Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ c4e2b826

History | View | Annotate | Download (69.6 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/avutil.h"
66
#include "libavutil/mathematics.h"
67
#include "libavutil/bswap.h"
68
#include "libavutil/pixdesc.h"
69

    
70
#undef MOVNTQ
71
#undef PAVGB
72

    
73
//#undef HAVE_MMX2
74
//#define HAVE_AMD3DNOW
75
//#undef HAVE_MMX
76
//#undef ARCH_X86
77
#define DITHER1XBPP
78

    
79
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
80

    
81
#define isPacked(x)         (       \
82
           (x)==PIX_FMT_PAL8        \
83
        || (x)==PIX_FMT_YUYV422     \
84
        || (x)==PIX_FMT_UYVY422     \
85
        || isAnyRGB(x)              \
86
    )
87

    
88
#define RGB2YUV_SHIFT 15
89
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
90
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
91
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
92
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
93
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
94
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
95
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
96
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
97
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
98

    
99
static const double rgb2yuv_table[8][9]={
100
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
101
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
102
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
103
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
104
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
105
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
106
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
107
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
108
};
109

    
110
/*
111
NOTES
112
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
113

114
TODO
115
more intelligent misalignment avoidance for the horizontal scaler
116
write special vertical cubic upscale version
117
optimize C code (YV12 / minmax)
118
add support for packed pixel YUV input & output
119
add support for Y8 output
120
optimize BGR24 & BGR32
121
add BGR4 output support
122
write special BGR->BGR scaler
123
*/
124

    
125
#if ARCH_X86
126
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
127
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
128
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
129
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
130
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
131
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
132
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
133
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
134

    
135
const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
136
        0x0103010301030103LL,
137
        0x0200020002000200LL,};
138

    
139
const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
140
        0x0602060206020602LL,
141
        0x0004000400040004LL,};
142

    
143
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
144
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
145
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
146
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
147
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
148
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
149

    
150
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
151
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
152
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
153

    
154
#ifdef FAST_BGR2YV12
155
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
156
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
157
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
158
#else
159
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
160
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
161
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
162
#endif /* FAST_BGR2YV12 */
163
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
164
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
165
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
166

    
167
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
168
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
169
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
170
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
171
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
172

    
173
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
174
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
175
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
176
};
177

    
178
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
179

    
180
#endif /* ARCH_X86 */
181

    
182
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
183
{  1,   3,   1,   3,   1,   3,   1,   3, },
184
{  2,   0,   2,   0,   2,   0,   2,   0, },
185
};
186

    
187
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
188
{  6,   2,   6,   2,   6,   2,   6,   2, },
189
{  0,   4,   0,   4,   0,   4,   0,   4, },
190
};
191

    
192
DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
193
{  8,   4,  11,   7,   8,   4,  11,   7, },
194
{  2,  14,   1,  13,   2,  14,   1,  13, },
195
{ 10,   6,   9,   5,  10,   6,   9,   5, },
196
{  0,  12,   3,  15,   0,  12,   3,  15, },
197
};
198

    
199
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
200
{ 17,   9,  23,  15,  16,   8,  22,  14, },
201
{  5,  29,   3,  27,   4,  28,   2,  26, },
202
{ 21,  13,  19,  11,  20,  12,  18,  10, },
203
{  0,  24,   6,  30,   1,  25,   7,  31, },
204
{ 16,   8,  22,  14,  17,   9,  23,  15, },
205
{  4,  28,   2,  26,   5,  29,   3,  27, },
206
{ 20,  12,  18,  10,  21,  13,  19,  11, },
207
{  1,  25,   7,  31,   0,  24,   6,  30, },
208
};
209

    
210
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
211
{  0,  55,  14,  68,   3,  58,  17,  72, },
212
{ 37,  18,  50,  32,  40,  22,  54,  35, },
213
{  9,  64,   5,  59,  13,  67,   8,  63, },
214
{ 46,  27,  41,  23,  49,  31,  44,  26, },
215
{  2,  57,  16,  71,   1,  56,  15,  70, },
216
{ 39,  21,  52,  34,  38,  19,  51,  33, },
217
{ 11,  66,   7,  62,  10,  65,   6,  60, },
218
{ 48,  30,  43,  25,  47,  29,  42,  24, },
219
};
220

    
221
#if 1
222
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
223
{117,  62, 158, 103, 113,  58, 155, 100, },
224
{ 34, 199,  21, 186,  31, 196,  17, 182, },
225
{144,  89, 131,  76, 141,  86, 127,  72, },
226
{  0, 165,  41, 206,  10, 175,  52, 217, },
227
{110,  55, 151,  96, 120,  65, 162, 107, },
228
{ 28, 193,  14, 179,  38, 203,  24, 189, },
229
{138,  83, 124,  69, 148,  93, 134,  79, },
230
{  7, 172,  48, 213,   3, 168,  45, 210, },
231
};
232
#elif 1
233
// tries to correct a gamma of 1.5
234
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
235
{  0, 143,  18, 200,   2, 156,  25, 215, },
236
{ 78,  28, 125,  64,  89,  36, 138,  74, },
237
{ 10, 180,   3, 161,  16, 195,   8, 175, },
238
{109,  51,  93,  38, 121,  60, 105,  47, },
239
{  1, 152,  23, 210,   0, 147,  20, 205, },
240
{ 85,  33, 134,  71,  81,  30, 130,  67, },
241
{ 14, 190,   6, 171,  12, 185,   5, 166, },
242
{117,  57, 101,  44, 113,  54,  97,  41, },
243
};
244
#elif 1
245
// tries to correct a gamma of 2.0
246
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
247
{  0, 124,   8, 193,   0, 140,  12, 213, },
248
{ 55,  14, 104,  42,  66,  19, 119,  52, },
249
{  3, 168,   1, 145,   6, 187,   3, 162, },
250
{ 86,  31,  70,  21,  99,  39,  82,  28, },
251
{  0, 134,  11, 206,   0, 129,   9, 200, },
252
{ 62,  17, 114,  48,  58,  16, 109,  45, },
253
{  5, 181,   2, 157,   4, 175,   1, 151, },
254
{ 95,  36,  78,  26,  90,  34,  74,  24, },
255
};
256
#else
257
// tries to correct a gamma of 2.5
258
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
259
{  0, 107,   3, 187,   0, 125,   6, 212, },
260
{ 39,   7,  86,  28,  49,  11, 102,  36, },
261
{  1, 158,   0, 131,   3, 180,   1, 151, },
262
{ 68,  19,  52,  12,  81,  25,  64,  17, },
263
{  0, 119,   5, 203,   0, 113,   4, 195, },
264
{ 45,   9,  96,  33,  42,   8,  91,  30, },
265
{  2, 172,   1, 144,   2, 165,   0, 137, },
266
{ 77,  23,  60,  15,  72,  21,  56,  14, },
267
};
268
#endif
269

    
270
static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
271
                                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
272
                                                    const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
273
                                                    int dstW, int chrDstW, int big_endian)
274
{
275
    //FIXME Optimize (just quickly written not optimized..)
276
    int i;
277

    
278
    for (i = 0; i < dstW; i++) {
279
        int val = 1 << 10;
280
        int j;
281

    
282
        for (j = 0; j < lumFilterSize; j++)
283
            val += lumSrc[j][i] * lumFilter[j];
284

    
285
        if (big_endian) {
286
            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
287
        } else {
288
            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
289
        }
290
    }
291

    
292
    if (uDest) {
293
        for (i = 0; i < chrDstW; i++) {
294
            int u = 1 << 10;
295
            int v = 1 << 10;
296
            int j;
297

    
298
            for (j = 0; j < chrFilterSize; j++) {
299
                u += chrSrc[j][i       ] * chrFilter[j];
300
                v += chrSrc[j][i + VOFW] * chrFilter[j];
301
            }
302

    
303
            if (big_endian) {
304
                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
305
                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
306
            } else {
307
                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
308
                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
309
            }
310
        }
311
    }
312

    
313
    if (CONFIG_SWSCALE_ALPHA && aDest) {
314
        for (i = 0; i < dstW; i++) {
315
            int val = 1 << 10;
316
            int j;
317

    
318
            for (j = 0; j < lumFilterSize; j++)
319
                val += alpSrc[j][i] * lumFilter[j];
320

    
321
            if (big_endian) {
322
                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
323
            } else {
324
                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
325
            }
326
        }
327
    }
328
}
329

    
330
static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
331
                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
332
                                 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
333
                                 enum PixelFormat dstFormat)
334
{
335
    if (isBE(dstFormat)) {
336
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
337
                               chrFilter, chrSrc, chrFilterSize,
338
                               alpSrc,
339
                               dest, uDest, vDest, aDest,
340
                               dstW, chrDstW, 1);
341
    } else {
342
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
343
                               chrFilter, chrSrc, chrFilterSize,
344
                               alpSrc,
345
                               dest, uDest, vDest, aDest,
346
                               dstW, chrDstW, 0);
347
    }
348
}
349

    
350
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
351
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
352
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
353
{
354
    //FIXME Optimize (just quickly written not optimized..)
355
    int i;
356
    for (i=0; i<dstW; i++) {
357
        int val=1<<18;
358
        int j;
359
        for (j=0; j<lumFilterSize; j++)
360
            val += lumSrc[j][i] * lumFilter[j];
361

    
362
        dest[i]= av_clip_uint8(val>>19);
363
    }
364

    
365
    if (uDest)
366
        for (i=0; i<chrDstW; i++) {
367
            int u=1<<18;
368
            int v=1<<18;
369
            int j;
370
            for (j=0; j<chrFilterSize; j++) {
371
                u += chrSrc[j][i] * chrFilter[j];
372
                v += chrSrc[j][i + VOFW] * chrFilter[j];
373
            }
374

    
375
            uDest[i]= av_clip_uint8(u>>19);
376
            vDest[i]= av_clip_uint8(v>>19);
377
        }
378

    
379
    if (CONFIG_SWSCALE_ALPHA && aDest)
380
        for (i=0; i<dstW; i++) {
381
            int val=1<<18;
382
            int j;
383
            for (j=0; j<lumFilterSize; j++)
384
                val += alpSrc[j][i] * lumFilter[j];
385

    
386
            aDest[i]= av_clip_uint8(val>>19);
387
        }
388

    
389
}
390

    
391
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
392
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
393
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
394
{
395
    //FIXME Optimize (just quickly written not optimized..)
396
    int i;
397
    for (i=0; i<dstW; i++) {
398
        int val=1<<18;
399
        int j;
400
        for (j=0; j<lumFilterSize; j++)
401
            val += lumSrc[j][i] * lumFilter[j];
402

    
403
        dest[i]= av_clip_uint8(val>>19);
404
    }
405

    
406
    if (!uDest)
407
        return;
408

    
409
    if (dstFormat == PIX_FMT_NV12)
410
        for (i=0; i<chrDstW; i++) {
411
            int u=1<<18;
412
            int v=1<<18;
413
            int j;
414
            for (j=0; j<chrFilterSize; j++) {
415
                u += chrSrc[j][i] * chrFilter[j];
416
                v += chrSrc[j][i + VOFW] * chrFilter[j];
417
            }
418

    
419
            uDest[2*i]= av_clip_uint8(u>>19);
420
            uDest[2*i+1]= av_clip_uint8(v>>19);
421
        }
422
    else
423
        for (i=0; i<chrDstW; i++) {
424
            int u=1<<18;
425
            int v=1<<18;
426
            int j;
427
            for (j=0; j<chrFilterSize; j++) {
428
                u += chrSrc[j][i] * chrFilter[j];
429
                v += chrSrc[j][i + VOFW] * chrFilter[j];
430
            }
431

    
432
            uDest[2*i]= av_clip_uint8(v>>19);
433
            uDest[2*i+1]= av_clip_uint8(u>>19);
434
        }
435
}
436

    
437
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
438
    for (i=0; i<(dstW>>1); i++) {\
439
        int j;\
440
        int Y1 = 1<<18;\
441
        int Y2 = 1<<18;\
442
        int U  = 1<<18;\
443
        int V  = 1<<18;\
444
        int av_unused A1, A2;\
445
        type av_unused *r, *b, *g;\
446
        const int i2= 2*i;\
447
        \
448
        for (j=0; j<lumFilterSize; j++) {\
449
            Y1 += lumSrc[j][i2] * lumFilter[j];\
450
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
451
        }\
452
        for (j=0; j<chrFilterSize; j++) {\
453
            U += chrSrc[j][i] * chrFilter[j];\
454
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
455
        }\
456
        Y1>>=19;\
457
        Y2>>=19;\
458
        U >>=19;\
459
        V >>=19;\
460
        if (alpha) {\
461
            A1 = 1<<18;\
462
            A2 = 1<<18;\
463
            for (j=0; j<lumFilterSize; j++) {\
464
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
465
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
466
            }\
467
            A1>>=19;\
468
            A2>>=19;\
469
        }
470

    
471
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
472
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
473
        if ((Y1|Y2|U|V)&256) {\
474
            if (Y1>255)   Y1=255; \
475
            else if (Y1<0)Y1=0;   \
476
            if (Y2>255)   Y2=255; \
477
            else if (Y2<0)Y2=0;   \
478
            if (U>255)    U=255;  \
479
            else if (U<0) U=0;    \
480
            if (V>255)    V=255;  \
481
            else if (V<0) V=0;    \
482
        }\
483
        if (alpha && ((A1|A2)&256)) {\
484
            A1=av_clip_uint8(A1);\
485
            A2=av_clip_uint8(A2);\
486
        }
487

    
488
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
489
    for (i=0; i<dstW; i++) {\
490
        int j;\
491
        int Y = 0;\
492
        int U = -128<<19;\
493
        int V = -128<<19;\
494
        int av_unused A;\
495
        int R,G,B;\
496
        \
497
        for (j=0; j<lumFilterSize; j++) {\
498
            Y += lumSrc[j][i     ] * lumFilter[j];\
499
        }\
500
        for (j=0; j<chrFilterSize; j++) {\
501
            U += chrSrc[j][i     ] * chrFilter[j];\
502
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
503
        }\
504
        Y >>=10;\
505
        U >>=10;\
506
        V >>=10;\
507
        if (alpha) {\
508
            A = rnd;\
509
            for (j=0; j<lumFilterSize; j++)\
510
                A += alpSrc[j][i     ] * lumFilter[j];\
511
            A >>=19;\
512
            if (A&256)\
513
                A = av_clip_uint8(A);\
514
        }
515

    
516
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
517
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
518
        Y-= c->yuv2rgb_y_offset;\
519
        Y*= c->yuv2rgb_y_coeff;\
520
        Y+= rnd;\
521
        R= Y + V*c->yuv2rgb_v2r_coeff;\
522
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
523
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
524
        if ((R|G|B)&(0xC0000000)) {\
525
            if (R>=(256<<22))   R=(256<<22)-1; \
526
            else if (R<0)R=0;   \
527
            if (G>=(256<<22))   G=(256<<22)-1; \
528
            else if (G<0)G=0;   \
529
            if (B>=(256<<22))   B=(256<<22)-1; \
530
            else if (B<0)B=0;   \
531
        }
532

    
533
#define YSCALE_YUV_2_GRAY16_C \
534
    for (i=0; i<(dstW>>1); i++) {\
535
        int j;\
536
        int Y1 = 1<<18;\
537
        int Y2 = 1<<18;\
538
        int U  = 1<<18;\
539
        int V  = 1<<18;\
540
        \
541
        const int i2= 2*i;\
542
        \
543
        for (j=0; j<lumFilterSize; j++) {\
544
            Y1 += lumSrc[j][i2] * lumFilter[j];\
545
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
546
        }\
547
        Y1>>=11;\
548
        Y2>>=11;\
549
        if ((Y1|Y2|U|V)&65536) {\
550
            if (Y1>65535)   Y1=65535; \
551
            else if (Y1<0)Y1=0;   \
552
            if (Y2>65535)   Y2=65535; \
553
            else if (Y2<0)Y2=0;   \
554
        }
555

    
556
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
557
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
558
    r = (type *)c->table_rV[V];   \
559
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
560
    b = (type *)c->table_bU[U];
561

    
562
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
563
    for (i=0; i<(dstW>>1); i++) { \
564
        const int i2= 2*i;       \
565
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
566
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
567
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
568
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
569
        type av_unused *r, *b, *g;                                    \
570
        int av_unused A1, A2;                                         \
571
        if (alpha) {\
572
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
573
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
574
        }
575

    
576
#define YSCALE_YUV_2_GRAY16_2_C   \
577
    for (i=0; i<(dstW>>1); i++) { \
578
        const int i2= 2*i;       \
579
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
580
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
581

    
582
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
583
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
584
    r = (type *)c->table_rV[V];\
585
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
586
    b = (type *)c->table_bU[U];
587

    
588
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
589
    for (i=0; i<(dstW>>1); i++) {\
590
        const int i2= 2*i;\
591
        int Y1= buf0[i2  ]>>7;\
592
        int Y2= buf0[i2+1]>>7;\
593
        int U= (uvbuf1[i     ])>>7;\
594
        int V= (uvbuf1[i+VOFW])>>7;\
595
        type av_unused *r, *b, *g;\
596
        int av_unused A1, A2;\
597
        if (alpha) {\
598
            A1= abuf0[i2  ]>>7;\
599
            A2= abuf0[i2+1]>>7;\
600
        }
601

    
602
#define YSCALE_YUV_2_GRAY16_1_C \
603
    for (i=0; i<(dstW>>1); i++) {\
604
        const int i2= 2*i;\
605
        int Y1= buf0[i2  ]<<1;\
606
        int Y2= buf0[i2+1]<<1;
607

    
608
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
609
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
610
    r = (type *)c->table_rV[V];\
611
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
612
    b = (type *)c->table_bU[U];
613

    
614
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
615
    for (i=0; i<(dstW>>1); i++) {\
616
        const int i2= 2*i;\
617
        int Y1= buf0[i2  ]>>7;\
618
        int Y2= buf0[i2+1]>>7;\
619
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
620
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
621
        type av_unused *r, *b, *g;\
622
        int av_unused A1, A2;\
623
        if (alpha) {\
624
            A1= abuf0[i2  ]>>7;\
625
            A2= abuf0[i2+1]>>7;\
626
        }
627

    
628
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
629
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
630
    r = (type *)c->table_rV[V];\
631
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
632
    b = (type *)c->table_bU[U];
633

    
634
#define YSCALE_YUV_2_MONO2_C \
635
    const uint8_t * const d128=dither_8x8_220[y&7];\
636
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
637
    for (i=0; i<dstW-7; i+=8) {\
638
        int acc;\
639
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
640
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
641
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
642
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
643
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
644
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
645
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
646
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
647
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
648
        dest++;\
649
    }
650

    
651
#define YSCALE_YUV_2_MONOX_C \
652
    const uint8_t * const d128=dither_8x8_220[y&7];\
653
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
654
    int acc=0;\
655
    for (i=0; i<dstW-1; i+=2) {\
656
        int j;\
657
        int Y1=1<<18;\
658
        int Y2=1<<18;\
659
\
660
        for (j=0; j<lumFilterSize; j++) {\
661
            Y1 += lumSrc[j][i] * lumFilter[j];\
662
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
663
        }\
664
        Y1>>=19;\
665
        Y2>>=19;\
666
        if ((Y1|Y2)&256) {\
667
            if (Y1>255)   Y1=255;\
668
            else if (Y1<0)Y1=0;\
669
            if (Y2>255)   Y2=255;\
670
            else if (Y2<0)Y2=0;\
671
        }\
672
        acc+= acc + g[Y1+d128[(i+0)&7]];\
673
        acc+= acc + g[Y2+d128[(i+1)&7]];\
674
        if ((i&7)==6) {\
675
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
676
            dest++;\
677
        }\
678
    }
679

    
680
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
681
    switch(c->dstFormat) {\
682
    case PIX_FMT_RGB48BE:\
683
    case PIX_FMT_RGB48LE:\
684
        func(uint8_t,0)\
685
            ((uint8_t*)dest)[ 0]= r[Y1];\
686
            ((uint8_t*)dest)[ 1]= r[Y1];\
687
            ((uint8_t*)dest)[ 2]= g[Y1];\
688
            ((uint8_t*)dest)[ 3]= g[Y1];\
689
            ((uint8_t*)dest)[ 4]= b[Y1];\
690
            ((uint8_t*)dest)[ 5]= b[Y1];\
691
            ((uint8_t*)dest)[ 6]= r[Y2];\
692
            ((uint8_t*)dest)[ 7]= r[Y2];\
693
            ((uint8_t*)dest)[ 8]= g[Y2];\
694
            ((uint8_t*)dest)[ 9]= g[Y2];\
695
            ((uint8_t*)dest)[10]= b[Y2];\
696
            ((uint8_t*)dest)[11]= b[Y2];\
697
            dest+=12;\
698
        }\
699
        break;\
700
    case PIX_FMT_RGBA:\
701
    case PIX_FMT_BGRA:\
702
        if (CONFIG_SMALL) {\
703
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
704
            func(uint32_t,needAlpha)\
705
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
706
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
707
            }\
708
        } else {\
709
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
710
                func(uint32_t,1)\
711
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
712
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
713
                }\
714
            } else {\
715
                func(uint32_t,0)\
716
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
717
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
718
                }\
719
            }\
720
        }\
721
        break;\
722
    case PIX_FMT_ARGB:\
723
    case PIX_FMT_ABGR:\
724
        if (CONFIG_SMALL) {\
725
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
726
            func(uint32_t,needAlpha)\
727
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
728
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
729
            }\
730
        } else {\
731
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
732
                func(uint32_t,1)\
733
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
734
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
735
                }\
736
            } else {\
737
                func(uint32_t,0)\
738
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
739
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
740
                }\
741
            }\
742
        }                \
743
        break;\
744
    case PIX_FMT_RGB24:\
745
        func(uint8_t,0)\
746
            ((uint8_t*)dest)[0]= r[Y1];\
747
            ((uint8_t*)dest)[1]= g[Y1];\
748
            ((uint8_t*)dest)[2]= b[Y1];\
749
            ((uint8_t*)dest)[3]= r[Y2];\
750
            ((uint8_t*)dest)[4]= g[Y2];\
751
            ((uint8_t*)dest)[5]= b[Y2];\
752
            dest+=6;\
753
        }\
754
        break;\
755
    case PIX_FMT_BGR24:\
756
        func(uint8_t,0)\
757
            ((uint8_t*)dest)[0]= b[Y1];\
758
            ((uint8_t*)dest)[1]= g[Y1];\
759
            ((uint8_t*)dest)[2]= r[Y1];\
760
            ((uint8_t*)dest)[3]= b[Y2];\
761
            ((uint8_t*)dest)[4]= g[Y2];\
762
            ((uint8_t*)dest)[5]= r[Y2];\
763
            dest+=6;\
764
        }\
765
        break;\
766
    case PIX_FMT_RGB565BE:\
767
    case PIX_FMT_RGB565LE:\
768
    case PIX_FMT_BGR565BE:\
769
    case PIX_FMT_BGR565LE:\
770
        {\
771
            const int dr1= dither_2x2_8[y&1    ][0];\
772
            const int dg1= dither_2x2_4[y&1    ][0];\
773
            const int db1= dither_2x2_8[(y&1)^1][0];\
774
            const int dr2= dither_2x2_8[y&1    ][1];\
775
            const int dg2= dither_2x2_4[y&1    ][1];\
776
            const int db2= dither_2x2_8[(y&1)^1][1];\
777
            func(uint16_t,0)\
778
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
779
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
780
            }\
781
        }\
782
        break;\
783
    case PIX_FMT_RGB555BE:\
784
    case PIX_FMT_RGB555LE:\
785
    case PIX_FMT_BGR555BE:\
786
    case PIX_FMT_BGR555LE:\
787
        {\
788
            const int dr1= dither_2x2_8[y&1    ][0];\
789
            const int dg1= dither_2x2_8[y&1    ][1];\
790
            const int db1= dither_2x2_8[(y&1)^1][0];\
791
            const int dr2= dither_2x2_8[y&1    ][1];\
792
            const int dg2= dither_2x2_8[y&1    ][0];\
793
            const int db2= dither_2x2_8[(y&1)^1][1];\
794
            func(uint16_t,0)\
795
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
796
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
797
            }\
798
        }\
799
        break;\
800
    case PIX_FMT_RGB444BE:\
801
    case PIX_FMT_RGB444LE:\
802
    case PIX_FMT_BGR444BE:\
803
    case PIX_FMT_BGR444LE:\
804
        {\
805
            const int dr1= dither_4x4_16[y&3    ][0];\
806
            const int dg1= dither_4x4_16[y&3    ][1];\
807
            const int db1= dither_4x4_16[(y&3)^3][0];\
808
            const int dr2= dither_4x4_16[y&3    ][1];\
809
            const int dg2= dither_4x4_16[y&3    ][0];\
810
            const int db2= dither_4x4_16[(y&3)^3][1];\
811
            func(uint16_t,0)\
812
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
813
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
814
            }\
815
        }\
816
        break;\
817
    case PIX_FMT_RGB8:\
818
    case PIX_FMT_BGR8:\
819
        {\
820
            const uint8_t * const d64= dither_8x8_73[y&7];\
821
            const uint8_t * const d32= dither_8x8_32[y&7];\
822
            func(uint8_t,0)\
823
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
824
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
825
            }\
826
        }\
827
        break;\
828
    case PIX_FMT_RGB4:\
829
    case PIX_FMT_BGR4:\
830
        {\
831
            const uint8_t * const d64= dither_8x8_73 [y&7];\
832
            const uint8_t * const d128=dither_8x8_220[y&7];\
833
            func(uint8_t,0)\
834
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
835
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
836
            }\
837
        }\
838
        break;\
839
    case PIX_FMT_RGB4_BYTE:\
840
    case PIX_FMT_BGR4_BYTE:\
841
        {\
842
            const uint8_t * const d64= dither_8x8_73 [y&7];\
843
            const uint8_t * const d128=dither_8x8_220[y&7];\
844
            func(uint8_t,0)\
845
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
846
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
847
            }\
848
        }\
849
        break;\
850
    case PIX_FMT_MONOBLACK:\
851
    case PIX_FMT_MONOWHITE:\
852
        {\
853
            func_monoblack\
854
        }\
855
        break;\
856
    case PIX_FMT_YUYV422:\
857
        func2\
858
            ((uint8_t*)dest)[2*i2+0]= Y1;\
859
            ((uint8_t*)dest)[2*i2+1]= U;\
860
            ((uint8_t*)dest)[2*i2+2]= Y2;\
861
            ((uint8_t*)dest)[2*i2+3]= V;\
862
        }                \
863
        break;\
864
    case PIX_FMT_UYVY422:\
865
        func2\
866
            ((uint8_t*)dest)[2*i2+0]= U;\
867
            ((uint8_t*)dest)[2*i2+1]= Y1;\
868
            ((uint8_t*)dest)[2*i2+2]= V;\
869
            ((uint8_t*)dest)[2*i2+3]= Y2;\
870
        }                \
871
        break;\
872
    case PIX_FMT_GRAY16BE:\
873
        func_g16\
874
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
875
            ((uint8_t*)dest)[2*i2+1]= Y1;\
876
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
877
            ((uint8_t*)dest)[2*i2+3]= Y2;\
878
        }                \
879
        break;\
880
    case PIX_FMT_GRAY16LE:\
881
        func_g16\
882
            ((uint8_t*)dest)[2*i2+0]= Y1;\
883
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
884
            ((uint8_t*)dest)[2*i2+2]= Y2;\
885
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
886
        }                \
887
        break;\
888
    }
889

    
890
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
891
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
892
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
893
{
894
    int i;
895
    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)
896
}
897

    
898
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
899
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
900
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
901
{
902
    int i;
903
    int step= c->dstFormatBpp/8;
904
    int aidx= 3;
905

    
906
    switch(c->dstFormat) {
907
    case PIX_FMT_ARGB:
908
        dest++;
909
        aidx= 0;
910
    case PIX_FMT_RGB24:
911
        aidx--;
912
    case PIX_FMT_RGBA:
913
        if (CONFIG_SMALL) {
914
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
915
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
916
                dest[aidx]= needAlpha ? A : 255;
917
                dest[0]= R>>22;
918
                dest[1]= G>>22;
919
                dest[2]= B>>22;
920
                dest+= step;
921
            }
922
        } else {
923
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
924
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
925
                    dest[aidx]= A;
926
                    dest[0]= R>>22;
927
                    dest[1]= G>>22;
928
                    dest[2]= B>>22;
929
                    dest+= step;
930
                }
931
            } else {
932
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
933
                    dest[aidx]= 255;
934
                    dest[0]= R>>22;
935
                    dest[1]= G>>22;
936
                    dest[2]= B>>22;
937
                    dest+= step;
938
                }
939
            }
940
        }
941
        break;
942
    case PIX_FMT_ABGR:
943
        dest++;
944
        aidx= 0;
945
    case PIX_FMT_BGR24:
946
        aidx--;
947
    case PIX_FMT_BGRA:
948
        if (CONFIG_SMALL) {
949
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
950
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
951
                dest[aidx]= needAlpha ? A : 255;
952
                dest[0]= B>>22;
953
                dest[1]= G>>22;
954
                dest[2]= R>>22;
955
                dest+= step;
956
            }
957
        } else {
958
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
959
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
960
                    dest[aidx]= A;
961
                    dest[0]= B>>22;
962
                    dest[1]= G>>22;
963
                    dest[2]= R>>22;
964
                    dest+= step;
965
                }
966
            } else {
967
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
968
                    dest[aidx]= 255;
969
                    dest[0]= B>>22;
970
                    dest[1]= G>>22;
971
                    dest[2]= R>>22;
972
                    dest+= step;
973
                }
974
            }
975
        }
976
        break;
977
    default:
978
        assert(0);
979
    }
980
}
981

    
982
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
983
{
984
    int i;
985
    uint8_t *ptr = plane + stride*y;
986
    for (i=0; i<height; i++) {
987
        memset(ptr, val, width);
988
        ptr += stride;
989
    }
990
}
991

    
992
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
993
                            uint32_t *unused)
994
{
995
    int i;
996
    for (i = 0; i < width; i++) {
997
        int r = src[i*6+0];
998
        int g = src[i*6+2];
999
        int b = src[i*6+4];
1000

    
1001
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1002
    }
1003
}
1004

    
1005
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1006
                             const uint8_t *src1, const uint8_t *src2,
1007
                             long width, uint32_t *unused)
1008
{
1009
    int i;
1010
    assert(src1==src2);
1011
    for (i = 0; i < width; i++) {
1012
        int r = src1[6*i + 0];
1013
        int g = src1[6*i + 2];
1014
        int b = src1[6*i + 4];
1015

    
1016
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1017
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1018
    }
1019
}
1020

    
1021
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1022
                                  const uint8_t *src1, const uint8_t *src2,
1023
                                  long width, uint32_t *unused)
1024
{
1025
    int i;
1026
    assert(src1==src2);
1027
    for (i = 0; i < width; i++) {
1028
        int r= src1[12*i + 0] + src1[12*i + 6];
1029
        int g= src1[12*i + 2] + src1[12*i + 8];
1030
        int b= src1[12*i + 4] + src1[12*i + 10];
1031

    
1032
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1033
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1034
    }
1035
}
1036

    
1037
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1038
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1039
{\
1040
    int i;\
1041
    for (i=0; i<width; i++) {\
1042
        int b= (((const type*)src)[i]>>shb)&maskb;\
1043
        int g= (((const type*)src)[i]>>shg)&maskg;\
1044
        int r= (((const type*)src)[i]>>shr)&maskr;\
1045
\
1046
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1047
    }\
1048
}
1049

    
1050
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1051
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1052
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1053
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1054
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1055
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1056

    
1057
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1058
{
1059
    int i;
1060
    for (i=0; i<width; i++) {
1061
        dst[i]= src[4*i];
1062
    }
1063
}
1064

    
1065
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1066
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1067
{\
1068
    int i;\
1069
    for (i=0; i<width; i++) {\
1070
        int b= (((const type*)src)[i]&maskb)>>shb;\
1071
        int g= (((const type*)src)[i]&maskg)>>shg;\
1072
        int r= (((const type*)src)[i]&maskr)>>shr;\
1073
\
1074
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1075
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1076
    }\
1077
}\
1078
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1079
{\
1080
    int i;\
1081
    for (i=0; i<width; i++) {\
1082
        int pix0= ((const type*)src)[2*i+0];\
1083
        int pix1= ((const type*)src)[2*i+1];\
1084
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1085
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1086
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1087
        g&= maskg|(2*maskg);\
1088
\
1089
        g>>=shg;\
1090
\
1091
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1092
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1093
    }\
1094
}
1095

    
1096
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1097
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1098
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1099
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1100
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1101
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1102

    
1103
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1104
{
1105
    int i;
1106
    for (i=0; i<width; i++) {
1107
        int d= src[i];
1108

    
1109
        dst[i]= pal[d] & 0xFF;
1110
    }
1111
}
1112

    
1113
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1114
                           const uint8_t *src1, const uint8_t *src2,
1115
                           long width, uint32_t *pal)
1116
{
1117
    int i;
1118
    assert(src1 == src2);
1119
    for (i=0; i<width; i++) {
1120
        int p= pal[src1[i]];
1121

    
1122
        dstU[i]= p>>8;
1123
        dstV[i]= p>>16;
1124
    }
1125
}
1126

    
1127
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1128
{
1129
    int i, j;
1130
    for (i=0; i<width/8; i++) {
1131
        int d= ~src[i];
1132
        for(j=0; j<8; j++)
1133
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1134
    }
1135
}
1136

    
1137
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1138
{
1139
    int i, j;
1140
    for (i=0; i<width/8; i++) {
1141
        int d= src[i];
1142
        for(j=0; j<8; j++)
1143
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1144
    }
1145
}
1146

    
1147
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1148
//Plain C versions
1149
#if (!HAVE_MMX && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1150
#define COMPILE_C
1151
#endif
1152

    
1153
#if ARCH_PPC
1154
#if HAVE_ALTIVEC
1155
#define COMPILE_ALTIVEC
1156
#endif
1157
#endif //ARCH_PPC
1158

    
1159
#if ARCH_X86
1160

    
1161
#if (HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT
1162
#define COMPILE_MMX
1163
#endif
1164

    
1165
#if HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT
1166
#define COMPILE_MMX2
1167
#endif
1168

    
1169
#if (HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT
1170
#define COMPILE_3DNOW
1171
#endif
1172
#endif //ARCH_X86
1173

    
1174
#define COMPILE_TEMPLATE_MMX 0
1175
#define COMPILE_TEMPLATE_MMX2 0
1176
#define COMPILE_TEMPLATE_AMD3DNOW 0
1177
#define COMPILE_TEMPLATE_ALTIVEC 0
1178

    
1179
#ifdef COMPILE_C
1180
#define RENAME(a) a ## _C
1181
#include "swscale_template.c"
1182
#endif
1183

    
1184
#ifdef COMPILE_ALTIVEC
1185
#undef RENAME
1186
#undef COMPILE_TEMPLATE_ALTIVEC
1187
#define COMPILE_TEMPLATE_ALTIVEC 1
1188
#define RENAME(a) a ## _altivec
1189
#include "swscale_template.c"
1190
#endif
1191

    
1192
#if ARCH_X86
1193

    
1194
//MMX versions
1195
#ifdef COMPILE_MMX
1196
#undef RENAME
1197
#undef COMPILE_TEMPLATE_MMX
1198
#undef COMPILE_TEMPLATE_MMX2
1199
#undef COMPILE_TEMPLATE_AMD3DNOW
1200
#define COMPILE_TEMPLATE_MMX 1
1201
#define COMPILE_TEMPLATE_MMX2 0
1202
#define COMPILE_TEMPLATE_AMD3DNOW 0
1203
#define RENAME(a) a ## _MMX
1204
#include "swscale_template.c"
1205
#endif
1206

    
1207
//MMX2 versions
1208
#ifdef COMPILE_MMX2
1209
#undef RENAME
1210
#undef COMPILE_TEMPLATE_MMX
1211
#undef COMPILE_TEMPLATE_MMX2
1212
#undef COMPILE_TEMPLATE_AMD3DNOW
1213
#define COMPILE_TEMPLATE_MMX 1
1214
#define COMPILE_TEMPLATE_MMX2 1
1215
#define COMPILE_TEMPLATE_AMD3DNOW 0
1216
#define RENAME(a) a ## _MMX2
1217
#include "swscale_template.c"
1218
#endif
1219

    
1220
//3DNOW versions
1221
#ifdef COMPILE_3DNOW
1222
#undef RENAME
1223
#undef COMPILE_TEMPLATE_MMX
1224
#undef COMPILE_TEMPLATE_MMX2
1225
#undef COMPILE_TEMPLATE_AMD3DNOW
1226
#define COMPILE_TEMPLATE_MMX 1
1227
#define COMPILE_TEMPLATE_MMX2 0
1228
#define COMPILE_TEMPLATE_AMD3DNOW 1
1229
#define RENAME(a) a ## _3DNow
1230
#include "swscale_template.c"
1231
#endif
1232

    
1233
#endif //ARCH_X86
1234

    
1235
SwsFunc ff_getSwsFunc(SwsContext *c)
1236
{
1237
#if CONFIG_RUNTIME_CPUDETECT
1238
    int flags = c->flags;
1239

    
1240
#if ARCH_X86
1241
    // ordered per speed fastest first
1242
    if (flags & SWS_CPU_CAPS_MMX2) {
1243
        sws_init_swScale_MMX2(c);
1244
        return swScale_MMX2;
1245
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1246
        sws_init_swScale_3DNow(c);
1247
        return swScale_3DNow;
1248
    } else if (flags & SWS_CPU_CAPS_MMX) {
1249
        sws_init_swScale_MMX(c);
1250
        return swScale_MMX;
1251
    } else {
1252
        sws_init_swScale_C(c);
1253
        return swScale_C;
1254
    }
1255

    
1256
#else
1257
#ifdef COMPILE_ALTIVEC
1258
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1259
        sws_init_swScale_altivec(c);
1260
        return swScale_altivec;
1261
    } else {
1262
        sws_init_swScale_C(c);
1263
        return swScale_C;
1264
    }
1265
#endif
1266
    sws_init_swScale_C(c);
1267
    return swScale_C;
1268
#endif /* ARCH_X86 */
1269
#else //CONFIG_RUNTIME_CPUDETECT
1270
#if   COMPILE_TEMPLATE_MMX2
1271
    sws_init_swScale_MMX2(c);
1272
    return swScale_MMX2;
1273
#elif COMPILE_TEMPLATE_AMD3DNOW
1274
    sws_init_swScale_3DNow(c);
1275
    return swScale_3DNow;
1276
#elif COMPILE_TEMPLATE_MMX
1277
    sws_init_swScale_MMX(c);
1278
    return swScale_MMX;
1279
#elif COMPILE_TEMPLATE_ALTIVEC
1280
    sws_init_swScale_altivec(c);
1281
    return swScale_altivec;
1282
#else
1283
    sws_init_swScale_C(c);
1284
    return swScale_C;
1285
#endif
1286
#endif //!CONFIG_RUNTIME_CPUDETECT
1287
}
1288

    
1289
static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1290
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1291
{
1292
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1293
    /* Copy Y plane */
1294
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1295
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1296
    else {
1297
        int i;
1298
        const uint8_t *srcPtr= src[0];
1299
        uint8_t *dstPtr= dst;
1300
        for (i=0; i<srcSliceH; i++) {
1301
            memcpy(dstPtr, srcPtr, c->srcW);
1302
            srcPtr+= srcStride[0];
1303
            dstPtr+= dstStride[0];
1304
        }
1305
    }
1306
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1307
    if (c->dstFormat == PIX_FMT_NV12)
1308
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1309
    else
1310
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1311

    
1312
    return srcSliceH;
1313
}
1314

    
1315
static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1316
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1317
{
1318
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1319

    
1320
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1321

    
1322
    return srcSliceH;
1323
}
1324

    
1325
static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1326
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1327
{
1328
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1329

    
1330
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1331

    
1332
    return srcSliceH;
1333
}
1334

    
1335
static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1336
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1337
{
1338
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1339

    
1340
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1341

    
1342
    return srcSliceH;
1343
}
1344

    
1345
static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1346
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1347
{
1348
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1349

    
1350
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1351

    
1352
    return srcSliceH;
1353
}
1354

    
1355
static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1356
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1357
{
1358
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1359
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1360
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1361

    
1362
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1363

    
1364
    if (dstParam[3])
1365
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1366

    
1367
    return srcSliceH;
1368
}
1369

    
1370
static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1371
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1372
{
1373
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1374
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1375
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1376

    
1377
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1378

    
1379
    return srcSliceH;
1380
}
1381

    
1382
static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1383
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1384
{
1385
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1386
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1387
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1388

    
1389
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1390

    
1391
    if (dstParam[3])
1392
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1393

    
1394
    return srcSliceH;
1395
}
1396

    
1397
static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1398
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1399
{
1400
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1401
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1402
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1403

    
1404
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1405

    
1406
    return srcSliceH;
1407
}
1408

    
1409
static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1410
                           int srcSliceH, uint8_t* dst[], int dstStride[])
1411
{
1412
    const enum PixelFormat srcFormat= c->srcFormat;
1413
    const enum PixelFormat dstFormat= c->dstFormat;
1414
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1415
                 const uint8_t *palette)=NULL;
1416
    int i;
1417
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1418
    const uint8_t *srcPtr= src[0];
1419

    
1420
    if (usePal(srcFormat)) {
1421
        switch (dstFormat) {
1422
        case PIX_FMT_RGB32  : conv = sws_convertPalette8ToPacked32; break;
1423
        case PIX_FMT_BGR32  : conv = sws_convertPalette8ToPacked32; break;
1424
        case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1425
        case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1426
        case PIX_FMT_RGB24  : conv = sws_convertPalette8ToPacked24; break;
1427
        case PIX_FMT_BGR24  : conv = sws_convertPalette8ToPacked24; break;
1428
        }
1429
    }
1430

    
1431
    if (!conv)
1432
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1433
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1434
    else {
1435
        for (i=0; i<srcSliceH; i++) {
1436
            conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1437
            srcPtr+= srcStride[0];
1438
            dstPtr+= dstStride[0];
1439
        }
1440
    }
1441

    
1442
    return srcSliceH;
1443
}
1444

    
1445
#define isRGBA32(x) (            \
1446
           (x) == PIX_FMT_ARGB   \
1447
        || (x) == PIX_FMT_RGBA   \
1448
        || (x) == PIX_FMT_BGRA   \
1449
        || (x) == PIX_FMT_ABGR   \
1450
        )
1451

    
1452
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1453
static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1454
                           int srcSliceH, uint8_t* dst[], int dstStride[])
1455
{
1456
    const enum PixelFormat srcFormat= c->srcFormat;
1457
    const enum PixelFormat dstFormat= c->dstFormat;
1458
    const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1459
    const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1460
    const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1461
    const int dstId= c->dstFormatBpp >> 2;
1462
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1463

    
1464
#define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1465

    
1466
    if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1467
        if (     CONV_IS(ABGR, RGBA)
1468
              || CONV_IS(ARGB, BGRA)
1469
              || CONV_IS(BGRA, ARGB)
1470
              || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1471
        else if (CONV_IS(ABGR, ARGB)
1472
              || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1473
        else if (CONV_IS(ABGR, BGRA)
1474
              || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1475
        else if (CONV_IS(BGRA, RGBA)
1476
              || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1477
        else if (CONV_IS(BGRA, ABGR)
1478
              || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1479
    } else
1480
    /* BGR -> BGR */
1481
    if (  (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1482
       || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1483
        switch(srcId | (dstId<<4)) {
1484
        case 0x34: conv= rgb16to15; break;
1485
        case 0x36: conv= rgb24to15; break;
1486
        case 0x38: conv= rgb32to15; break;
1487
        case 0x43: conv= rgb15to16; break;
1488
        case 0x46: conv= rgb24to16; break;
1489
        case 0x48: conv= rgb32to16; break;
1490
        case 0x63: conv= rgb15to24; break;
1491
        case 0x64: conv= rgb16to24; break;
1492
        case 0x68: conv= rgb32to24; break;
1493
        case 0x83: conv= rgb15to32; break;
1494
        case 0x84: conv= rgb16to32; break;
1495
        case 0x86: conv= rgb24to32; break;
1496
        }
1497
    } else if (  (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1498
             || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1499
        switch(srcId | (dstId<<4)) {
1500
        case 0x33: conv= rgb15tobgr15; break;
1501
        case 0x34: conv= rgb16tobgr15; break;
1502
        case 0x36: conv= rgb24tobgr15; break;
1503
        case 0x38: conv= rgb32tobgr15; break;
1504
        case 0x43: conv= rgb15tobgr16; break;
1505
        case 0x44: conv= rgb16tobgr16; break;
1506
        case 0x46: conv= rgb24tobgr16; break;
1507
        case 0x48: conv= rgb32tobgr16; break;
1508
        case 0x63: conv= rgb15tobgr24; break;
1509
        case 0x64: conv= rgb16tobgr24; break;
1510
        case 0x66: conv= rgb24tobgr24; break;
1511
        case 0x68: conv= rgb32tobgr24; break;
1512
        case 0x83: conv= rgb15tobgr32; break;
1513
        case 0x84: conv= rgb16tobgr32; break;
1514
        case 0x86: conv= rgb24tobgr32; break;
1515
        }
1516
    }
1517

    
1518
    if (!conv) {
1519
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1520
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1521
    } else {
1522
        const uint8_t *srcPtr= src[0];
1523
              uint8_t *dstPtr= dst[0];
1524
        if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1525
            srcPtr += ALT32_CORR;
1526

    
1527
        if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1528
            dstPtr += ALT32_CORR;
1529

    
1530
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1531
            conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1532
        else {
1533
            int i;
1534
            dstPtr += dstStride[0]*srcSliceY;
1535

    
1536
            for (i=0; i<srcSliceH; i++) {
1537
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
1538
                srcPtr+= srcStride[0];
1539
                dstPtr+= dstStride[0];
1540
            }
1541
        }
1542
    }
1543
    return srcSliceH;
1544
}
1545

    
1546
static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1547
                              int srcSliceH, uint8_t* dst[], int dstStride[])
1548
{
1549
    rgb24toyv12(
1550
        src[0],
1551
        dst[0]+ srcSliceY    *dstStride[0],
1552
        dst[1]+(srcSliceY>>1)*dstStride[1],
1553
        dst[2]+(srcSliceY>>1)*dstStride[2],
1554
        c->srcW, srcSliceH,
1555
        dstStride[0], dstStride[1], srcStride[0]);
1556
    if (dst[3])
1557
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1558
    return srcSliceH;
1559
}
1560

    
1561
static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1562
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1563
{
1564
    int i;
1565

    
1566
    /* copy Y */
1567
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1568
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1569
    else {
1570
        const uint8_t *srcPtr= src[0];
1571
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1572

    
1573
        for (i=0; i<srcSliceH; i++) {
1574
            memcpy(dstPtr, srcPtr, c->srcW);
1575
            srcPtr+= srcStride[0];
1576
            dstPtr+= dstStride[0];
1577
        }
1578
    }
1579

    
1580
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
1581
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1582
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
1583
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1584
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
1585
    } else {
1586
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1587
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
1588
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1589
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
1590
    }
1591
    if (dst[3])
1592
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1593
    return srcSliceH;
1594
}
1595

    
1596
/* unscaled copy like stuff (assumes nearly identical formats) */
1597
static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1598
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1599
{
1600
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1601
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1602
    else {
1603
        int i;
1604
        const uint8_t *srcPtr= src[0];
1605
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1606
        int length=0;
1607

    
1608
        /* universal length finder */
1609
        while(length+c->srcW <= FFABS(dstStride[0])
1610
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1611
        assert(length!=0);
1612

    
1613
        for (i=0; i<srcSliceH; i++) {
1614
            memcpy(dstPtr, srcPtr, length);
1615
            srcPtr+= srcStride[0];
1616
            dstPtr+= dstStride[0];
1617
        }
1618
    }
1619
    return srcSliceH;
1620
}
1621

    
1622
static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1623
                             int srcSliceH, uint8_t* dst[], int dstStride[])
1624
{
1625
    int plane, i, j;
1626
    for (plane=0; plane<4; plane++) {
1627
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1628
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1629
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1630
        const uint8_t *srcPtr= src[plane];
1631
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1632

    
1633
        if (!dst[plane]) continue;
1634
        // ignore palette for GRAY8
1635
        if (plane == 1 && !dst[2]) continue;
1636
        if (!src[plane] || (plane == 1 && !src[2])) {
1637
            if(is16BPS(c->dstFormat))
1638
                length*=2;
1639
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1640
        } else {
1641
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1642
                if (!isBE(c->srcFormat)) srcPtr++;
1643
                for (i=0; i<height; i++) {
1644
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1645
                    srcPtr+= srcStride[plane];
1646
                    dstPtr+= dstStride[plane];
1647
                }
1648
            } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1649
                for (i=0; i<height; i++) {
1650
                    for (j=0; j<length; j++) {
1651
                        dstPtr[ j<<1   ] = srcPtr[j];
1652
                        dstPtr[(j<<1)+1] = srcPtr[j];
1653
                    }
1654
                    srcPtr+= srcStride[plane];
1655
                    dstPtr+= dstStride[plane];
1656
                }
1657
            } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1658
                  && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1659

    
1660
                for (i=0; i<height; i++) {
1661
                    for (j=0; j<length; j++)
1662
                        ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1663
                    srcPtr+= srcStride[plane];
1664
                    dstPtr+= dstStride[plane];
1665
                }
1666
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0) {
1667
                if (height > 0)
1668
                    memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1669
                           (height - 1)*dstStride[plane] + length);
1670
            } else {
1671
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1672
                    length*=2;
1673
                for (i=0; i<height; i++) {
1674
                    memcpy(dstPtr, srcPtr, length);
1675
                    srcPtr+= srcStride[plane];
1676
                    dstPtr+= dstStride[plane];
1677
                }
1678
            }
1679
        }
1680
    }
1681
    return srcSliceH;
1682
}
1683

    
1684
int ff_hardcodedcpuflags(void)
1685
{
1686
    int flags = 0;
1687
#if   COMPILE_TEMPLATE_MMX2
1688
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1689
#elif COMPILE_TEMPLATE_AMD3DNOW
1690
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1691
#elif COMPILE_TEMPLATE_MMX
1692
    flags |= SWS_CPU_CAPS_MMX;
1693
#elif COMPILE_TEMPLATE_ALTIVEC
1694
    flags |= SWS_CPU_CAPS_ALTIVEC;
1695
#elif ARCH_BFIN
1696
    flags |= SWS_CPU_CAPS_BFIN;
1697
#endif
1698
    return flags;
1699
}
1700

    
1701
void ff_get_unscaled_swscale(SwsContext *c)
1702
{
1703
    const enum PixelFormat srcFormat = c->srcFormat;
1704
    const enum PixelFormat dstFormat = c->dstFormat;
1705
    const int flags = c->flags;
1706
    const int dstH = c->dstH;
1707
    int needsDither;
1708

    
1709
    needsDither= isAnyRGB(dstFormat)
1710
        &&  c->dstFormatBpp < 24
1711
        && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1712

    
1713
    /* yv12_to_nv12 */
1714
    if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1715
        c->swScale= planarToNv12Wrapper;
1716
    }
1717
    /* yuv2bgr */
1718
    if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1719
        && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1720
        c->swScale= ff_yuv2rgb_get_func_ptr(c);
1721
    }
1722

    
1723
    if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1724
        c->swScale= yvu9ToYv12Wrapper;
1725
    }
1726

    
1727
    /* bgr24toYV12 */
1728
    if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1729
        c->swScale= bgr24ToYv12Wrapper;
1730

    
1731
    /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1732
    if (   isAnyRGB(srcFormat)
1733
        && isAnyRGB(dstFormat)
1734
        && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
1735
        && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
1736
        && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
1737
        && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
1738
        && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1739
        && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1740
        && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1741
        && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1742
        && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
1743
        && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
1744
        && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1745
        c->swScale= rgbToRgbWrapper;
1746

    
1747
    if ((usePal(srcFormat) && (
1748
        dstFormat == PIX_FMT_RGB32   ||
1749
        dstFormat == PIX_FMT_RGB32_1 ||
1750
        dstFormat == PIX_FMT_RGB24   ||
1751
        dstFormat == PIX_FMT_BGR32   ||
1752
        dstFormat == PIX_FMT_BGR32_1 ||
1753
        dstFormat == PIX_FMT_BGR24)))
1754
        c->swScale= palToRgbWrapper;
1755

    
1756
    if (srcFormat == PIX_FMT_YUV422P) {
1757
        if (dstFormat == PIX_FMT_YUYV422)
1758
            c->swScale= yuv422pToYuy2Wrapper;
1759
        else if (dstFormat == PIX_FMT_UYVY422)
1760
            c->swScale= yuv422pToUyvyWrapper;
1761
    }
1762

    
1763
    /* LQ converters if -sws 0 or -sws 4*/
1764
    if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1765
        /* yv12_to_yuy2 */
1766
        if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1767
            if (dstFormat == PIX_FMT_YUYV422)
1768
                c->swScale= planarToYuy2Wrapper;
1769
            else if (dstFormat == PIX_FMT_UYVY422)
1770
                c->swScale= planarToUyvyWrapper;
1771
        }
1772
    }
1773
    if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1774
        c->swScale= yuyvToYuv420Wrapper;
1775
    if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1776
        c->swScale= uyvyToYuv420Wrapper;
1777
    if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1778
        c->swScale= yuyvToYuv422Wrapper;
1779
    if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1780
        c->swScale= uyvyToYuv422Wrapper;
1781

    
1782
#ifdef COMPILE_ALTIVEC
1783
    if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1784
        !(c->flags & SWS_BITEXACT) &&
1785
        srcFormat == PIX_FMT_YUV420P) {
1786
        // unscaled YV12 -> packed YUV, we want speed
1787
        if (dstFormat == PIX_FMT_YUYV422)
1788
            c->swScale= yv12toyuy2_unscaled_altivec;
1789
        else if (dstFormat == PIX_FMT_UYVY422)
1790
            c->swScale= yv12touyvy_unscaled_altivec;
1791
    }
1792
#endif
1793

    
1794
    /* simple copy */
1795
    if (  srcFormat == dstFormat
1796
        || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1797
        || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1798
        || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1799
        || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1800
        || (isGray(dstFormat) && isGray(srcFormat))
1801
        || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1802
            && c->chrDstHSubSample == c->chrSrcHSubSample
1803
            && c->chrDstVSubSample == c->chrSrcVSubSample
1804
            && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1805
            && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1806
    {
1807
        if (isPacked(c->srcFormat))
1808
            c->swScale= packedCopyWrapper;
1809
        else /* Planar YUV or gray */
1810
            c->swScale= planarCopyWrapper;
1811
    }
1812
#if ARCH_BFIN
1813
    if (flags & SWS_CPU_CAPS_BFIN)
1814
        ff_bfin_get_unscaled_swscale (c);
1815
#endif
1816
}
1817

    
1818
static void reset_ptr(const uint8_t* src[], int format)
1819
{
1820
    if(!isALPHA(format))
1821
        src[3]=NULL;
1822
    if(!isPlanarYUV(format)) {
1823
        src[3]=src[2]=NULL;
1824

    
1825
        if (!usePal(format))
1826
            src[1]= NULL;
1827
    }
1828
}
1829

    
1830
/**
1831
 * swscale wrapper, so we don't need to export the SwsContext.
1832
 * Assumes planar YUV to be in YUV order instead of YVU.
1833
 */
1834
int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1835
              int srcSliceH, uint8_t* const dst[], const int dstStride[])
1836
{
1837
    int i;
1838
    const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1839
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1840

    
1841
    // do not mess up sliceDir if we have a "trailing" 0-size slice
1842
    if (srcSliceH == 0)
1843
        return 0;
1844

    
1845
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1846
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1847
        return 0;
1848
    }
1849
    if (c->sliceDir == 0) {
1850
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1851
    }
1852

    
1853
    if (usePal(c->srcFormat)) {
1854
        for (i=0; i<256; i++) {
1855
            int p, r, g, b,y,u,v;
1856
            if(c->srcFormat == PIX_FMT_PAL8) {
1857
                p=((const uint32_t*)(src[1]))[i];
1858
                r= (p>>16)&0xFF;
1859
                g= (p>> 8)&0xFF;
1860
                b=  p     &0xFF;
1861
            } else if(c->srcFormat == PIX_FMT_RGB8) {
1862
                r= (i>>5    )*36;
1863
                g= ((i>>2)&7)*36;
1864
                b= (i&3     )*85;
1865
            } else if(c->srcFormat == PIX_FMT_BGR8) {
1866
                b= (i>>6    )*85;
1867
                g= ((i>>3)&7)*36;
1868
                r= (i&7     )*36;
1869
            } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1870
                r= (i>>3    )*255;
1871
                g= ((i>>1)&3)*85;
1872
                b= (i&1     )*255;
1873
            } else if(c->srcFormat == PIX_FMT_GRAY8) {
1874
                r = g = b = i;
1875
            } else {
1876
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1877
                b= (i>>3    )*255;
1878
                g= ((i>>1)&3)*85;
1879
                r= (i&1     )*255;
1880
            }
1881
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1882
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1883
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1884
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
1885

    
1886
            switch(c->dstFormat) {
1887
            case PIX_FMT_BGR32:
1888
#if !HAVE_BIGENDIAN
1889
            case PIX_FMT_RGB24:
1890
#endif
1891
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
1892
                break;
1893
            case PIX_FMT_BGR32_1:
1894
#if HAVE_BIGENDIAN
1895
            case PIX_FMT_BGR24:
1896
#endif
1897
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1898
                break;
1899
            case PIX_FMT_RGB32_1:
1900
#if HAVE_BIGENDIAN
1901
            case PIX_FMT_RGB24:
1902
#endif
1903
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1904
                break;
1905
            case PIX_FMT_RGB32:
1906
#if !HAVE_BIGENDIAN
1907
            case PIX_FMT_BGR24:
1908
#endif
1909
            default:
1910
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
1911
            }
1912
        }
1913
    }
1914

    
1915
    // copy strides, so they can safely be modified
1916
    if (c->sliceDir == 1) {
1917
        // slices go from top to bottom
1918
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1919
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1920

    
1921
        reset_ptr(src2, c->srcFormat);
1922
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
1923

    
1924
        /* reset slice direction at end of frame */
1925
        if (srcSliceY + srcSliceH == c->srcH)
1926
            c->sliceDir = 0;
1927

    
1928
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
1929
    } else {
1930
        // slices go from bottom to top => we flip the image internally
1931
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
1932
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
1933

    
1934
        src2[0] += (srcSliceH-1)*srcStride[0];
1935
        if (!usePal(c->srcFormat))
1936
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
1937
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
1938
        src2[3] += (srcSliceH-1)*srcStride[3];
1939
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
1940
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
1941
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
1942
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
1943

    
1944
        reset_ptr(src2, c->srcFormat);
1945
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
1946

    
1947
        /* reset slice direction at end of frame */
1948
        if (!srcSliceY)
1949
            c->sliceDir = 0;
1950

    
1951
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
1952
    }
1953
}
1954

    
1955
#if LIBSWSCALE_VERSION_MAJOR < 1
1956
int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
1957
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1958
{
1959
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
1960
}
1961
#endif
1962

    
1963
/* Convert the palette to the same packed 32-bit format as the palette */
1964
void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1965
{
1966
    long i;
1967

    
1968
    for (i=0; i<num_pixels; i++)
1969
        ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
1970
}
1971

    
1972
/* Palette format: ABCD -> dst format: ABC */
1973
void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1974
{
1975
    long i;
1976

    
1977
    for (i=0; i<num_pixels; i++) {
1978
        //FIXME slow?
1979
        dst[0]= palette[src[i]*4+0];
1980
        dst[1]= palette[src[i]*4+1];
1981
        dst[2]= palette[src[i]*4+2];
1982
        dst+= 3;
1983
    }
1984
}