Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ aa2ead82

History | View | Annotate | Download (67.2 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 modify
7
 * it under the terms of the GNU General Public License as published by
8
 * the Free Software Foundation; either version 2 of the License, or
9
 * (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
14
 * GNU General Public License for more details.
15
 *
16
 * You should have received a copy of the GNU General Public License
17
 * 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
 * the C code (not assembly, mmx, ...) of this file can be used
21
 * under the LGPL license too
22
 */
23

    
24
/*
25
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
26
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
27
  {BGR,RGB}{1,4,8,15,16} support dithering
28

29
  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
30
  YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
31
  x -> x
32
  YUV9 -> YV12
33
  YUV9/YV12 -> Y800
34
  Y800 -> YUV9/YV12
35
  BGR24 -> BGR32 & RGB24 -> RGB32
36
  BGR32 -> BGR24 & RGB32 -> RGB24
37
  BGR15 -> BGR16
38
*/
39

    
40
/*
41
tested special converters (most are tested actually, but I did not write it down ...)
42
 YV12 -> BGR16
43
 YV12 -> YV12
44
 BGR15 -> BGR16
45
 BGR16 -> BGR16
46
 YVU9 -> YV12
47

48
untested special converters
49
  YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
50
  YV12/I420 -> YV12/I420
51
  YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
52
  BGR24 -> BGR32 & RGB24 -> RGB32
53
  BGR32 -> BGR24 & RGB32 -> RGB24
54
  BGR24 -> YV12
55
*/
56

    
57
#include <inttypes.h>
58
#include <string.h>
59
#include <math.h>
60
#include <stdio.h>
61
#include "config.h"
62
#include <assert.h>
63
#include "swscale.h"
64
#include "swscale_internal.h"
65
#include "rgb2rgb.h"
66
#include "libavutil/intreadwrite.h"
67
#include "libavutil/x86_cpu.h"
68
#include "libavutil/avutil.h"
69
#include "libavutil/bswap.h"
70
#include "libavutil/pixdesc.h"
71

    
72
#undef MOVNTQ
73
#undef PAVGB
74

    
75
//#undef HAVE_MMX2
76
//#define HAVE_AMD3DNOW
77
//#undef HAVE_MMX
78
//#undef ARCH_X86
79
#define DITHER1XBPP
80

    
81
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
82

    
83
#ifdef M_PI
84
#define PI M_PI
85
#else
86
#define PI 3.14159265358979323846
87
#endif
88

    
89
#define isPacked(x)         (       \
90
           (x)==PIX_FMT_PAL8        \
91
        || (x)==PIX_FMT_YUYV422     \
92
        || (x)==PIX_FMT_UYVY422     \
93
        || isRGB(x)                 \
94
        || isBGR(x)                 \
95
    )
96
#define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)
97

    
98
#define RGB2YUV_SHIFT 15
99
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
100
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
101
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
102
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
103
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
104
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
105
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
106
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
107
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
108

    
109
static const double rgb2yuv_table[8][9]={
110
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
111
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
112
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
113
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
114
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
115
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
116
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
117
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
118
};
119

    
120
/*
121
NOTES
122
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
123

124
TODO
125
more intelligent misalignment avoidance for the horizontal scaler
126
write special vertical cubic upscale version
127
optimize C code (YV12 / minmax)
128
add support for packed pixel YUV input & output
129
add support for Y8 output
130
optimize BGR24 & BGR32
131
add BGR4 output support
132
write special BGR->BGR scaler
133
*/
134

    
135
#if ARCH_X86 && CONFIG_GPL
136
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
137
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
138
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
139
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
140
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
141
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
142
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
143
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
144

    
145
const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
146
        0x0103010301030103LL,
147
        0x0200020002000200LL,};
148

    
149
const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
150
        0x0602060206020602LL,
151
        0x0004000400040004LL,};
152

    
153
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
154
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
155
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
156
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
157
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
158
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
159

    
160
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
161
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
162
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
163

    
164
#ifdef FAST_BGR2YV12
165
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
166
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
167
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
168
#else
169
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
170
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
171
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
172
#endif /* FAST_BGR2YV12 */
173
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
174
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
175
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
176

    
177
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
178
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
179
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
180
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
181
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
182

    
183
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
184
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
185
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
186
};
187

    
188
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
189

    
190
#endif /* ARCH_X86 && CONFIG_GPL */
191

    
192
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
193
{  1,   3,   1,   3,   1,   3,   1,   3, },
194
{  2,   0,   2,   0,   2,   0,   2,   0, },
195
};
196

    
197
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
198
{  6,   2,   6,   2,   6,   2,   6,   2, },
199
{  0,   4,   0,   4,   0,   4,   0,   4, },
200
};
201

    
202
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
203
{ 17,   9,  23,  15,  16,   8,  22,  14, },
204
{  5,  29,   3,  27,   4,  28,   2,  26, },
205
{ 21,  13,  19,  11,  20,  12,  18,  10, },
206
{  0,  24,   6,  30,   1,  25,   7,  31, },
207
{ 16,   8,  22,  14,  17,   9,  23,  15, },
208
{  4,  28,   2,  26,   5,  29,   3,  27, },
209
{ 20,  12,  18,  10,  21,  13,  19,  11, },
210
{  1,  25,   7,  31,   0,  24,   6,  30, },
211
};
212

    
213
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
214
{  0,  55,  14,  68,   3,  58,  17,  72, },
215
{ 37,  18,  50,  32,  40,  22,  54,  35, },
216
{  9,  64,   5,  59,  13,  67,   8,  63, },
217
{ 46,  27,  41,  23,  49,  31,  44,  26, },
218
{  2,  57,  16,  71,   1,  56,  15,  70, },
219
{ 39,  21,  52,  34,  38,  19,  51,  33, },
220
{ 11,  66,   7,  62,  10,  65,   6,  60, },
221
{ 48,  30,  43,  25,  47,  29,  42,  24, },
222
};
223

    
224
#if 1
225
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
226
{117,  62, 158, 103, 113,  58, 155, 100, },
227
{ 34, 199,  21, 186,  31, 196,  17, 182, },
228
{144,  89, 131,  76, 141,  86, 127,  72, },
229
{  0, 165,  41, 206,  10, 175,  52, 217, },
230
{110,  55, 151,  96, 120,  65, 162, 107, },
231
{ 28, 193,  14, 179,  38, 203,  24, 189, },
232
{138,  83, 124,  69, 148,  93, 134,  79, },
233
{  7, 172,  48, 213,   3, 168,  45, 210, },
234
};
235
#elif 1
236
// tries to correct a gamma of 1.5
237
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
238
{  0, 143,  18, 200,   2, 156,  25, 215, },
239
{ 78,  28, 125,  64,  89,  36, 138,  74, },
240
{ 10, 180,   3, 161,  16, 195,   8, 175, },
241
{109,  51,  93,  38, 121,  60, 105,  47, },
242
{  1, 152,  23, 210,   0, 147,  20, 205, },
243
{ 85,  33, 134,  71,  81,  30, 130,  67, },
244
{ 14, 190,   6, 171,  12, 185,   5, 166, },
245
{117,  57, 101,  44, 113,  54,  97,  41, },
246
};
247
#elif 1
248
// tries to correct a gamma of 2.0
249
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
250
{  0, 124,   8, 193,   0, 140,  12, 213, },
251
{ 55,  14, 104,  42,  66,  19, 119,  52, },
252
{  3, 168,   1, 145,   6, 187,   3, 162, },
253
{ 86,  31,  70,  21,  99,  39,  82,  28, },
254
{  0, 134,  11, 206,   0, 129,   9, 200, },
255
{ 62,  17, 114,  48,  58,  16, 109,  45, },
256
{  5, 181,   2, 157,   4, 175,   1, 151, },
257
{ 95,  36,  78,  26,  90,  34,  74,  24, },
258
};
259
#else
260
// tries to correct a gamma of 2.5
261
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
262
{  0, 107,   3, 187,   0, 125,   6, 212, },
263
{ 39,   7,  86,  28,  49,  11, 102,  36, },
264
{  1, 158,   0, 131,   3, 180,   1, 151, },
265
{ 68,  19,  52,  12,  81,  25,  64,  17, },
266
{  0, 119,   5, 203,   0, 113,   4, 195, },
267
{ 45,   9,  96,  33,  42,   8,  91,  30, },
268
{  2, 172,   1, 144,   2, 165,   0, 137, },
269
{ 77,  23,  60,  15,  72,  21,  56,  14, },
270
};
271
#endif
272

    
273
static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
274
                                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
275
                                                    const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
276
                                                    int dstW, int chrDstW, int big_endian)
277
{
278
    //FIXME Optimize (just quickly written not optimized..)
279
    int i;
280

    
281
    for (i = 0; i < dstW; i++) {
282
        int val = 1 << 10;
283
        int j;
284

    
285
        for (j = 0; j < lumFilterSize; j++)
286
            val += lumSrc[j][i] * lumFilter[j];
287

    
288
        if (big_endian) {
289
            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
290
        } else {
291
            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
292
        }
293
    }
294

    
295
    if (uDest) {
296
        for (i = 0; i < chrDstW; i++) {
297
            int u = 1 << 10;
298
            int v = 1 << 10;
299
            int j;
300

    
301
            for (j = 0; j < chrFilterSize; j++) {
302
                u += chrSrc[j][i       ] * chrFilter[j];
303
                v += chrSrc[j][i + VOFW] * chrFilter[j];
304
            }
305

    
306
            if (big_endian) {
307
                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
308
                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
309
            } else {
310
                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
311
                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
312
            }
313
        }
314
    }
315

    
316
    if (CONFIG_SWSCALE_ALPHA && aDest) {
317
        for (i = 0; i < dstW; i++) {
318
            int val = 1 << 10;
319
            int j;
320

    
321
            for (j = 0; j < lumFilterSize; j++)
322
                val += alpSrc[j][i] * lumFilter[j];
323

    
324
            if (big_endian) {
325
                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
326
            } else {
327
                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
328
            }
329
        }
330
    }
331
}
332

    
333
static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
334
                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
335
                                 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
336
                                 enum PixelFormat dstFormat)
337
{
338
    if (isBE(dstFormat)) {
339
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
340
                               chrFilter, chrSrc, chrFilterSize,
341
                               alpSrc,
342
                               dest, uDest, vDest, aDest,
343
                               dstW, chrDstW, 1);
344
    } else {
345
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
346
                               chrFilter, chrSrc, chrFilterSize,
347
                               alpSrc,
348
                               dest, uDest, vDest, aDest,
349
                               dstW, chrDstW, 0);
350
    }
351
}
352

    
353
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
354
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
355
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
356
{
357
    //FIXME Optimize (just quickly written not optimized..)
358
    int i;
359
    for (i=0; i<dstW; i++) {
360
        int val=1<<18;
361
        int j;
362
        for (j=0; j<lumFilterSize; j++)
363
            val += lumSrc[j][i] * lumFilter[j];
364

    
365
        dest[i]= av_clip_uint8(val>>19);
366
    }
367

    
368
    if (uDest)
369
        for (i=0; i<chrDstW; i++) {
370
            int u=1<<18;
371
            int v=1<<18;
372
            int j;
373
            for (j=0; j<chrFilterSize; j++) {
374
                u += chrSrc[j][i] * chrFilter[j];
375
                v += chrSrc[j][i + VOFW] * chrFilter[j];
376
            }
377

    
378
            uDest[i]= av_clip_uint8(u>>19);
379
            vDest[i]= av_clip_uint8(v>>19);
380
        }
381

    
382
    if (CONFIG_SWSCALE_ALPHA && aDest)
383
        for (i=0; i<dstW; i++) {
384
            int val=1<<18;
385
            int j;
386
            for (j=0; j<lumFilterSize; j++)
387
                val += alpSrc[j][i] * lumFilter[j];
388

    
389
            aDest[i]= av_clip_uint8(val>>19);
390
        }
391

    
392
}
393

    
394
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
395
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
396
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
397
{
398
    //FIXME Optimize (just quickly written not optimized..)
399
    int i;
400
    for (i=0; i<dstW; i++) {
401
        int val=1<<18;
402
        int j;
403
        for (j=0; j<lumFilterSize; j++)
404
            val += lumSrc[j][i] * lumFilter[j];
405

    
406
        dest[i]= av_clip_uint8(val>>19);
407
    }
408

    
409
    if (!uDest)
410
        return;
411

    
412
    if (dstFormat == PIX_FMT_NV12)
413
        for (i=0; i<chrDstW; i++) {
414
            int u=1<<18;
415
            int v=1<<18;
416
            int j;
417
            for (j=0; j<chrFilterSize; j++) {
418
                u += chrSrc[j][i] * chrFilter[j];
419
                v += chrSrc[j][i + VOFW] * chrFilter[j];
420
            }
421

    
422
            uDest[2*i]= av_clip_uint8(u>>19);
423
            uDest[2*i+1]= av_clip_uint8(v>>19);
424
        }
425
    else
426
        for (i=0; i<chrDstW; i++) {
427
            int u=1<<18;
428
            int v=1<<18;
429
            int j;
430
            for (j=0; j<chrFilterSize; j++) {
431
                u += chrSrc[j][i] * chrFilter[j];
432
                v += chrSrc[j][i + VOFW] * chrFilter[j];
433
            }
434

    
435
            uDest[2*i]= av_clip_uint8(v>>19);
436
            uDest[2*i+1]= av_clip_uint8(u>>19);
437
        }
438
}
439

    
440
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
441
    for (i=0; i<(dstW>>1); i++) {\
442
        int j;\
443
        int Y1 = 1<<18;\
444
        int Y2 = 1<<18;\
445
        int U  = 1<<18;\
446
        int V  = 1<<18;\
447
        int av_unused A1, A2;\
448
        type av_unused *r, *b, *g;\
449
        const int i2= 2*i;\
450
        \
451
        for (j=0; j<lumFilterSize; j++) {\
452
            Y1 += lumSrc[j][i2] * lumFilter[j];\
453
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
454
        }\
455
        for (j=0; j<chrFilterSize; j++) {\
456
            U += chrSrc[j][i] * chrFilter[j];\
457
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
458
        }\
459
        Y1>>=19;\
460
        Y2>>=19;\
461
        U >>=19;\
462
        V >>=19;\
463
        if (alpha) {\
464
            A1 = 1<<18;\
465
            A2 = 1<<18;\
466
            for (j=0; j<lumFilterSize; j++) {\
467
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
468
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
469
            }\
470
            A1>>=19;\
471
            A2>>=19;\
472
        }
473

    
474
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
475
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
476
        if ((Y1|Y2|U|V)&256) {\
477
            if (Y1>255)   Y1=255; \
478
            else if (Y1<0)Y1=0;   \
479
            if (Y2>255)   Y2=255; \
480
            else if (Y2<0)Y2=0;   \
481
            if (U>255)    U=255;  \
482
            else if (U<0) U=0;    \
483
            if (V>255)    V=255;  \
484
            else if (V<0) V=0;    \
485
        }\
486
        if (alpha && ((A1|A2)&256)) {\
487
            A1=av_clip_uint8(A1);\
488
            A2=av_clip_uint8(A2);\
489
        }
490

    
491
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
492
    for (i=0; i<dstW; i++) {\
493
        int j;\
494
        int Y = 0;\
495
        int U = -128<<19;\
496
        int V = -128<<19;\
497
        int av_unused A;\
498
        int R,G,B;\
499
        \
500
        for (j=0; j<lumFilterSize; j++) {\
501
            Y += lumSrc[j][i     ] * lumFilter[j];\
502
        }\
503
        for (j=0; j<chrFilterSize; j++) {\
504
            U += chrSrc[j][i     ] * chrFilter[j];\
505
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
506
        }\
507
        Y >>=10;\
508
        U >>=10;\
509
        V >>=10;\
510
        if (alpha) {\
511
            A = rnd;\
512
            for (j=0; j<lumFilterSize; j++)\
513
                A += alpSrc[j][i     ] * lumFilter[j];\
514
            A >>=19;\
515
            if (A&256)\
516
                A = av_clip_uint8(A);\
517
        }
518

    
519
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
520
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
521
        Y-= c->yuv2rgb_y_offset;\
522
        Y*= c->yuv2rgb_y_coeff;\
523
        Y+= rnd;\
524
        R= Y + V*c->yuv2rgb_v2r_coeff;\
525
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
526
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
527
        if ((R|G|B)&(0xC0000000)) {\
528
            if (R>=(256<<22))   R=(256<<22)-1; \
529
            else if (R<0)R=0;   \
530
            if (G>=(256<<22))   G=(256<<22)-1; \
531
            else if (G<0)G=0;   \
532
            if (B>=(256<<22))   B=(256<<22)-1; \
533
            else if (B<0)B=0;   \
534
        }
535

    
536
#define YSCALE_YUV_2_GRAY16_C \
537
    for (i=0; i<(dstW>>1); i++) {\
538
        int j;\
539
        int Y1 = 1<<18;\
540
        int Y2 = 1<<18;\
541
        int U  = 1<<18;\
542
        int V  = 1<<18;\
543
        \
544
        const int i2= 2*i;\
545
        \
546
        for (j=0; j<lumFilterSize; j++) {\
547
            Y1 += lumSrc[j][i2] * lumFilter[j];\
548
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
549
        }\
550
        Y1>>=11;\
551
        Y2>>=11;\
552
        if ((Y1|Y2|U|V)&65536) {\
553
            if (Y1>65535)   Y1=65535; \
554
            else if (Y1<0)Y1=0;   \
555
            if (Y2>65535)   Y2=65535; \
556
            else if (Y2<0)Y2=0;   \
557
        }
558

    
559
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
560
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
561
    r = (type *)c->table_rV[V];   \
562
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
563
    b = (type *)c->table_bU[U];
564

    
565
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
566
    for (i=0; i<(dstW>>1); i++) { \
567
        const int i2= 2*i;       \
568
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
569
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
570
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
571
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
572
        type av_unused *r, *b, *g;                                    \
573
        int av_unused A1, A2;                                         \
574
        if (alpha) {\
575
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
576
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
577
        }
578

    
579
#define YSCALE_YUV_2_GRAY16_2_C   \
580
    for (i=0; i<(dstW>>1); i++) { \
581
        const int i2= 2*i;       \
582
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
583
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
584

    
585
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
586
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
587
    r = (type *)c->table_rV[V];\
588
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
589
    b = (type *)c->table_bU[U];
590

    
591
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
592
    for (i=0; i<(dstW>>1); i++) {\
593
        const int i2= 2*i;\
594
        int Y1= buf0[i2  ]>>7;\
595
        int Y2= buf0[i2+1]>>7;\
596
        int U= (uvbuf1[i     ])>>7;\
597
        int V= (uvbuf1[i+VOFW])>>7;\
598
        type av_unused *r, *b, *g;\
599
        int av_unused A1, A2;\
600
        if (alpha) {\
601
            A1= abuf0[i2  ]>>7;\
602
            A2= abuf0[i2+1]>>7;\
603
        }
604

    
605
#define YSCALE_YUV_2_GRAY16_1_C \
606
    for (i=0; i<(dstW>>1); i++) {\
607
        const int i2= 2*i;\
608
        int Y1= buf0[i2  ]<<1;\
609
        int Y2= buf0[i2+1]<<1;
610

    
611
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
612
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
613
    r = (type *)c->table_rV[V];\
614
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
615
    b = (type *)c->table_bU[U];
616

    
617
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
618
    for (i=0; i<(dstW>>1); i++) {\
619
        const int i2= 2*i;\
620
        int Y1= buf0[i2  ]>>7;\
621
        int Y2= buf0[i2+1]>>7;\
622
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
623
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
624
        type av_unused *r, *b, *g;\
625
        int av_unused A1, A2;\
626
        if (alpha) {\
627
            A1= abuf0[i2  ]>>7;\
628
            A2= abuf0[i2+1]>>7;\
629
        }
630

    
631
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
632
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
633
    r = (type *)c->table_rV[V];\
634
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
635
    b = (type *)c->table_bU[U];
636

    
637
#define YSCALE_YUV_2_MONO2_C \
638
    const uint8_t * const d128=dither_8x8_220[y&7];\
639
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
640
    for (i=0; i<dstW-7; i+=8) {\
641
        int acc;\
642
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
643
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
644
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
645
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
646
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
647
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
648
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
649
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
650
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
651
        dest++;\
652
    }
653

    
654
#define YSCALE_YUV_2_MONOX_C \
655
    const uint8_t * const d128=dither_8x8_220[y&7];\
656
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
657
    int acc=0;\
658
    for (i=0; i<dstW-1; i+=2) {\
659
        int j;\
660
        int Y1=1<<18;\
661
        int Y2=1<<18;\
662
\
663
        for (j=0; j<lumFilterSize; j++) {\
664
            Y1 += lumSrc[j][i] * lumFilter[j];\
665
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
666
        }\
667
        Y1>>=19;\
668
        Y2>>=19;\
669
        if ((Y1|Y2)&256) {\
670
            if (Y1>255)   Y1=255;\
671
            else if (Y1<0)Y1=0;\
672
            if (Y2>255)   Y2=255;\
673
            else if (Y2<0)Y2=0;\
674
        }\
675
        acc+= acc + g[Y1+d128[(i+0)&7]];\
676
        acc+= acc + g[Y2+d128[(i+1)&7]];\
677
        if ((i&7)==6) {\
678
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
679
            dest++;\
680
        }\
681
    }
682

    
683
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
684
    switch(c->dstFormat) {\
685
    case PIX_FMT_RGB48BE:\
686
    case PIX_FMT_RGB48LE:\
687
        func(uint8_t,0)\
688
            ((uint8_t*)dest)[ 0]= r[Y1];\
689
            ((uint8_t*)dest)[ 1]= r[Y1];\
690
            ((uint8_t*)dest)[ 2]= g[Y1];\
691
            ((uint8_t*)dest)[ 3]= g[Y1];\
692
            ((uint8_t*)dest)[ 4]= b[Y1];\
693
            ((uint8_t*)dest)[ 5]= b[Y1];\
694
            ((uint8_t*)dest)[ 6]= r[Y2];\
695
            ((uint8_t*)dest)[ 7]= r[Y2];\
696
            ((uint8_t*)dest)[ 8]= g[Y2];\
697
            ((uint8_t*)dest)[ 9]= g[Y2];\
698
            ((uint8_t*)dest)[10]= b[Y2];\
699
            ((uint8_t*)dest)[11]= b[Y2];\
700
            dest+=12;\
701
        }\
702
        break;\
703
    case PIX_FMT_RGBA:\
704
    case PIX_FMT_BGRA:\
705
        if (CONFIG_SMALL) {\
706
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
707
            func(uint32_t,needAlpha)\
708
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
709
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
710
            }\
711
        } else {\
712
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
713
                func(uint32_t,1)\
714
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
715
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
716
                }\
717
            } else {\
718
                func(uint32_t,0)\
719
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
720
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
721
                }\
722
            }\
723
        }\
724
        break;\
725
    case PIX_FMT_ARGB:\
726
    case PIX_FMT_ABGR:\
727
        if (CONFIG_SMALL) {\
728
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
729
            func(uint32_t,needAlpha)\
730
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
731
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
732
            }\
733
        } else {\
734
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
735
                func(uint32_t,1)\
736
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
737
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
738
                }\
739
            } else {\
740
                func(uint32_t,0)\
741
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
742
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
743
                }\
744
            }\
745
        }                \
746
        break;\
747
    case PIX_FMT_RGB24:\
748
        func(uint8_t,0)\
749
            ((uint8_t*)dest)[0]= r[Y1];\
750
            ((uint8_t*)dest)[1]= g[Y1];\
751
            ((uint8_t*)dest)[2]= b[Y1];\
752
            ((uint8_t*)dest)[3]= r[Y2];\
753
            ((uint8_t*)dest)[4]= g[Y2];\
754
            ((uint8_t*)dest)[5]= b[Y2];\
755
            dest+=6;\
756
        }\
757
        break;\
758
    case PIX_FMT_BGR24:\
759
        func(uint8_t,0)\
760
            ((uint8_t*)dest)[0]= b[Y1];\
761
            ((uint8_t*)dest)[1]= g[Y1];\
762
            ((uint8_t*)dest)[2]= r[Y1];\
763
            ((uint8_t*)dest)[3]= b[Y2];\
764
            ((uint8_t*)dest)[4]= g[Y2];\
765
            ((uint8_t*)dest)[5]= r[Y2];\
766
            dest+=6;\
767
        }\
768
        break;\
769
    case PIX_FMT_RGB565:\
770
    case PIX_FMT_BGR565:\
771
        {\
772
            const int dr1= dither_2x2_8[y&1    ][0];\
773
            const int dg1= dither_2x2_4[y&1    ][0];\
774
            const int db1= dither_2x2_8[(y&1)^1][0];\
775
            const int dr2= dither_2x2_8[y&1    ][1];\
776
            const int dg2= dither_2x2_4[y&1    ][1];\
777
            const int db2= dither_2x2_8[(y&1)^1][1];\
778
            func(uint16_t,0)\
779
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
780
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
781
            }\
782
        }\
783
        break;\
784
    case PIX_FMT_RGB555:\
785
    case PIX_FMT_BGR555:\
786
        {\
787
            const int dr1= dither_2x2_8[y&1    ][0];\
788
            const int dg1= dither_2x2_8[y&1    ][1];\
789
            const int db1= dither_2x2_8[(y&1)^1][0];\
790
            const int dr2= dither_2x2_8[y&1    ][1];\
791
            const int dg2= dither_2x2_8[y&1    ][0];\
792
            const int db2= dither_2x2_8[(y&1)^1][1];\
793
            func(uint16_t,0)\
794
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
795
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
796
            }\
797
        }\
798
        break;\
799
    case PIX_FMT_RGB8:\
800
    case PIX_FMT_BGR8:\
801
        {\
802
            const uint8_t * const d64= dither_8x8_73[y&7];\
803
            const uint8_t * const d32= dither_8x8_32[y&7];\
804
            func(uint8_t,0)\
805
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
806
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
807
            }\
808
        }\
809
        break;\
810
    case PIX_FMT_RGB4:\
811
    case PIX_FMT_BGR4:\
812
        {\
813
            const uint8_t * const d64= dither_8x8_73 [y&7];\
814
            const uint8_t * const d128=dither_8x8_220[y&7];\
815
            func(uint8_t,0)\
816
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
817
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
818
            }\
819
        }\
820
        break;\
821
    case PIX_FMT_RGB4_BYTE:\
822
    case PIX_FMT_BGR4_BYTE:\
823
        {\
824
            const uint8_t * const d64= dither_8x8_73 [y&7];\
825
            const uint8_t * const d128=dither_8x8_220[y&7];\
826
            func(uint8_t,0)\
827
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
828
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
829
            }\
830
        }\
831
        break;\
832
    case PIX_FMT_MONOBLACK:\
833
    case PIX_FMT_MONOWHITE:\
834
        {\
835
            func_monoblack\
836
        }\
837
        break;\
838
    case PIX_FMT_YUYV422:\
839
        func2\
840
            ((uint8_t*)dest)[2*i2+0]= Y1;\
841
            ((uint8_t*)dest)[2*i2+1]= U;\
842
            ((uint8_t*)dest)[2*i2+2]= Y2;\
843
            ((uint8_t*)dest)[2*i2+3]= V;\
844
        }                \
845
        break;\
846
    case PIX_FMT_UYVY422:\
847
        func2\
848
            ((uint8_t*)dest)[2*i2+0]= U;\
849
            ((uint8_t*)dest)[2*i2+1]= Y1;\
850
            ((uint8_t*)dest)[2*i2+2]= V;\
851
            ((uint8_t*)dest)[2*i2+3]= Y2;\
852
        }                \
853
        break;\
854
    case PIX_FMT_GRAY16BE:\
855
        func_g16\
856
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
857
            ((uint8_t*)dest)[2*i2+1]= Y1;\
858
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
859
            ((uint8_t*)dest)[2*i2+3]= Y2;\
860
        }                \
861
        break;\
862
    case PIX_FMT_GRAY16LE:\
863
        func_g16\
864
            ((uint8_t*)dest)[2*i2+0]= Y1;\
865
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
866
            ((uint8_t*)dest)[2*i2+2]= Y2;\
867
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
868
        }                \
869
        break;\
870
    }
871

    
872
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
873
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
874
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
875
{
876
    int i;
877
    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)
878
}
879

    
880
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
881
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
882
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
883
{
884
    int i;
885
    int step= c->dstFormatBpp/8;
886
    int aidx= 3;
887

    
888
    switch(c->dstFormat) {
889
    case PIX_FMT_ARGB:
890
        dest++;
891
        aidx= 0;
892
    case PIX_FMT_RGB24:
893
        aidx--;
894
    case PIX_FMT_RGBA:
895
        if (CONFIG_SMALL) {
896
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
897
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
898
                dest[aidx]= needAlpha ? A : 255;
899
                dest[0]= R>>22;
900
                dest[1]= G>>22;
901
                dest[2]= B>>22;
902
                dest+= step;
903
            }
904
        } else {
905
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
906
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
907
                    dest[aidx]= A;
908
                    dest[0]= R>>22;
909
                    dest[1]= G>>22;
910
                    dest[2]= B>>22;
911
                    dest+= step;
912
                }
913
            } else {
914
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
915
                    dest[aidx]= 255;
916
                    dest[0]= R>>22;
917
                    dest[1]= G>>22;
918
                    dest[2]= B>>22;
919
                    dest+= step;
920
                }
921
            }
922
        }
923
        break;
924
    case PIX_FMT_ABGR:
925
        dest++;
926
        aidx= 0;
927
    case PIX_FMT_BGR24:
928
        aidx--;
929
    case PIX_FMT_BGRA:
930
        if (CONFIG_SMALL) {
931
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
932
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
933
                dest[aidx]= needAlpha ? A : 255;
934
                dest[0]= B>>22;
935
                dest[1]= G>>22;
936
                dest[2]= R>>22;
937
                dest+= step;
938
            }
939
        } else {
940
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
941
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
942
                    dest[aidx]= A;
943
                    dest[0]= B>>22;
944
                    dest[1]= G>>22;
945
                    dest[2]= R>>22;
946
                    dest+= step;
947
                }
948
            } else {
949
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
950
                    dest[aidx]= 255;
951
                    dest[0]= B>>22;
952
                    dest[1]= G>>22;
953
                    dest[2]= R>>22;
954
                    dest+= step;
955
                }
956
            }
957
        }
958
        break;
959
    default:
960
        assert(0);
961
    }
962
}
963

    
964
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
965
{
966
    int i;
967
    uint8_t *ptr = plane + stride*y;
968
    for (i=0; i<height; i++) {
969
        memset(ptr, val, width);
970
        ptr += stride;
971
    }
972
}
973

    
974
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
975
                            uint32_t *unused)
976
{
977
    int i;
978
    for (i = 0; i < width; i++) {
979
        int r = src[i*6+0];
980
        int g = src[i*6+2];
981
        int b = src[i*6+4];
982

    
983
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
984
    }
985
}
986

    
987
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
988
                             const uint8_t *src1, const uint8_t *src2,
989
                             int width, uint32_t *unused)
990
{
991
    int i;
992
    assert(src1==src2);
993
    for (i = 0; i < width; i++) {
994
        int r = src1[6*i + 0];
995
        int g = src1[6*i + 2];
996
        int b = src1[6*i + 4];
997

    
998
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
999
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1000
    }
1001
}
1002

    
1003
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1004
                                  const uint8_t *src1, const uint8_t *src2,
1005
                                  int width, uint32_t *unused)
1006
{
1007
    int i;
1008
    assert(src1==src2);
1009
    for (i = 0; i < width; i++) {
1010
        int r= src1[12*i + 0] + src1[12*i + 6];
1011
        int g= src1[12*i + 2] + src1[12*i + 8];
1012
        int b= src1[12*i + 4] + src1[12*i + 10];
1013

    
1014
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1015
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1016
    }
1017
}
1018

    
1019
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1020
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1021
{\
1022
    int i;\
1023
    for (i=0; i<width; i++) {\
1024
        int b= (((const type*)src)[i]>>shb)&maskb;\
1025
        int g= (((const type*)src)[i]>>shg)&maskg;\
1026
        int r= (((const type*)src)[i]>>shr)&maskr;\
1027
\
1028
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1029
    }\
1030
}
1031

    
1032
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1033
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1034
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1035
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1036
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1037
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1038

    
1039
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1040
{
1041
    int i;
1042
    for (i=0; i<width; i++) {
1043
        dst[i]= src[4*i];
1044
    }
1045
}
1046

    
1047
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1048
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1049
{\
1050
    int i;\
1051
    for (i=0; i<width; i++) {\
1052
        int b= (((const type*)src)[i]&maskb)>>shb;\
1053
        int g= (((const type*)src)[i]&maskg)>>shg;\
1054
        int r= (((const type*)src)[i]&maskr)>>shr;\
1055
\
1056
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1057
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1058
    }\
1059
}\
1060
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1061
{\
1062
    int i;\
1063
    for (i=0; i<width; i++) {\
1064
        int pix0= ((const type*)src)[2*i+0];\
1065
        int pix1= ((const type*)src)[2*i+1];\
1066
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1067
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1068
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1069
        g&= maskg|(2*maskg);\
1070
\
1071
        g>>=shg;\
1072
\
1073
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1074
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1075
    }\
1076
}
1077

    
1078
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1079
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1080
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1081
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1082
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1083
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1084

    
1085
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1086
{
1087
    int i;
1088
    for (i=0; i<width; i++) {
1089
        int d= src[i];
1090

    
1091
        dst[i]= pal[d] & 0xFF;
1092
    }
1093
}
1094

    
1095
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1096
                           const uint8_t *src1, const uint8_t *src2,
1097
                           long width, uint32_t *pal)
1098
{
1099
    int i;
1100
    assert(src1 == src2);
1101
    for (i=0; i<width; i++) {
1102
        int p= pal[src1[i]];
1103

    
1104
        dstU[i]= p>>8;
1105
        dstV[i]= p>>16;
1106
    }
1107
}
1108

    
1109
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1110
{
1111
    int i, j;
1112
    for (i=0; i<width/8; i++) {
1113
        int d= ~src[i];
1114
        for(j=0; j<8; j++)
1115
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1116
    }
1117
}
1118

    
1119
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1120
{
1121
    int i, j;
1122
    for (i=0; i<width/8; i++) {
1123
        int d= src[i];
1124
        for(j=0; j<8; j++)
1125
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1126
    }
1127
}
1128

    
1129
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1130
//Plain C versions
1131
#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1132
#define COMPILE_C
1133
#endif
1134

    
1135
#if ARCH_PPC
1136
#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1137
#define COMPILE_ALTIVEC
1138
#endif
1139
#endif //ARCH_PPC
1140

    
1141
#if ARCH_X86
1142

    
1143
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1144
#define COMPILE_MMX
1145
#endif
1146

    
1147
#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1148
#define COMPILE_MMX2
1149
#endif
1150

    
1151
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1152
#define COMPILE_3DNOW
1153
#endif
1154
#endif //ARCH_X86
1155

    
1156
#define COMPILE_TEMPLATE_MMX 0
1157
#define COMPILE_TEMPLATE_MMX2 0
1158
#define COMPILE_TEMPLATE_AMD3DNOW 0
1159
#define COMPILE_TEMPLATE_ALTIVEC 0
1160

    
1161
#ifdef COMPILE_C
1162
#define RENAME(a) a ## _C
1163
#include "swscale_template.c"
1164
#endif
1165

    
1166
#ifdef COMPILE_ALTIVEC
1167
#undef RENAME
1168
#undef COMPILE_TEMPLATE_ALTIVEC
1169
#define COMPILE_TEMPLATE_ALTIVEC 1
1170
#define RENAME(a) a ## _altivec
1171
#include "swscale_template.c"
1172
#endif
1173

    
1174
#if ARCH_X86
1175

    
1176
//MMX versions
1177
#ifdef COMPILE_MMX
1178
#undef RENAME
1179
#undef COMPILE_TEMPLATE_MMX
1180
#undef COMPILE_TEMPLATE_MMX2
1181
#undef COMPILE_TEMPLATE_AMD3DNOW
1182
#define COMPILE_TEMPLATE_MMX 1
1183
#define COMPILE_TEMPLATE_MMX2 0
1184
#define COMPILE_TEMPLATE_AMD3DNOW 0
1185
#define RENAME(a) a ## _MMX
1186
#include "swscale_template.c"
1187
#endif
1188

    
1189
//MMX2 versions
1190
#ifdef COMPILE_MMX2
1191
#undef RENAME
1192
#undef COMPILE_TEMPLATE_MMX
1193
#undef COMPILE_TEMPLATE_MMX2
1194
#undef COMPILE_TEMPLATE_AMD3DNOW
1195
#define COMPILE_TEMPLATE_MMX 1
1196
#define COMPILE_TEMPLATE_MMX2 1
1197
#define COMPILE_TEMPLATE_AMD3DNOW 0
1198
#define RENAME(a) a ## _MMX2
1199
#include "swscale_template.c"
1200
#endif
1201

    
1202
//3DNOW versions
1203
#ifdef COMPILE_3DNOW
1204
#undef RENAME
1205
#undef COMPILE_TEMPLATE_MMX
1206
#undef COMPILE_TEMPLATE_MMX2
1207
#undef COMPILE_TEMPLATE_AMD3DNOW
1208
#define COMPILE_TEMPLATE_MMX 1
1209
#define COMPILE_TEMPLATE_MMX2 0
1210
#define COMPILE_TEMPLATE_AMD3DNOW 1
1211
#define RENAME(a) a ## _3DNow
1212
#include "swscale_template.c"
1213
#endif
1214

    
1215
#endif //ARCH_X86
1216

    
1217
SwsFunc ff_getSwsFunc(SwsContext *c)
1218
{
1219
#if CONFIG_RUNTIME_CPUDETECT
1220
    int flags = c->flags;
1221

    
1222
#if ARCH_X86 && CONFIG_GPL
1223
    // ordered per speed fastest first
1224
    if (flags & SWS_CPU_CAPS_MMX2) {
1225
        sws_init_swScale_MMX2(c);
1226
        return swScale_MMX2;
1227
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1228
        sws_init_swScale_3DNow(c);
1229
        return swScale_3DNow;
1230
    } else if (flags & SWS_CPU_CAPS_MMX) {
1231
        sws_init_swScale_MMX(c);
1232
        return swScale_MMX;
1233
    } else {
1234
        sws_init_swScale_C(c);
1235
        return swScale_C;
1236
    }
1237

    
1238
#else
1239
#if ARCH_PPC
1240
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1241
        sws_init_swScale_altivec(c);
1242
        return swScale_altivec;
1243
    } else {
1244
        sws_init_swScale_C(c);
1245
        return swScale_C;
1246
    }
1247
#endif
1248
    sws_init_swScale_C(c);
1249
    return swScale_C;
1250
#endif /* ARCH_X86 && CONFIG_GPL */
1251
#else //CONFIG_RUNTIME_CPUDETECT
1252
#if   COMPILE_TEMPLATE_MMX2
1253
    sws_init_swScale_MMX2(c);
1254
    return swScale_MMX2;
1255
#elif COMPILE_TEMPLATE_AMD3DNOW
1256
    sws_init_swScale_3DNow(c);
1257
    return swScale_3DNow;
1258
#elif COMPILE_TEMPLATE_MMX
1259
    sws_init_swScale_MMX(c);
1260
    return swScale_MMX;
1261
#elif COMPILE_TEMPLATE_ALTIVEC
1262
    sws_init_swScale_altivec(c);
1263
    return swScale_altivec;
1264
#else
1265
    sws_init_swScale_C(c);
1266
    return swScale_C;
1267
#endif
1268
#endif //!CONFIG_RUNTIME_CPUDETECT
1269
}
1270

    
1271
static int PlanarToNV12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1272
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1273
{
1274
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1275
    /* Copy Y plane */
1276
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1277
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1278
    else {
1279
        int i;
1280
        const uint8_t *srcPtr= src[0];
1281
        uint8_t *dstPtr= dst;
1282
        for (i=0; i<srcSliceH; i++) {
1283
            memcpy(dstPtr, srcPtr, c->srcW);
1284
            srcPtr+= srcStride[0];
1285
            dstPtr+= dstStride[0];
1286
        }
1287
    }
1288
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1289
    if (c->dstFormat == PIX_FMT_NV12)
1290
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1291
    else
1292
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1293

    
1294
    return srcSliceH;
1295
}
1296

    
1297
static int PlanarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1298
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1299
{
1300
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1301

    
1302
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1303

    
1304
    return srcSliceH;
1305
}
1306

    
1307
static int PlanarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1308
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1309
{
1310
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1311

    
1312
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1313

    
1314
    return srcSliceH;
1315
}
1316

    
1317
static int YUV422PToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1318
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1319
{
1320
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1321

    
1322
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1323

    
1324
    return srcSliceH;
1325
}
1326

    
1327
static int YUV422PToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1328
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1329
{
1330
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1331

    
1332
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1333

    
1334
    return srcSliceH;
1335
}
1336

    
1337
static int YUYV2YUV420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1338
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1339
{
1340
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1341
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1342
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1343

    
1344
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1345

    
1346
    if (dstParam[3])
1347
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1348

    
1349
    return srcSliceH;
1350
}
1351

    
1352
static int YUYV2YUV422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1353
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1354
{
1355
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1356
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1357
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1358

    
1359
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1360

    
1361
    return srcSliceH;
1362
}
1363

    
1364
static int UYVY2YUV420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1365
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1366
{
1367
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1368
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1369
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1370

    
1371
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1372

    
1373
    if (dstParam[3])
1374
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1375

    
1376
    return srcSliceH;
1377
}
1378

    
1379
static int UYVY2YUV422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1380
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1381
{
1382
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1383
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1384
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1385

    
1386
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1387

    
1388
    return srcSliceH;
1389
}
1390

    
1391
static int pal2rgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1392
                          int srcSliceH, uint8_t* dst[], int dstStride[])
1393
{
1394
    const enum PixelFormat srcFormat= c->srcFormat;
1395
    const enum PixelFormat dstFormat= c->dstFormat;
1396
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1397
                 const uint8_t *palette)=NULL;
1398
    int i;
1399
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1400
    const uint8_t *srcPtr= src[0];
1401

    
1402
    if (usePal(srcFormat)) {
1403
        switch (dstFormat) {
1404
        case PIX_FMT_RGB32  : conv = palette8topacked32; break;
1405
        case PIX_FMT_BGR32  : conv = palette8topacked32; break;
1406
        case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1407
        case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1408
        case PIX_FMT_RGB24  : conv = palette8topacked24; break;
1409
        case PIX_FMT_BGR24  : conv = palette8topacked24; break;
1410
        }
1411
    }
1412

    
1413
    if (!conv)
1414
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1415
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1416
    else {
1417
        for (i=0; i<srcSliceH; i++) {
1418
            conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1419
            srcPtr+= srcStride[0];
1420
            dstPtr+= dstStride[0];
1421
        }
1422
    }
1423

    
1424
    return srcSliceH;
1425
}
1426

    
1427
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1428
static int rgb2rgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1429
                          int srcSliceH, uint8_t* dst[], int dstStride[])
1430
{
1431
    const enum PixelFormat srcFormat= c->srcFormat;
1432
    const enum PixelFormat dstFormat= c->dstFormat;
1433
    const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1434
    const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1435
    const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1436
    const int dstId= c->dstFormatBpp >> 2;
1437
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1438

    
1439
    /* BGR -> BGR */
1440
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
1441
       || (isRGB(srcFormat) && isRGB(dstFormat))) {
1442
        switch(srcId | (dstId<<4)) {
1443
        case 0x34: conv= rgb16to15; break;
1444
        case 0x36: conv= rgb24to15; break;
1445
        case 0x38: conv= rgb32to15; break;
1446
        case 0x43: conv= rgb15to16; break;
1447
        case 0x46: conv= rgb24to16; break;
1448
        case 0x48: conv= rgb32to16; break;
1449
        case 0x63: conv= rgb15to24; break;
1450
        case 0x64: conv= rgb16to24; break;
1451
        case 0x68: conv= rgb32to24; break;
1452
        case 0x83: conv= rgb15to32; break;
1453
        case 0x84: conv= rgb16to32; break;
1454
        case 0x86: conv= rgb24to32; break;
1455
        }
1456
    } else if (  (isBGR(srcFormat) && isRGB(dstFormat))
1457
             || (isRGB(srcFormat) && isBGR(dstFormat))) {
1458
        switch(srcId | (dstId<<4)) {
1459
        case 0x33: conv= rgb15tobgr15; break;
1460
        case 0x34: conv= rgb16tobgr15; break;
1461
        case 0x36: conv= rgb24tobgr15; break;
1462
        case 0x38: conv= rgb32tobgr15; break;
1463
        case 0x43: conv= rgb15tobgr16; break;
1464
        case 0x44: conv= rgb16tobgr16; break;
1465
        case 0x46: conv= rgb24tobgr16; break;
1466
        case 0x48: conv= rgb32tobgr16; break;
1467
        case 0x63: conv= rgb15tobgr24; break;
1468
        case 0x64: conv= rgb16tobgr24; break;
1469
        case 0x66: conv= rgb24tobgr24; break;
1470
        case 0x68: conv= rgb32tobgr24; break;
1471
        case 0x83: conv= rgb15tobgr32; break;
1472
        case 0x84: conv= rgb16tobgr32; break;
1473
        case 0x86: conv= rgb24tobgr32; break;
1474
        case 0x88: conv= rgb32tobgr32; break;
1475
        }
1476
    }
1477

    
1478
    if (!conv) {
1479
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1480
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1481
    } else {
1482
        const uint8_t *srcPtr= src[0];
1483
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
1484
            srcPtr += ALT32_CORR;
1485

    
1486
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1487
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1488
        else {
1489
            int i;
1490
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1491

    
1492
            for (i=0; i<srcSliceH; i++) {
1493
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
1494
                srcPtr+= srcStride[0];
1495
                dstPtr+= dstStride[0];
1496
            }
1497
        }
1498
    }
1499
    return srcSliceH;
1500
}
1501

    
1502
static int bgr24toyv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1503
                              int srcSliceH, uint8_t* dst[], int dstStride[])
1504
{
1505
    rgb24toyv12(
1506
        src[0],
1507
        dst[0]+ srcSliceY    *dstStride[0],
1508
        dst[1]+(srcSliceY>>1)*dstStride[1],
1509
        dst[2]+(srcSliceY>>1)*dstStride[2],
1510
        c->srcW, srcSliceH,
1511
        dstStride[0], dstStride[1], srcStride[0]);
1512
    if (dst[3])
1513
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1514
    return srcSliceH;
1515
}
1516

    
1517
static int yvu9toyv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1518
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1519
{
1520
    int i;
1521

    
1522
    /* copy Y */
1523
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1524
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1525
    else {
1526
        const uint8_t *srcPtr= src[0];
1527
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1528

    
1529
        for (i=0; i<srcSliceH; i++) {
1530
            memcpy(dstPtr, srcPtr, c->srcW);
1531
            srcPtr+= srcStride[0];
1532
            dstPtr+= dstStride[0];
1533
        }
1534
    }
1535

    
1536
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
1537
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1538
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
1539
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1540
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
1541
    } else {
1542
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1543
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
1544
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1545
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
1546
    }
1547
    if (dst[3])
1548
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1549
    return srcSliceH;
1550
}
1551

    
1552
/* unscaled copy like stuff (assumes nearly identical formats) */
1553
static int packedCopy(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1554
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1555
{
1556
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1557
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1558
    else {
1559
        int i;
1560
        const uint8_t *srcPtr= src[0];
1561
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1562
        int length=0;
1563

    
1564
        /* universal length finder */
1565
        while(length+c->srcW <= FFABS(dstStride[0])
1566
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1567
        assert(length!=0);
1568

    
1569
        for (i=0; i<srcSliceH; i++) {
1570
            memcpy(dstPtr, srcPtr, length);
1571
            srcPtr+= srcStride[0];
1572
            dstPtr+= dstStride[0];
1573
        }
1574
    }
1575
    return srcSliceH;
1576
}
1577

    
1578
static int planarCopy(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1579
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1580
{
1581
    int plane, i, j;
1582
    for (plane=0; plane<4; plane++) {
1583
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1584
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1585
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1586
        const uint8_t *srcPtr= src[plane];
1587
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1588

    
1589
        if (!dst[plane]) continue;
1590
        // ignore palette for GRAY8
1591
        if (plane == 1 && !dst[2]) continue;
1592
        if (!src[plane] || (plane == 1 && !src[2])) {
1593
            if(is16BPS(c->dstFormat))
1594
                length*=2;
1595
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1596
        } else {
1597
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1598
                if (!isBE(c->srcFormat)) srcPtr++;
1599
                for (i=0; i<height; i++) {
1600
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1601
                    srcPtr+= srcStride[plane];
1602
                    dstPtr+= dstStride[plane];
1603
                }
1604
            } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1605
                for (i=0; i<height; i++) {
1606
                    for (j=0; j<length; j++) {
1607
                        dstPtr[ j<<1   ] = srcPtr[j];
1608
                        dstPtr[(j<<1)+1] = srcPtr[j];
1609
                    }
1610
                    srcPtr+= srcStride[plane];
1611
                    dstPtr+= dstStride[plane];
1612
                }
1613
            } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1614
                  && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1615

    
1616
                for (i=0; i<height; i++) {
1617
                    for (j=0; j<length; j++)
1618
                        ((uint16_t*)dstPtr)[j] = bswap_16(((const uint16_t*)srcPtr)[j]);
1619
                    srcPtr+= srcStride[plane];
1620
                    dstPtr+= dstStride[plane];
1621
                }
1622
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1623
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1624
            else {
1625
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1626
                    length*=2;
1627
                for (i=0; i<height; i++) {
1628
                    memcpy(dstPtr, srcPtr, length);
1629
                    srcPtr+= srcStride[plane];
1630
                    dstPtr+= dstStride[plane];
1631
                }
1632
            }
1633
        }
1634
    }
1635
    return srcSliceH;
1636
}
1637

    
1638
int ff_hardcodedcpuflags(void)
1639
{
1640
    int flags = 0;
1641
#if   COMPILE_TEMPLATE_MMX2
1642
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1643
#elif COMPILE_TEMPLATE_AMD3DNOW
1644
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1645
#elif COMPILE_TEMPLATE_MMX
1646
    flags |= SWS_CPU_CAPS_MMX;
1647
#elif COMPILE_TEMPLATE_ALTIVEC
1648
    flags |= SWS_CPU_CAPS_ALTIVEC;
1649
#elif ARCH_BFIN
1650
    flags |= SWS_CPU_CAPS_BFIN;
1651
#endif
1652
    return flags;
1653
}
1654

    
1655
void ff_get_unscaled_swscale(SwsContext *c)
1656
{
1657
    const enum PixelFormat srcFormat = c->srcFormat;
1658
    const enum PixelFormat dstFormat = c->dstFormat;
1659
    const int flags = c->flags;
1660
    const int dstH = c->dstH;
1661
    int needsDither;
1662

    
1663
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
1664
        &&  c->dstFormatBpp < 24
1665
        && (c->dstFormatBpp < c->srcFormatBpp || (!(isRGB(srcFormat) || isBGR(srcFormat))));
1666

    
1667
    /* yv12_to_nv12 */
1668
    if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1669
        c->swScale= PlanarToNV12Wrapper;
1670
    }
1671
    /* yuv2bgr */
1672
    if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
1673
        && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1674
        c->swScale= ff_yuv2rgb_get_func_ptr(c);
1675
    }
1676

    
1677
    if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1678
        c->swScale= yvu9toyv12Wrapper;
1679
    }
1680

    
1681
    /* bgr24toYV12 */
1682
    if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1683
        c->swScale= bgr24toyv12Wrapper;
1684

    
1685
    /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1686
    if (  (isBGR(srcFormat) || isRGB(srcFormat))
1687
        && (isBGR(dstFormat) || isRGB(dstFormat))
1688
        && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
1689
        && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
1690
        && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
1691
        && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
1692
        && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1693
        && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1694
        && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1695
        && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1696
                                          && dstFormat != PIX_FMT_RGB32_1
1697
                                          && dstFormat != PIX_FMT_BGR32_1
1698
        && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
1699
        && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
1700
        && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1701
        c->swScale= rgb2rgbWrapper;
1702

    
1703
    if ((usePal(srcFormat) && (
1704
        dstFormat == PIX_FMT_RGB32   ||
1705
        dstFormat == PIX_FMT_RGB32_1 ||
1706
        dstFormat == PIX_FMT_RGB24   ||
1707
        dstFormat == PIX_FMT_BGR32   ||
1708
        dstFormat == PIX_FMT_BGR32_1 ||
1709
        dstFormat == PIX_FMT_BGR24)))
1710
        c->swScale= pal2rgbWrapper;
1711

    
1712
    if (srcFormat == PIX_FMT_YUV422P) {
1713
        if (dstFormat == PIX_FMT_YUYV422)
1714
            c->swScale= YUV422PToYuy2Wrapper;
1715
        else if (dstFormat == PIX_FMT_UYVY422)
1716
            c->swScale= YUV422PToUyvyWrapper;
1717
    }
1718

    
1719
    /* LQ converters if -sws 0 or -sws 4*/
1720
    if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1721
        /* yv12_to_yuy2 */
1722
        if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1723
            if (dstFormat == PIX_FMT_YUYV422)
1724
                c->swScale= PlanarToYuy2Wrapper;
1725
            else if (dstFormat == PIX_FMT_UYVY422)
1726
                c->swScale= PlanarToUyvyWrapper;
1727
        }
1728
    }
1729
    if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1730
        c->swScale= YUYV2YUV420Wrapper;
1731
    if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1732
        c->swScale= UYVY2YUV420Wrapper;
1733
    if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1734
        c->swScale= YUYV2YUV422Wrapper;
1735
    if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1736
        c->swScale= UYVY2YUV422Wrapper;
1737

    
1738
#ifdef COMPILE_ALTIVEC
1739
    if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1740
        !(c->flags & SWS_BITEXACT) &&
1741
        srcFormat == PIX_FMT_YUV420P) {
1742
        // unscaled YV12 -> packed YUV, we want speed
1743
        if (dstFormat == PIX_FMT_YUYV422)
1744
            c->swScale= yv12toyuy2_unscaled_altivec;
1745
        else if (dstFormat == PIX_FMT_UYVY422)
1746
            c->swScale= yv12touyvy_unscaled_altivec;
1747
    }
1748
#endif
1749

    
1750
    /* simple copy */
1751
    if (  srcFormat == dstFormat
1752
        || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1753
        || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1754
        || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1755
        || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1756
        || (isGray(dstFormat) && isGray(srcFormat))
1757
        || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1758
            && c->chrDstHSubSample == c->chrSrcHSubSample
1759
            && c->chrDstVSubSample == c->chrSrcVSubSample
1760
            && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1761
            && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1762
    {
1763
        if (isPacked(c->srcFormat))
1764
            c->swScale= packedCopy;
1765
        else /* Planar YUV or gray */
1766
            c->swScale= planarCopy;
1767
    }
1768
#if ARCH_BFIN
1769
    if (flags & SWS_CPU_CAPS_BFIN)
1770
        ff_bfin_get_unscaled_swscale (c);
1771
#endif
1772
}
1773

    
1774
static void reset_ptr(const uint8_t* src[], int format)
1775
{
1776
    if(!isALPHA(format))
1777
        src[3]=NULL;
1778
    if(!isPlanarYUV(format)) {
1779
        src[3]=src[2]=NULL;
1780

    
1781
        if (!usePal(format))
1782
            src[1]= NULL;
1783
    }
1784
}
1785

    
1786
/**
1787
 * swscale wrapper, so we don't need to export the SwsContext.
1788
 * Assumes planar YUV to be in YUV order instead of YVU.
1789
 */
1790
int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1791
              int srcSliceH, uint8_t* const dst[], const int dstStride[])
1792
{
1793
    int i;
1794
    const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1795
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1796

    
1797
    // do not mess up sliceDir if we have a "trailing" 0-size slice
1798
    if (srcSliceH == 0)
1799
        return 0;
1800

    
1801
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1802
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1803
        return 0;
1804
    }
1805
    if (c->sliceDir == 0) {
1806
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1807
    }
1808

    
1809
    if (usePal(c->srcFormat)) {
1810
        for (i=0; i<256; i++) {
1811
            int p, r, g, b,y,u,v;
1812
            if(c->srcFormat == PIX_FMT_PAL8) {
1813
                p=((const uint32_t*)(src[1]))[i];
1814
                r= (p>>16)&0xFF;
1815
                g= (p>> 8)&0xFF;
1816
                b=  p     &0xFF;
1817
            } else if(c->srcFormat == PIX_FMT_RGB8) {
1818
                r= (i>>5    )*36;
1819
                g= ((i>>2)&7)*36;
1820
                b= (i&3     )*85;
1821
            } else if(c->srcFormat == PIX_FMT_BGR8) {
1822
                b= (i>>6    )*85;
1823
                g= ((i>>3)&7)*36;
1824
                r= (i&7     )*36;
1825
            } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1826
                r= (i>>3    )*255;
1827
                g= ((i>>1)&3)*85;
1828
                b= (i&1     )*255;
1829
            } else {
1830
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1831
                b= (i>>3    )*255;
1832
                g= ((i>>1)&3)*85;
1833
                r= (i&1     )*255;
1834
            }
1835
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1836
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1837
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1838
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
1839

    
1840
            switch(c->dstFormat) {
1841
            case PIX_FMT_BGR32:
1842
#if !HAVE_BIGENDIAN
1843
            case PIX_FMT_RGB24:
1844
#endif
1845
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
1846
                break;
1847
            case PIX_FMT_BGR32_1:
1848
#if HAVE_BIGENDIAN
1849
            case PIX_FMT_BGR24:
1850
#endif
1851
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1852
                break;
1853
            case PIX_FMT_RGB32_1:
1854
#if HAVE_BIGENDIAN
1855
            case PIX_FMT_RGB24:
1856
#endif
1857
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1858
                break;
1859
            case PIX_FMT_RGB32:
1860
#if !HAVE_BIGENDIAN
1861
            case PIX_FMT_BGR24:
1862
#endif
1863
            default:
1864
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
1865
            }
1866
        }
1867
    }
1868

    
1869
    // copy strides, so they can safely be modified
1870
    if (c->sliceDir == 1) {
1871
        // slices go from top to bottom
1872
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1873
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1874

    
1875
        reset_ptr(src2, c->srcFormat);
1876
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
1877

    
1878
        /* reset slice direction at end of frame */
1879
        if (srcSliceY + srcSliceH == c->srcH)
1880
            c->sliceDir = 0;
1881

    
1882
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
1883
    } else {
1884
        // slices go from bottom to top => we flip the image internally
1885
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
1886
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
1887

    
1888
        src2[0] += (srcSliceH-1)*srcStride[0];
1889
        if (!usePal(c->srcFormat))
1890
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
1891
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
1892
        src2[3] += (srcSliceH-1)*srcStride[3];
1893
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
1894
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
1895
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
1896
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
1897

    
1898
        reset_ptr(src2, c->srcFormat);
1899
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
1900

    
1901
        /* reset slice direction at end of frame */
1902
        if (!srcSliceY)
1903
            c->sliceDir = 0;
1904

    
1905
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
1906
    }
1907
}
1908

    
1909
#if LIBSWSCALE_VERSION_MAJOR < 1
1910
int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
1911
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1912
{
1913
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
1914
}
1915
#endif