        int Y1 = 1<<18;\
539
        int Y2 = 1<<18;\
540
        int U  = 1<<18;\
541
        int V  = 1<<18;\
542
        \
543
        const int i2= 2*i;\
544
        \
545
        for (j=0; j<lumFilterSize; j++) {\
546
            Y1 += lumSrc[j][i2] * lumFilter[j];\
547
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
548
        }\
549
        Y1>>=11;\
550
        Y2>>=11;\
551
        if ((Y1|Y2|U|V)&65536) {\
552
            if (Y1>65535)   Y1=65535; \
553
            else if (Y1<0)Y1=0;   \
554
            if (Y2>65535)   Y2=65535; \
555
            else if (Y2<0)Y2=0;   \
556
        }
557

    
558
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
559
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
560
    r = (type *)c->table_rV[V];   \
561
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
562
    b = (type *)c->table_bU[U];
563

    
564
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
565
    for (i=0; i<(dstW>>1); i++) { \
566
        const int i2= 2*i;       \
567
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
568
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
569
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
570
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
571
        type av_unused *r, *b, *g;                                    \
572
        int av_unused A1, A2;                                         \
573
        if (alpha) {\
574
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
575
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
576
        }
577

    
578
#define YSCALE_YUV_2_GRAY16_2_C   \
579
    for (i=0; i<(dstW>>1); i++) { \
580
        const int i2= 2*i;       \
581
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
582
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
583

    
584
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
585
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
586
    r = (type *)c->table_rV[V];\
587
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
588
    b = (type *)c->table_bU[U];
589

    
590
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
591
    for (i=0; i<(dstW>>1); i++) {\
592
        const int i2= 2*i;\
593
        int Y1= buf0[i2  ]>>7;\
594
        int Y2= buf0[i2+1]>>7;\
595
        int U= (uvbuf1[i     ])>>7;\
596
        int V= (uvbuf1[i+VOFW])>>7;\
597
        type av_unused *r, *b, *g;\
598
        int av_unused A1, A2;\
599
        if (alpha) {\
600
            A1= abuf0[i2  ]>>7;\
601
            A2= abuf0[i2+1]>>7;\
602
        }
603

    
604
#define YSCALE_YUV_2_GRAY16_1_C \
605
    for (i=0; i<(dstW>>1); i++) {\
606
        const int i2= 2*i;\
607
        int Y1= buf0[i2  ]<<1;\
608
        int Y2= buf0[i2+1]<<1;
609

    
610
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
611
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
612
    r = (type *)c->table_rV[V];\
613
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
614
    b = (type *)c->table_bU[U];
615

    
616
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
617
    for (i=0; i<(dstW>>1); i++) {\
618
        const int i2= 2*i;\
619
        int Y1= buf0[i2  ]>>7;\
620
        int Y2= buf0[i2+1]>>7;\
621
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
622
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
623
        type av_unused *r, *b, *g;\
624
        int av_unused A1, A2;\
625
        if (alpha) {\
626
            A1= abuf0[i2  ]>>7;\
627
            A2= abuf0[i2+1]>>7;\
628
        }
629

    
630
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
631
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
632
    r = (type *)c->table_rV[V];\
633
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
634
    b = (type *)c->table_bU[U];
635

    
636
#define YSCALE_YUV_2_MONO2_C \
637
    const uint8_t * const d128=dither_8x8_220[y&7];\
638
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
639
    for (i=0; i<dstW-7; i+=8) {\
640
        int acc;\
641
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
642
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
643
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
644
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
645
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
646
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
647
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
648
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
649
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
650
        dest++;\
651
    }
652

    
653
#define YSCALE_YUV_2_MONOX_C \
654
    const uint8_t * const d128=dither_8x8_220[y&7];\
655
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
656
    int acc=0;\
657
    for (i=0; i<dstW-1; i+=2) {\
658
        int j;\
659
        int Y1=1<<18;\
660
        int Y2=1<<18;\
661
\
662
        for (j=0; j<lumFilterSize; j++) {\
663
            Y1 += lumSrc[j][i] * lumFilter[j];\
664
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
665
        }\
666
        Y1>>=19;\
667
        Y2>>=19;\
668
        if ((Y1|Y2)&256) {\
669
            if (Y1>255)   Y1=255;\
670
            else if (Y1<0)Y1=0;\
671
            if (Y2>255)   Y2=255;\
672
            else if (Y2<0)Y2=0;\
673
        }\
674
        acc+= acc + g[Y1+d128[(i+0)&7]];\
675
        acc+= acc + g[Y2+d128[(i+1)&7]];\
676
        if ((i&7)==6) {\
677
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
678
            dest++;\
679
        }\
680
    }
681

    
682
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
683
    switch(c->dstFormat) {\
684
    case PIX_FMT_RGB48BE:\
685
    case PIX_FMT_RGB48LE:\
686
        func(uint8_t,0)\
687
            ((uint8_t*)dest)[ 0]= r[Y1];\
688
            ((uint8_t*)dest)[ 1]= r[Y1];\
689
            ((uint8_t*)dest)[ 2]= g[Y1];\
690
            ((uint8_t*)dest)[ 3]= g[Y1];\
691
            ((uint8_t*)dest)[ 4]= b[Y1];\
692
            ((uint8_t*)dest)[ 5]= b[Y1];\
693
            ((uint8_t*)dest)[ 6]= r[Y2];\
694
            ((uint8_t*)dest)[ 7]= r[Y2];\
695
            ((uint8_t*)dest)[ 8]= g[Y2];\
696
            ((uint8_t*)dest)[ 9]= g[Y2];\
697
            ((uint8_t*)dest)[10]= b[Y2];\
698
            ((uint8_t*)dest)[11]= b[Y2];\
699
            dest+=12;\
700
        }\
701
        break;\
702
    case PIX_FMT_BGR48BE:\
703
    case PIX_FMT_BGR48LE:\
704
        func(uint8_t,0)\
705
            ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
706
            ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
707
            ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
708
            ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
709
            ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
710
            ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
711
            dest+=12;\
712
        }\
713
        break;\
714
    case PIX_FMT_RGBA:\
715
    case PIX_FMT_BGRA:\
716
        if (CONFIG_SMALL) {\
717
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
718
            func(uint32_t,needAlpha)\
719
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
720
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
721
            }\
722
        } else {\
723
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
724
                func(uint32_t,1)\
725
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
726
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
727
                }\
728
            } else {\
729
                func(uint32_t,0)\
730
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
731
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
732
                }\
733
            }\
734
        }\
735
        break;\
736
    case PIX_FMT_ARGB:\
737
    case PIX_FMT_ABGR:\
738
        if (CONFIG_SMALL) {\
739
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
740
            func(uint32_t,needAlpha)\
741
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
742
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
743
            }\
744
        } else {\
745
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
746
                func(uint32_t,1)\
747
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
748
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
749
                }\
750
            } else {\
751
                func(uint32_t,0)\
752
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
753
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
754
                }\
755
            }\
756
        }                \
757
        break;\
758
    case PIX_FMT_RGB24:\
759
        func(uint8_t,0)\
760
            ((uint8_t*)dest)[0]= r[Y1];\
761
            ((uint8_t*)dest)[1]= g[Y1];\
762
            ((uint8_t*)dest)[2]= b[Y1];\
763
            ((uint8_t*)dest)[3]= r[Y2];\
764
            ((uint8_t*)dest)[4]= g[Y2];\
765
            ((uint8_t*)dest)[5]= b[Y2];\
766
            dest+=6;\
767
        }\
768
        break;\
769
    case PIX_FMT_BGR24:\
770
        func(uint8_t,0)\
771
            ((uint8_t*)dest)[0]= b[Y1];\
772
            ((uint8_t*)dest)[1]= g[Y1];\
773
            ((uint8_t*)dest)[2]= r[Y1];\
774
            ((uint8_t*)dest)[3]= b[Y2];\
775
            ((uint8_t*)dest)[4]= g[Y2];\
776
            ((uint8_t*)dest)[5]= r[Y2];\
777
            dest+=6;\
778
        }\
779
        break;\
780
    case PIX_FMT_RGB565BE:\
781
    case PIX_FMT_RGB565LE:\
782
    case PIX_FMT_BGR565BE:\
783
    case PIX_FMT_BGR565LE:\
784
        {\
785
            const int dr1= dither_2x2_8[y&1    ][0];\
786
            const int dg1= dither_2x2_4[y&1    ][0];\
787
            const int db1= dither_2x2_8[(y&1)^1][0];\
788
            const int dr2= dither_2x2_8[y&1    ][1];\
789
            const int dg2= dither_2x2_4[y&1    ][1];\
790
            const int db2= dither_2x2_8[(y&1)^1][1];\
791
            func(uint16_t,0)\
792
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
793
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
794
            }\
795
        }\
796
        break;\
797
    case PIX_FMT_RGB555BE:\
798
    case PIX_FMT_RGB555LE:\
799
    case PIX_FMT_BGR555BE:\
800
    case PIX_FMT_BGR555LE:\
801
        {\
802
            const int dr1= dither_2x2_8[y&1    ][0];\
803
            const int dg1= dither_2x2_8[y&1    ][1];\
804
            const int db1= dither_2x2_8[(y&1)^1][0];\
805
            const int dr2= dither_2x2_8[y&1    ][1];\
806
            const int dg2= dither_2x2_8[y&1    ][0];\
807
            const int db2= dither_2x2_8[(y&1)^1][1];\
808
            func(uint16_t,0)\
809
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
810
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
811
            }\
812
        }\
813
        break;\
814
    case PIX_FMT_RGB444BE:\
815
    case PIX_FMT_RGB444LE:\
816
    case PIX_FMT_BGR444BE:\
817
    case PIX_FMT_BGR444LE:\
818
        {\
819
            const int dr1= dither_4x4_16[y&3    ][0];\
820
            const int dg1= dither_4x4_16[y&3    ][1];\
821
            const int db1= dither_4x4_16[(y&3)^3][0];\
822
            const int dr2= dither_4x4_16[y&3    ][1];\
823
            const int dg2= dither_4x4_16[y&3    ][0];\
824
            const int db2= dither_4x4_16[(y&3)^3][1];\
825
            func(uint16_t,0)\
826
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
827
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
828
            }\
829
        }\
830
        break;\
831
    case PIX_FMT_RGB8:\
832
    case PIX_FMT_BGR8:\
833
        {\
834
            const uint8_t * const d64= dither_8x8_73[y&7];\
835
            const uint8_t * const d32= dither_8x8_32[y&7];\
836
            func(uint8_t,0)\
837
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
838
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
839
            }\
840
        }\
841
        break;\
842
    case PIX_FMT_RGB4:\
843
    case PIX_FMT_BGR4:\
844
        {\
845
            const uint8_t * const d64= dither_8x8_73 [y&7];\
846
            const uint8_t * const d128=dither_8x8_220[y&7];\
847
            func(uint8_t,0)\
848
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
849
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
850
            }\
851
        }\
852
        break;\
853
    case PIX_FMT_RGB4_BYTE:\
854
    case PIX_FMT_BGR4_BYTE:\
855
        {\
856
            const uint8_t * const d64= dither_8x8_73 [y&7];\
857
            const uint8_t * const d128=dither_8x8_220[y&7];\
858
            func(uint8_t,0)\
859
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
860
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
861
            }\
862
        }\
863
        break;\
864
    case PIX_FMT_MONOBLACK:\
865
    case PIX_FMT_MONOWHITE:\
866
        {\
867
            func_monoblack\
868
        }\
869
        break;\
870
    case PIX_FMT_YUYV422:\
871
        func2\
872
            ((uint8_t*)dest)[2*i2+0]= Y1;\
873
            ((uint8_t*)dest)[2*i2+1]= U;\
874
            ((uint8_t*)dest)[2*i2+2]= Y2;\
875
            ((uint8_t*)dest)[2*i2+3]= V;\
876
        }                \
877
        break;\
878
    case PIX_FMT_UYVY422:\
879
        func2\
880
            ((uint8_t*)dest)[2*i2+0]= U;\
881
            ((uint8_t*)dest)[2*i2+1]= Y1;\
882
            ((uint8_t*)dest)[2*i2+2]= V;\
883
            ((uint8_t*)dest)[2*i2+3]= Y2;\
884
        }                \
885
        break;\
886
    case PIX_FMT_GRAY16BE:\
887
        func_g16\
888
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
889
            ((uint8_t*)dest)[2*i2+1]= Y1;\
890
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
891
            ((uint8_t*)dest)[2*i2+3]= Y2;\
892
        }                \
893
        break;\
894
    case PIX_FMT_GRAY16LE:\
895
        func_g16\
896
            ((uint8_t*)dest)[2*i2+0]= Y1;\
897
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
898
            ((uint8_t*)dest)[2*i2+2]= Y2;\
899
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
900
        }                \
901
        break;\
902
    }
903

    
904
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
905
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
906
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
907
{
908
    int i;
909
    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
910
}
911

    
912
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
913
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
914
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
915
{
916
    int i;
917
    int step= c->dstFormatBpp/8;
918
    int aidx= 3;
919

    
920
    switch(c->dstFormat) {
921
    case PIX_FMT_ARGB:
922
        dest++;
923
        aidx= 0;
924
    case PIX_FMT_RGB24:
925
        aidx--;
926
    case PIX_FMT_RGBA:
927
        if (CONFIG_SMALL) {
928
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
929
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
930
                dest[aidx]= needAlpha ? A : 255;
931
                dest[0]= R>>22;
932
                dest[1]= G>>22;
933
                dest[2]= B>>22;
934
                dest+= step;
935
            }
936
        } else {
937
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
938
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
939
                    dest[aidx]= A;
940
                    dest[0]= R>>22;
941
                    dest[1]= G>>22;
942
                    dest[2]= B>>22;
943
                    dest+= step;
944
                }
945
            } else {
946
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
947
                    dest[aidx]= 255;
948
                    dest[0]= R>>22;
949
                    dest[1]= G>>22;
950
                    dest[2]= B>>22;
951
                    dest+= step;
952
                }
953
            }
954
        }
955
        break;
956
    case PIX_FMT_ABGR:
957
        dest++;
958
        aidx= 0;
959
    case PIX_FMT_BGR24:
960
        aidx--;
961
    case PIX_FMT_BGRA:
962
        if (CONFIG_SMALL) {
963
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
964
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
965
                dest[aidx]= needAlpha ? A : 255;
966
                dest[0]= B>>22;
967
                dest[1]= G>>22;
968
                dest[2]= R>>22;
969
                dest+= step;
970
            }
971
        } else {
972
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
973
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
974
                    dest[aidx]= A;
975
                    dest[0]= B>>22;
976
                    dest[1]= G>>22;
977
                    dest[2]= R>>22;
978
                    dest+= step;
979
                }
980
            } else {
981
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
982
                    dest[aidx]= 255;
983
                    dest[0]= B>>22;
984
                    dest[1]= G>>22;
985
                    dest[2]= R>>22;
986
                    dest+= step;
987
                }
988
            }
989
        }
990
        break;
991
    default:
992
        assert(0);
993
    }
994
}
995

    
996
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
997
{
998
    int i;
999
    uint8_t *ptr = plane + stride*y;
1000
    for (i=0; i<height; i++) {
1001
        memset(ptr, val, width);
1002
        ptr += stride;
1003
    }
1004
}
1005

    
1006
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
1007
                            uint32_t *unused)
1008
{
1009
    int i;
1010
    for (i = 0; i < width; i++) {
1011
        int r = src[i*6+0];
1012
        int g = src[i*6+2];
1013
        int b = src[i*6+4];
1014

    
1015
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1016
    }
1017
}
1018

    
1019
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1020
                             const uint8_t *src1, const uint8_t *src2,
1021
                             long width, uint32_t *unused)
1022
{
1023
    int i;
1024
    assert(src1==src2);
1025
    for (i = 0; i < width; i++) {
1026
        int r = src1[6*i + 0];
1027
        int g = src1[6*i + 2];
1028
        int b = src1[6*i + 4];
1029

    
1030
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1031
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1032
    }
1033
}
1034

    
1035
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1036
                                  const uint8_t *src1, const uint8_t *src2,
1037
                                  long width, uint32_t *unused)
1038
{
1039
    int i;
1040
    assert(src1==src2);
1041
    for (i = 0; i < width; i++) {
1042
        int r= src1[12*i + 0] + src1[12*i + 6];
1043
        int g= src1[12*i + 2] + src1[12*i + 8];
1044
        int b= src1[12*i + 4] + src1[12*i + 10];
1045

    
1046
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1047
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1048
    }
1049
}
1050

    
1051
static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,
1052
                            uint32_t *unused)
1053
{
1054
    int i;
1055
    for (i = 0; i < width; i++) {
1056
        int b = src[i*6+0];
1057
        int g = src[i*6+2];
1058
        int r = src[i*6+4];
1059

    
1060
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1061
    }
1062
}
1063

    
1064
static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1065
                             const uint8_t *src1, const uint8_t *src2,
1066
                             long width, uint32_t *unused)
1067
{
1068
    int i;
1069
    for (i = 0; i < width; i++) {
1070
        int b = src1[6*i + 0];
1071
        int g = src1[6*i + 2];
1072
        int r = src1[6*i + 4];
1073

    
1074
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1075
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1076
    }
1077
}
1078

    
1079
static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1080
                                  const uint8_t *src1, const uint8_t *src2,
1081
                                  long width, uint32_t *unused)
1082
{
1083
    int i;
1084
    for (i = 0; i < width; i++) {
1085
        int b= src1[12*i + 0] + src1[12*i + 6];
1086
        int g= src1[12*i + 2] + src1[12*i + 8];
1087
        int r= src1[12*i + 4] + src1[12*i + 10];
1088

    
1089
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1090
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1091
    }
1092
}
1093

    
1094
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1095
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1096
{\
1097
    int i;\
1098
    for (i=0; i<width; i++) {\
1099
        int b= (((const type*)src)[i]>>shb)&maskb;\
1100
        int g= (((const type*)src)[i]>>shg)&maskg;\
1101
        int r= (((const type*)src)[i]>>shr)&maskr;\
1102
\
1103
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1104
    }\
1105
}
1106

    
1107
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1108
BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)
1109
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1110
BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)
1111
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1112
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1113
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1114
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1115

    
1116
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1117
{
1118
    int i;
1119
    for (i=0; i<width; i++) {
1120
        dst[i]= src[4*i];
1121
    }
1122
}
1123

    
1124
#define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1125
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1126
{\
1127
    int i;\
1128
    for (i=0; i<width; i++) {\
1129
        int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1130
        int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1131
        int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1132
\
1133
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1134
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1135
    }\
1136
}\
1137
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1138
{\
1139
    int i;\
1140
    for (i=0; i<width; i++) {\
1141
        int pix0= ((const type*)src)[2*i+0]>>shp;\
1142
        int pix1= ((const type*)src)[2*i+1]>>shp;\
1143
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1144
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1145
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1146
        g&= maskg|(2*maskg);\
1147
\
1148
        g>>=shg;\
1149
\
1150
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1151
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1152
    }\
1153
}
1154

    
1155
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1156
BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1157
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1158
BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1159
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1160
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1161
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1162
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1163

    
1164
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1165
{
1166
    int i;
1167
    for (i=0; i<width; i++) {
1168
        int d= src[i];
1169

    
1170
        dst[i]= pal[d] & 0xFF;
1171
    }
1172
}
1173

    
1174
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1175
                           const uint8_t *src1, const uint8_t *src2,
1176
                           long width, uint32_t *pal)
1177
{
1178
    int i;
1179
    assert(src1 == src2);
1180
    for (i=0; i<width; i++) {
1181
        int p= pal[src1[i]];
1182

    
1183
        dstU[i]= p>>8;
1184
        dstV[i]= p>>16;
1185
    }
1186
}
1187

    
1188
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1189
{
1190
    int i, j;
1191
    for (i=0; i<width/8; i++) {
1192
        int d= ~src[i];
1193
        for(j=0; j<8; j++)
1194
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1195
    }
1196
}
1197

    
1198
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1199
{
1200
    int i, j;
1201
    for (i=0; i<width/8; i++) {
1202
        int d= src[i];
1203
        for(j=0; j<8; j++)
1204
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1205
    }
1206
}
1207

    
1208
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1209
//Plain C versions
1210
#if CONFIG_RUNTIME_CPUDETECT
1211
#  define COMPILE_C 1
1212
#  if   ARCH_X86
1213
#    define COMPILE_MMX     1
1214
#    define COMPILE_MMX2    1
1215
#    define COMPILE_3DNOW   1
1216
#  elif ARCH_PPC
1217
#    define COMPILE_ALTIVEC HAVE_ALTIVEC
1218
#  endif
1219
#else /* CONFIG_RUNTIME_CPUDETECT */
1220
#  if   ARCH_X86
1221
#    if   HAVE_MMX2
1222
#      define COMPILE_MMX2  1
1223
#    elif HAVE_AMD3DNOW
1224
#      define COMPILE_3DNOW 1
1225
#    elif HAVE_MMX
1226
#      define COMPILE_MMX   1
1227
#    else
1228
#      define COMPILE_C     1
1229
#    endif
1230
#  elif ARCH_PPC && HAVE_ALTIVEC
1231
#    define COMPILE_ALTIVEC 1
1232
#  else
1233
#    define COMPILE_C       1
1234
#  endif
1235
#endif
1236

    
1237
#ifndef COMPILE_C
1238
#  define COMPILE_C 0
1239
#endif
1240
#ifndef COMPILE_MMX
1241
#  define COMPILE_MMX 0
1242
#endif
1243
#ifndef COMPILE_MMX2
1244
#  define COMPILE_MMX2 0
1245
#endif
1246
#ifndef COMPILE_3DNOW
1247
#  define COMPILE_3DNOW 0
1248
#endif
1249
#ifndef COMPILE_ALTIVEC
1250
#  define COMPILE_ALTIVEC 0
1251
#endif
1252

    
1253
#define COMPILE_TEMPLATE_MMX 0
1254
#define COMPILE_TEMPLATE_MMX2 0
1255
#define COMPILE_TEMPLATE_AMD3DNOW 0
1256
#define COMPILE_TEMPLATE_ALTIVEC 0
1257

    
1258
#if COMPILE_C
1259
#define RENAME(a) a ## _C
1260
#include "swscale_template.c"
1261
#endif
1262

    
1263
#if COMPILE_ALTIVEC
1264
#undef RENAME
1265
#undef COMPILE_TEMPLATE_ALTIVEC
1266
#define COMPILE_TEMPLATE_ALTIVEC 1
1267
#define RENAME(a) a ## _altivec
1268
#include "swscale_template.c"
1269
#endif
1270

    
1271
#if ARCH_X86
1272

    
1273
//MMX versions
1274
#if COMPILE_MMX
1275
#undef RENAME
1276
#undef COMPILE_TEMPLATE_MMX
1277
#undef COMPILE_TEMPLATE_MMX2
1278
#undef COMPILE_TEMPLATE_AMD3DNOW
1279
#define COMPILE_TEMPLATE_MMX 1
1280
#define COMPILE_TEMPLATE_MMX2 0
1281
#define COMPILE_TEMPLATE_AMD3DNOW 0
1282
#define RENAME(a) a ## _MMX
1283
#include "swscale_template.c"
1284
#endif
1285

    
1286
//MMX2 versions
1287
#if COMPILE_MMX2
1288
#undef RENAME
1289
#undef COMPILE_TEMPLATE_MMX
1290
#undef COMPILE_TEMPLATE_MMX2
1291
#undef COMPILE_TEMPLATE_AMD3DNOW
1292
#define COMPILE_TEMPLATE_MMX 1
1293
#define COMPILE_TEMPLATE_MMX2 1
1294
#define COMPILE_TEMPLATE_AMD3DNOW 0
1295
#define RENAME(a) a ## _MMX2
1296
#include "swscale_template.c"
1297
#endif
1298

    
1299
//3DNOW versions
1300
#if COMPILE_3DNOW
1301
#undef RENAME
1302
#undef COMPILE_TEMPLATE_MMX
1303
#undef COMPILE_TEMPLATE_MMX2
1304
#undef COMPILE_TEMPLATE_AMD3DNOW
1305
#define COMPILE_TEMPLATE_MMX 1
1306
#define COMPILE_TEMPLATE_MMX2 0
1307
#define COMPILE_TEMPLATE_AMD3DNOW 1
1308
#define RENAME(a) a ## _3DNow
1309
#include "swscale_template.c"
1310
#endif
1311

    
1312
#endif //ARCH_X86
1313

    
1314
SwsFunc ff_getSwsFunc(SwsContext *c)
1315
{
1316
#if CONFIG_RUNTIME_CPUDETECT
1317
    int flags = c->flags;
1318

    
1319
    int cpuflags = av_get_cpu_flags();
1320

    
1321
    flags |= (cpuflags & AV_CPU_FLAG_MMX ? SWS_CPU_CAPS_MMX : 0);
1322
    flags |= (cpuflags & AV_CPU_FLAG_MMX2 ? SWS_CPU_CAPS_MMX2 : 0);
1323
    flags |= (cpuflags & AV_CPU_FLAG_3DNOW ? SWS_CPU_CAPS_3DNOW : 0);
1324

    
1325
#if ARCH_X86
1326
    // ordered per speed fastest first
1327
    if (flags & SWS_CPU_CAPS_MMX2) {
1328
        sws_init_swScale_MMX2(c);
1329
        return swScale_MMX2;
1330
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1331
        sws_init_swScale_3DNow(c);
1332
        return swScale_3DNow;
1333
    } else if (flags & SWS_CPU_CAPS_MMX) {
1334
        sws_init_swScale_MMX(c);
1335
        return swScale_MMX;
1336
    } else {
1337
        sws_init_swScale_C(c);
1338
        return swScale_C;
1339
    }
1340

    
1341
#else
1342
#if COMPILE_ALTIVEC
1343
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1344
        sws_init_swScale_altivec(c);
1345
        return swScale_altivec;
1346
    } else {
1347
        sws_init_swScale_C(c);
1348
        return swScale_C;
1349
    }
1350
#endif
1351
    sws_init_swScale_C(c);
1352
    return swScale_C;
1353
#endif /* ARCH_X86 */
1354
#else //CONFIG_RUNTIME_CPUDETECT
1355
#if   COMPILE_TEMPLATE_MMX2
1356
    sws_init_swScale_MMX2(c);
1357
    return swScale_MMX2;
1358
#elif COMPILE_TEMPLATE_AMD3DNOW
1359
    sws_init_swScale_3DNow(c);
1360
    return swScale_3DNow;
1361
#elif COMPILE_TEMPLATE_MMX
1362
    sws_init_swScale_MMX(c);
1363
    return swScale_MMX;
1364
#elif COMPILE_TEMPLATE_ALTIVEC
1365
    sws_init_swScale_altivec(c);
1366
    return swScale_altivec;
1367
#else
1368
    sws_init_swScale_C(c);
1369
    return swScale_C;
1370
#endif
1371
#endif //!CONFIG_RUNTIME_CPUDETECT
1372
}
1373

    
1374
static void copyPlane(const uint8_t *src, int srcStride,
1375
                      int srcSliceY, int srcSliceH, int width,
1376
                      uint8_t *dst, int dstStride)
1377
{
1378
    dst += dstStride * srcSliceY;
1379
    if (dstStride == srcStride && srcStride > 0) {
1380
        memcpy(dst, src, srcSliceH * dstStride);
1381
    } else {
1382
        int i;
1383
        for (i=0; i<srcSliceH; i++) {
1384
            memcpy(dst, src, width);
1385
            src += srcStride;
1386
            dst += dstStride;
1387
        }
1388
    }
1389
}
1390

    
1391
static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1392
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1393
{
1394
    uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1395

    
1396
    copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1397
              dstParam[0], dstStride[0]);
1398

    
1399
    if (c->dstFormat == PIX_FMT_NV12)
1400
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1401
    else
1402
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1403

    
1404
    return srcSliceH;
1405
}
1406

    
1407
static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1408
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1409
{
1410
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1411

    
1412
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1413

    
1414
    return srcSliceH;
1415
}
1416

    
1417
static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1418
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1419
{
1420
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1421

    
1422
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1423

    
1424
    return srcSliceH;
1425
}
1426

    
1427
static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1428
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1429
{
1430
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1431

    
1432
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1433

    
1434
    return srcSliceH;
1435
}
1436

    
1437
static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1438
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1439
{
1440
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1441

    
1442
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1443

    
1444
    return srcSliceH;
1445
}
1446

    
1447
static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1448
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1449
{
1450
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1451
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1452
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1453

    
1454
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1455

    
1456
    if (dstParam[3])
1457
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1458

    
1459
    return srcSliceH;
1460
}
1461

    
1462
static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1463
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1464
{
1465
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1466
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1467
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1468

    
1469
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1470

    
1471
    return srcSliceH;
1472
}
1473

    
1474
static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1475
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1476
{
1477
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1478
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1479
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1480

    
1481
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1482

    
1483
    if (dstParam[3])
1484
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1485

    
1486
    return srcSliceH;
1487
}
1488

    
1489
static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1490
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1491
{
1492
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1493
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1494
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1495

    
1496
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1497

    
1498
    return srcSliceH;
1499
}
1500

    
1501
static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1502
{
1503
    long i;
1504
    for (i=0; i<num_pixels; i++)
1505
        ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1506
}
1507

    
1508
static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1509
{
1510
    long i;
1511

    
1512
    for (i=0; i<num_pixels; i++)
1513
        ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1514
}
1515

    
1516
static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1517
{
1518
    long i;
1519

    
1520
    for (i=0; i<num_pixels; i++) {
1521
        //FIXME slow?
1522
        dst[0]= palette[src[i<<1]*4+0];
1523
        dst[1]= palette[src[i<<1]*4+1];
1524
        dst[2]= palette[src[i<<1]*4+2];
1525
        dst+= 3;
1526
    }
1527
}
1528

    
1529
static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1530
                           int srcSliceH, uint8_t* dst[], int dstStride[])
1531
{
1532
    const enum PixelFormat srcFormat= c->srcFormat;
1533
    const enum PixelFormat dstFormat= c->dstFormat;
1534
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1535
                 const uint8_t *palette)=NULL;
1536
    int i;
1537
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1538
    const uint8_t *srcPtr= src[0];
1539

    
1540
    if (srcFormat == PIX_FMT_Y400A) {
1541
        switch (dstFormat) {
1542
        case PIX_FMT_RGB32  : conv = gray8aToPacked32; break;
1543
        case PIX_FMT_BGR32  : conv = gray8aToPacked32; break;
1544
        case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1545
        case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1546
        case PIX_FMT_RGB24  : conv = gray8aToPacked24; break;
1547
        case PIX_FMT_BGR24  : conv = gray8aToPacked24; break;
1548
        }
1549
    } else if (usePal(srcFormat)) {
1550
        switch (dstFormat) {
1551
        case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;
1552
        case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;
1553
        case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1554
        case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1555
        case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;
1556
        case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;
1557
        }
1558
    }
1559

    
1560
    if (!conv)
1561
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1562
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1563
    else {
1564
        for (i=0; i<srcSliceH; i++) {
1565
            conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1566
            srcPtr+= srcStride[0];
1567
            dstPtr+= dstStride[0];
1568
        }
1569
    }
1570

    
1571
    return srcSliceH;
1572
}
1573

    
1574
#define isRGBA32(x) (            \
1575
           (x) == PIX_FMT_ARGB   \
1576
        || (x) == PIX_FMT_RGBA   \
1577
        || (x) == PIX_FMT_BGRA   \
1578
        || (x) == PIX_FMT_ABGR   \
1579
        )
1580

    
1581
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1582
static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1583
                           int srcSliceH, uint8_t* dst[], int dstStride[])
1584
{
1585
    const enum PixelFormat srcFormat= c->srcFormat;
1586
    const enum PixelFormat dstFormat= c->dstFormat;
1587
    const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1588
    const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1589
    const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1590
    const int dstId= c->dstFormatBpp >> 2;
1591
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1592

    
1593
#define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1594

    
1595
    if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1596
        if (     CONV_IS(ABGR, RGBA)
1597
              || CONV_IS(ARGB, BGRA)
1598
              || CONV_IS(BGRA, ARGB)
1599
              || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1600
        else if (CONV_IS(ABGR, ARGB)
1601
              || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1602
        else if (CONV_IS(ABGR, BGRA)
1603
              || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1604
        else if (CONV_IS(BGRA, RGBA)
1605
              || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1606
        else if (CONV_IS(BGRA, ABGR)
1607
              || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1608
    } else
1609
    /* BGR -> BGR */
1610
    if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1611
       || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1612
        switch(srcId | (dstId<<4)) {
1613
        case 0x34: conv= rgb16to15; break;
1614
        case 0x36: conv= rgb24to15; break;
1615
        case 0x38: conv= rgb32to15; break;
1616
        case 0x43: conv= rgb15to16; break;
1617
        case 0x46: conv= rgb24to16; break;
1618
        case 0x48: conv= rgb32to16; break;
1619
        case 0x63: conv= rgb15to24; break;
1620
        case 0x64: conv= rgb16to24; break;
1621
        case 0x68: conv= rgb32to24; break;
1622
        case 0x83: conv= rgb15to32; break;
1623
        case 0x84: conv= rgb16to32; break;
1624
        case 0x86: conv= rgb24to32; break;
1625
        }
1626
    } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1627
             || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1628
        switch(srcId | (dstId<<4)) {
1629
        case 0x33: conv= rgb15tobgr15; break;
1630
        case 0x34: conv= rgb16tobgr15; break;
1631
        case 0x36: conv= rgb24tobgr15; break;
1632
        case 0x38: conv= rgb32tobgr15; break;
1633
        case 0x43: conv= rgb15tobgr16; break;
1634
        case 0x44: conv= rgb16tobgr16; break;
1635
        case 0x46: conv= rgb24tobgr16; break;
1636
        case 0x48: conv= rgb32tobgr16; break;
1637
        case 0x63: conv= rgb15tobgr24; break;
1638
        case 0x64: conv= rgb16tobgr24; break;
1639
        case 0x66: conv= rgb24tobgr24; break;
1640
        case 0x68: conv= rgb32tobgr24; break;
1641
        case 0x83: conv= rgb15tobgr32; break;
1642
        case 0x84: conv= rgb16tobgr32; break;
1643
        case 0x86: conv= rgb24tobgr32; break;
1644
        }
1645
    }
1646

    
1647
    if (!conv) {
1648
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1649
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1650
    } else {
1651
        const uint8_t *srcPtr= src[0];
1652
              uint8_t *dstPtr= dst[0];
1653
        if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1654
            srcPtr += ALT32_CORR;
1655

    
1656
        if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1657
            dstPtr += ALT32_CORR;
1658

    
1659
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1660
            conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1661
        else {
1662
            int i;
1663
            dstPtr += dstStride[0]*srcSliceY;
1664

    
1665
            for (i=0; i<srcSliceH; i++) {
1666
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
1667
                srcPtr+= srcStride[0];
1668
                dstPtr+= dstStride[0];
1669
            }
1670
        }
1671
    }
1672
    return srcSliceH;
1673
}
1674

    
1675
static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1676
                              int srcSliceH, uint8_t* dst[], int dstStride[])
1677
{
1678
    rgb24toyv12(
1679
        src[0],
1680
        dst[0]+ srcSliceY    *dstStride[0],
1681
        dst[1]+(srcSliceY>>1)*dstStride[1],
1682
        dst[2]+(srcSliceY>>1)*dstStride[2],
1683
        c->srcW, srcSliceH,
1684
        dstStride[0], dstStride[1], srcStride[0]);
1685
    if (dst[3])
1686
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1687
    return srcSliceH;
1688
}
1689

    
1690
static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1691
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1692
{
1693
    copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1694
              dst[0], dstStride[0]);
1695

    
1696
    planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1697
             srcSliceH >> 2, srcStride[1], dstStride[1]);
1698
    planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1699
             srcSliceH >> 2, srcStride[2], dstStride[2]);
1700
    if (dst[3])
1701
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1702
    return srcSliceH;
1703
}
1704

    
1705
/* unscaled copy like stuff (assumes nearly identical formats) */
1706
static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1707
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1708
{
1709
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1710
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1711
    else {
1712
        int i;
1713
        const uint8_t *srcPtr= src[0];
1714
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1715
        int length=0;
1716

    
1717
        /* universal length finder */
1718
        while(length+c->srcW <= FFABS(dstStride[0])
1719
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1720
        assert(length!=0);
1721

    
1722
        for (i=0; i<srcSliceH; i++) {
1723
            memcpy(dstPtr, srcPtr, length);
1724
            srcPtr+= srcStride[0];
1725
            dstPtr+= dstStride[0];
1726
        }
1727
    }
1728
    return srcSliceH;
1729
}
1730

    
1731
static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1732
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1733
{
1734
    int plane, i, j;
1735
    for (plane=0; plane<4; plane++) {
1736
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1737
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1738
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1739
        const uint8_t *srcPtr= src[plane];
1740
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1741

    
1742
        if (!dst[plane]) continue;
1743
        // ignore palette for GRAY8
1744
        if (plane == 1 && !dst[2]) continue;
1745
        if (!src[plane] || (plane == 1 && !src[2])) {
1746
            if(is16BPS(c->dstFormat))
1747
                length*=2;
1748
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1749
        } else {
1750
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1751
                if (!isBE(c->srcFormat)) srcPtr++;
1752
                for (i=0; i<height; i++) {
1753
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1754
                    srcPtr+= srcStride[plane];
1755
                    dstPtr+= dstStride[plane];
1756
                }
1757
            } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1758
                for (i=0; i<height; i++) {
1759
                    for (j=0; j<length; j++) {
1760
                        dstPtr[ j<<1   ] = srcPtr[j];
1761
                        dstPtr[(j<<1)+1] = srcPtr[j];
1762
                    }
1763
                    srcPtr+= srcStride[plane];
1764
                    dstPtr+= dstStride[plane];
1765
                }
1766
            } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1767
                  && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1768

    
1769
                for (i=0; i<height; i++) {
1770
                    for (j=0; j<length; j++)
1771
                        ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1772
                    srcPtr+= srcStride[plane];
1773
                    dstPtr+= dstStride[plane];
1774
                }
1775
            } else if (dstStride[plane] == srcStride[plane] &&
1776
                       srcStride[plane] > 0 && srcStride[plane] == length) {
1777
                memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1778
                       height*dstStride[plane]);
1779
            } else {
1780
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1781
                    length*=2;
1782
                for (i=0; i<height; i++) {
1783
                    memcpy(dstPtr, srcPtr, length);
1784
                    srcPtr+= srcStride[plane];
1785
                    dstPtr+= dstStride[plane];
1786
                }
1787
            }
1788
        }
1789
    }
1790
    return srcSliceH;
1791
}
1792

    
1793
int ff_hardcodedcpuflags(void)
1794
{
1795
    int flags = 0;
1796
#if   COMPILE_TEMPLATE_MMX2
1797
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1798
#elif COMPILE_TEMPLATE_AMD3DNOW
1799
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1800
#elif COMPILE_TEMPLATE_MMX
1801
    flags |= SWS_CPU_CAPS_MMX;
1802
#elif COMPILE_TEMPLATE_ALTIVEC
1803
    flags |= SWS_CPU_CAPS_ALTIVEC;
1804
#elif ARCH_BFIN
1805
    flags |= SWS_CPU_CAPS_BFIN;
1806
#endif
1807
    return flags;
1808
}
1809

    
1810
void ff_get_unscaled_swscale(SwsContext *c)
1811
{
1812
    const enum PixelFormat srcFormat = c->srcFormat;
1813
    const enum PixelFormat dstFormat = c->dstFormat;
1814
    const int flags = c->flags;
1815
    const int dstH = c->dstH;
1816
    int needsDither;
1817

    
1818
    needsDither= isAnyRGB(dstFormat)
1819
        &&  c->dstFormatBpp < 24
1820
        && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1821

    
1822
    /* yv12_to_nv12 */
1823
    if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1824
        c->swScale= planarToNv12Wrapper;
1825
    }
1826
    /* yuv2bgr */
1827
    if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1828
        && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1829
        c->swScale= ff_yuv2rgb_get_func_ptr(c);
1830
    }
1831

    
1832
    if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1833
        c->swScale= yvu9ToYv12Wrapper;
1834
    }
1835

    
1836
    /* bgr24toYV12 */
1837
    if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1838
        c->swScale= bgr24ToYv12Wrapper;
1839

    
1840
    /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1841
    if (   isAnyRGB(srcFormat)
1842
        && isAnyRGB(dstFormat)
1843
        && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
1844
        && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
1845
        && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
1846
        && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
1847
        && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1848
        && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1849
        && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1850
        && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1851
        && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
1852
        && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
1853
        && srcFormat != PIX_FMT_BGR48LE   && dstFormat != PIX_FMT_BGR48LE
1854
        && srcFormat != PIX_FMT_BGR48BE   && dstFormat != PIX_FMT_BGR48BE
1855
        && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1856
        c->swScale= rgbToRgbWrapper;
1857

    
1858
    if ((usePal(srcFormat) && (
1859
        dstFormat == PIX_FMT_RGB32   ||
1860
        dstFormat == PIX_FMT_RGB32_1 ||
1861
        dstFormat == PIX_FMT_RGB24   ||
1862
        dstFormat == PIX_FMT_BGR32   ||
1863
        dstFormat == PIX_FMT_BGR32_1 ||
1864
        dstFormat == PIX_FMT_BGR24)))
1865
        c->swScale= palToRgbWrapper;
1866

    
1867
    if (srcFormat == PIX_FMT_YUV422P) {
1868
        if (dstFormat == PIX_FMT_YUYV422)
1869
            c->swScale= yuv422pToYuy2Wrapper;
1870
        else if (dstFormat == PIX_FMT_UYVY422)
1871
            c->swScale= yuv422pToUyvyWrapper;
1872
    }
1873

    
1874
    /* LQ converters if -sws 0 or -sws 4*/
1875
    if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1876
        /* yv12_to_yuy2 */
1877
        if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1878
            if (dstFormat == PIX_FMT_YUYV422)
1879
                c->swScale= planarToYuy2Wrapper;
1880
            else if (dstFormat == PIX_FMT_UYVY422)
1881
                c->swScale= planarToUyvyWrapper;
1882
        }
1883
    }
1884
    if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1885
        c->swScale= yuyvToYuv420Wrapper;
1886
    if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1887
        c->swScale= uyvyToYuv420Wrapper;
1888
    if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1889
        c->swScale= yuyvToYuv422Wrapper;
1890
    if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1891
        c->swScale= uyvyToYuv422Wrapper;
1892

    
1893
#if COMPILE_ALTIVEC
1894
    if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1895
        !(c->flags & SWS_BITEXACT) &&
1896
        srcFormat == PIX_FMT_YUV420P) {
1897
        // unscaled YV12 -> packed YUV, we want speed
1898
        if (dstFormat == PIX_FMT_YUYV422)
1899
            c->swScale= yv12toyuy2_unscaled_altivec;
1900
        else if (dstFormat == PIX_FMT_UYVY422)
1901
            c->swScale= yv12touyvy_unscaled_altivec;
1902
    }
1903
#endif
1904

    
1905
    /* simple copy */
1906
    if (  srcFormat == dstFormat
1907
        || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1908
        || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1909
        || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1910
        || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1911
        || (isGray(dstFormat) && isGray(srcFormat))
1912
        || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1913
            && c->chrDstHSubSample == c->chrSrcHSubSample
1914
            && c->chrDstVSubSample == c->chrSrcVSubSample
1915
            && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1916
            && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1917
    {
1918
        if (isPacked(c->srcFormat))
1919
            c->swScale= packedCopyWrapper;
1920
        else /* Planar YUV or gray */
1921
            c->swScale= planarCopyWrapper;
1922
    }
1923
#if ARCH_BFIN
1924
    if (flags & SWS_CPU_CAPS_BFIN)
1925
        ff_bfin_get_unscaled_swscale (c);
1926
#endif
1927
}
1928

    
1929
static void reset_ptr(const uint8_t* src[], int format)
1930
{
1931
    if(!isALPHA(format))
1932
        src[3]=NULL;
1933
    if(!isPlanarYUV(format)) {
1934
        src[3]=src[2]=NULL;
1935

    
1936
        if (!usePal(format))
1937
            src[1]= NULL;
1938
    }
1939
}
1940

    
1941
static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1942
                                const int linesizes[4])
1943
{
1944
    const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1945
    int i;
1946

    
1947
    for (i = 0; i < 4; i++) {
1948
        int plane = desc->comp[i].plane;
1949
        if (!data[plane] || !linesizes[plane])
1950
            return 0;
1951
    }
1952

    
1953
    return 1;
1954
}
1955

    
1956
/**
1957
 * swscale wrapper, so we don't need to export the SwsContext.
1958
 * Assumes planar YUV to be in YUV order instead of YVU.
1959
 */
1960
int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1961
              int srcSliceH, uint8_t* const dst[], const int dstStride[])
1962
{
1963
    int i;
1964
    const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1965
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1966

    
1967
    // do not mess up sliceDir if we have a "trailing" 0-size slice
1968
    if (srcSliceH == 0)
1969
        return 0;
1970

    
1971
    if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1972
        av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1973
        return 0;
1974
    }
1975
    if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1976
        av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1977
        return 0;
1978
    }
1979

    
1980
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1981
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1982
        return 0;
1983
    }
1984
    if (c->sliceDir == 0) {
1985
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1986
    }
1987

    
1988
    if (usePal(c->srcFormat)) {
1989
        for (i=0; i<256; i++) {
1990
            int p, r, g, b,y,u,v;
1991
            if(c->srcFormat == PIX_FMT_PAL8) {
1992
                p=((const uint32_t*)(src[1]))[i];
1993
                r= (p>>16)&0xFF;
1994
                g= (p>> 8)&0xFF;
1995
                b=  p     &0xFF;
1996
            } else if(c->srcFormat == PIX_FMT_RGB8) {
1997
                r= (i>>5    )*36;
1998
                g= ((i>>2)&7)*36;
1999
                b= (i&3     )*85;
2000
            } else if(c->srcFormat == PIX_FMT_BGR8) {
2001
                b= (i>>6    )*85;
2002
                g= ((i>>3)&7)*36;
2003
                r= (i&7     )*36;
2004
            } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
2005
                r= (i>>3    )*255;
2006
                g= ((i>>1)&3)*85;
2007
                b= (i&1     )*255;
2008
            } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
2009
                r = g = b = i;
2010
            } else {
2011
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2012
                b= (i>>3    )*255;
2013
                g= ((i>>1)&3)*85;
2014
                r= (i&1     )*255;
2015
            }
2016
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2017
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2018
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2019
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
2020

    
2021
            switch(c->dstFormat) {
2022
            case PIX_FMT_BGR32:
2023
#if !HAVE_BIGENDIAN
2024
            case PIX_FMT_RGB24:
2025
#endif
2026
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
2027
                break;
2028
            case PIX_FMT_BGR32_1:
2029
#if HAVE_BIGENDIAN
2030
            case PIX_FMT_BGR24:
2031
#endif
2032
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2033
                break;
2034
            case PIX_FMT_RGB32_1:
2035
#if HAVE_BIGENDIAN
2036
            case PIX_FMT_RGB24:
2037
#endif
2038
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2039
                break;
2040
            case PIX_FMT_RGB32:
2041
#if !HAVE_BIGENDIAN
2042
            case PIX_FMT_BGR24:
2043
#endif
2044
            default:
2045
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
2046
            }
2047
        }
2048
    }
2049

    
2050
    // copy strides, so they can safely be modified
2051
    if (c->sliceDir == 1) {
2052
        // slices go from top to bottom
2053
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2054
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2055

    
2056
        reset_ptr(src2, c->srcFormat);
2057
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
2058

    
2059
        /* reset slice direction at end of frame */
2060
        if (srcSliceY + srcSliceH == c->srcH)
2061
            c->sliceDir = 0;
2062

    
2063
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2064
    } else {
2065
        // slices go from bottom to top => we flip the image internally
2066
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2067
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2068

    
2069
        src2[0] += (srcSliceH-1)*srcStride[0];
2070
        if (!usePal(c->srcFormat))
2071
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2072
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2073
        src2[3] += (srcSliceH-1)*srcStride[3];
2074
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
2075
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2076
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2077
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
2078

    
2079
        reset_ptr(src2, c->srcFormat);
2080
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
2081

    
2082
        /* reset slice direction at end of frame */
2083
        if (!srcSliceY)
2084
            c->sliceDir = 0;
2085

    
2086
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2087
    }
2088
}
2089

    
2090
#if LIBSWSCALE_VERSION_MAJOR < 1
2091
int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
2092
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2093
{
2094
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2095
}
2096
#endif
2097

    
2098
/* Convert the palette to the same packed 32-bit format as the palette */
2099
void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2100
{
2101
    long i;
2102

    
2103
    for (i=0; i<num_pixels; i++)
2104
        ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2105
}
2106

    
2107
/* Palette format: ABCD -> dst format: ABC */
2108
void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2109
{
2110
    long i;
2111

    
2112
    for (i=0; i<num_pixels; i++) {
2113
        //FIXME slow?
2114
        dst[0]= palette[src[i]*4+0];
2115
        dst[1]= palette[src[i]*4+1];
2116
        dst[2]= palette[src[i]*4+2];
2117
        dst+= 3;
2118
    }
2119
}