Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 80853cfc

History | View | Annotate | Download (122 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
#define _SVID_SOURCE //needed for MAP_ANONYMOUS
58
#include <inttypes.h>
59
#include <string.h>
60
#include <math.h>
61
#include <stdio.h>
62
#include "config.h"
63
#include <assert.h>
64
#if HAVE_SYS_MMAN_H
65
#include <sys/mman.h>
66
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
67
#define MAP_ANONYMOUS MAP_ANON
68
#endif
69
#endif
70
#if HAVE_VIRTUALALLOC
71
#define WIN32_LEAN_AND_MEAN
72
#include <windows.h>
73
#endif
74
#include "swscale.h"
75
#include "swscale_internal.h"
76
#include "rgb2rgb.h"
77
#include "libavutil/intreadwrite.h"
78
#include "libavutil/x86_cpu.h"
79
#include "libavutil/avutil.h"
80
#include "libavutil/bswap.h"
81
#include "libavutil/pixdesc.h"
82

    
83
unsigned swscale_version(void)
84
{
85
    return LIBSWSCALE_VERSION_INT;
86
}
87

    
88
const char *swscale_configuration(void)
89
{
90
    return FFMPEG_CONFIGURATION;
91
}
92

    
93
const char *swscale_license(void)
94
{
95
#define LICENSE_PREFIX "libswscale license: "
96
    return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
97
}
98

    
99
#undef MOVNTQ
100
#undef PAVGB
101

    
102
//#undef HAVE_MMX2
103
//#define HAVE_AMD3DNOW
104
//#undef HAVE_MMX
105
//#undef ARCH_X86
106
#define DITHER1XBPP
107

    
108
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
109

    
110
#define RET 0xC3 //near return opcode for x86
111

    
112
#ifdef M_PI
113
#define PI M_PI
114
#else
115
#define PI 3.14159265358979323846
116
#endif
117

    
118
#define isSupportedIn(x)    (       \
119
           (x)==PIX_FMT_YUV420P     \
120
        || (x)==PIX_FMT_YUVA420P    \
121
        || (x)==PIX_FMT_YUYV422     \
122
        || (x)==PIX_FMT_UYVY422     \
123
        || (x)==PIX_FMT_RGB48BE     \
124
        || (x)==PIX_FMT_RGB48LE     \
125
        || (x)==PIX_FMT_RGB32       \
126
        || (x)==PIX_FMT_RGB32_1     \
127
        || (x)==PIX_FMT_BGR24       \
128
        || (x)==PIX_FMT_BGR565      \
129
        || (x)==PIX_FMT_BGR555      \
130
        || (x)==PIX_FMT_BGR32       \
131
        || (x)==PIX_FMT_BGR32_1     \
132
        || (x)==PIX_FMT_RGB24       \
133
        || (x)==PIX_FMT_RGB565      \
134
        || (x)==PIX_FMT_RGB555      \
135
        || (x)==PIX_FMT_GRAY8       \
136
        || (x)==PIX_FMT_YUV410P     \
137
        || (x)==PIX_FMT_YUV440P     \
138
        || (x)==PIX_FMT_NV12        \
139
        || (x)==PIX_FMT_NV21        \
140
        || (x)==PIX_FMT_GRAY16BE    \
141
        || (x)==PIX_FMT_GRAY16LE    \
142
        || (x)==PIX_FMT_YUV444P     \
143
        || (x)==PIX_FMT_YUV422P     \
144
        || (x)==PIX_FMT_YUV411P     \
145
        || (x)==PIX_FMT_PAL8        \
146
        || (x)==PIX_FMT_BGR8        \
147
        || (x)==PIX_FMT_RGB8        \
148
        || (x)==PIX_FMT_BGR4_BYTE   \
149
        || (x)==PIX_FMT_RGB4_BYTE   \
150
        || (x)==PIX_FMT_YUV440P     \
151
        || (x)==PIX_FMT_MONOWHITE   \
152
        || (x)==PIX_FMT_MONOBLACK   \
153
        || (x)==PIX_FMT_YUV420P16LE   \
154
        || (x)==PIX_FMT_YUV422P16LE   \
155
        || (x)==PIX_FMT_YUV444P16LE   \
156
        || (x)==PIX_FMT_YUV420P16BE   \
157
        || (x)==PIX_FMT_YUV422P16BE   \
158
        || (x)==PIX_FMT_YUV444P16BE   \
159
    )
160

    
161
int sws_isSupportedInput(enum PixelFormat pix_fmt)
162
{
163
    return isSupportedIn(pix_fmt);
164
}
165

    
166
#define isSupportedOut(x)   (       \
167
           (x)==PIX_FMT_YUV420P     \
168
        || (x)==PIX_FMT_YUVA420P    \
169
        || (x)==PIX_FMT_YUYV422     \
170
        || (x)==PIX_FMT_UYVY422     \
171
        || (x)==PIX_FMT_YUV444P     \
172
        || (x)==PIX_FMT_YUV422P     \
173
        || (x)==PIX_FMT_YUV411P     \
174
        || isRGB(x)                 \
175
        || isBGR(x)                 \
176
        || (x)==PIX_FMT_NV12        \
177
        || (x)==PIX_FMT_NV21        \
178
        || (x)==PIX_FMT_GRAY16BE    \
179
        || (x)==PIX_FMT_GRAY16LE    \
180
        || (x)==PIX_FMT_GRAY8       \
181
        || (x)==PIX_FMT_YUV410P     \
182
        || (x)==PIX_FMT_YUV440P     \
183
        || (x)==PIX_FMT_YUV420P16LE   \
184
        || (x)==PIX_FMT_YUV422P16LE   \
185
        || (x)==PIX_FMT_YUV444P16LE   \
186
        || (x)==PIX_FMT_YUV420P16BE   \
187
        || (x)==PIX_FMT_YUV422P16BE   \
188
        || (x)==PIX_FMT_YUV444P16BE   \
189
    )
190

    
191
int sws_isSupportedOutput(enum PixelFormat pix_fmt)
192
{
193
    return isSupportedOut(pix_fmt);
194
}
195

    
196
#define isPacked(x)         (       \
197
           (x)==PIX_FMT_PAL8        \
198
        || (x)==PIX_FMT_YUYV422     \
199
        || (x)==PIX_FMT_UYVY422     \
200
        || isRGB(x)                 \
201
        || isBGR(x)                 \
202
    )
203
#define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)
204

    
205
#define RGB2YUV_SHIFT 15
206
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
207
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
208
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
209
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
210
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
211
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
212
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
213
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
214
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
215

    
216
extern const int32_t ff_yuv2rgb_coeffs[8][4];
217

    
218
static const double rgb2yuv_table[8][9]={
219
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
220
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
221
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
222
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
223
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
224
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
225
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
226
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
227
};
228

    
229
/*
230
NOTES
231
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
232

233
TODO
234
more intelligent misalignment avoidance for the horizontal scaler
235
write special vertical cubic upscale version
236
optimize C code (YV12 / minmax)
237
add support for packed pixel YUV input & output
238
add support for Y8 output
239
optimize BGR24 & BGR32
240
add BGR4 output support
241
write special BGR->BGR scaler
242
*/
243

    
244
#if ARCH_X86 && CONFIG_GPL
245
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
246
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
247
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
248
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
249
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
250
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
251
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
252
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
253

    
254
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
255
        0x0103010301030103LL,
256
        0x0200020002000200LL,};
257

    
258
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
259
        0x0602060206020602LL,
260
        0x0004000400040004LL,};
261

    
262
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
263
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
264
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
265
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
266
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
267
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
268

    
269
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
270
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
271
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
272

    
273
#ifdef FAST_BGR2YV12
274
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
275
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
276
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
277
#else
278
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
279
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
280
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
281
#endif /* FAST_BGR2YV12 */
282
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
283
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
284
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
285

    
286
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
287
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
288
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
289
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
290
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
291

    
292
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
293
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
294
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
295
};
296

    
297
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
298

    
299
#endif /* ARCH_X86 && CONFIG_GPL */
300

    
301
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
302

    
303
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4[2][8])={
304
{  1,   3,   1,   3,   1,   3,   1,   3, },
305
{  2,   0,   2,   0,   2,   0,   2,   0, },
306
};
307

    
308
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8[2][8])={
309
{  6,   2,   6,   2,   6,   2,   6,   2, },
310
{  0,   4,   0,   4,   0,   4,   0,   4, },
311
};
312

    
313
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32[8][8])={
314
{ 17,   9,  23,  15,  16,   8,  22,  14, },
315
{  5,  29,   3,  27,   4,  28,   2,  26, },
316
{ 21,  13,  19,  11,  20,  12,  18,  10, },
317
{  0,  24,   6,  30,   1,  25,   7,  31, },
318
{ 16,   8,  22,  14,  17,   9,  23,  15, },
319
{  4,  28,   2,  26,   5,  29,   3,  27, },
320
{ 20,  12,  18,  10,  21,  13,  19,  11, },
321
{  1,  25,   7,  31,   0,  24,   6,  30, },
322
};
323

    
324
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73[8][8])={
325
{  0,  55,  14,  68,   3,  58,  17,  72, },
326
{ 37,  18,  50,  32,  40,  22,  54,  35, },
327
{  9,  64,   5,  59,  13,  67,   8,  63, },
328
{ 46,  27,  41,  23,  49,  31,  44,  26, },
329
{  2,  57,  16,  71,   1,  56,  15,  70, },
330
{ 39,  21,  52,  34,  38,  19,  51,  33, },
331
{ 11,  66,   7,  62,  10,  65,   6,  60, },
332
{ 48,  30,  43,  25,  47,  29,  42,  24, },
333
};
334

    
335
#if 1
336
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
337
{117,  62, 158, 103, 113,  58, 155, 100, },
338
{ 34, 199,  21, 186,  31, 196,  17, 182, },
339
{144,  89, 131,  76, 141,  86, 127,  72, },
340
{  0, 165,  41, 206,  10, 175,  52, 217, },
341
{110,  55, 151,  96, 120,  65, 162, 107, },
342
{ 28, 193,  14, 179,  38, 203,  24, 189, },
343
{138,  83, 124,  69, 148,  93, 134,  79, },
344
{  7, 172,  48, 213,   3, 168,  45, 210, },
345
};
346
#elif 1
347
// tries to correct a gamma of 1.5
348
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
349
{  0, 143,  18, 200,   2, 156,  25, 215, },
350
{ 78,  28, 125,  64,  89,  36, 138,  74, },
351
{ 10, 180,   3, 161,  16, 195,   8, 175, },
352
{109,  51,  93,  38, 121,  60, 105,  47, },
353
{  1, 152,  23, 210,   0, 147,  20, 205, },
354
{ 85,  33, 134,  71,  81,  30, 130,  67, },
355
{ 14, 190,   6, 171,  12, 185,   5, 166, },
356
{117,  57, 101,  44, 113,  54,  97,  41, },
357
};
358
#elif 1
359
// tries to correct a gamma of 2.0
360
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
361
{  0, 124,   8, 193,   0, 140,  12, 213, },
362
{ 55,  14, 104,  42,  66,  19, 119,  52, },
363
{  3, 168,   1, 145,   6, 187,   3, 162, },
364
{ 86,  31,  70,  21,  99,  39,  82,  28, },
365
{  0, 134,  11, 206,   0, 129,   9, 200, },
366
{ 62,  17, 114,  48,  58,  16, 109,  45, },
367
{  5, 181,   2, 157,   4, 175,   1, 151, },
368
{ 95,  36,  78,  26,  90,  34,  74,  24, },
369
};
370
#else
371
// tries to correct a gamma of 2.5
372
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
373
{  0, 107,   3, 187,   0, 125,   6, 212, },
374
{ 39,   7,  86,  28,  49,  11, 102,  36, },
375
{  1, 158,   0, 131,   3, 180,   1, 151, },
376
{ 68,  19,  52,  12,  81,  25,  64,  17, },
377
{  0, 119,   5, 203,   0, 113,   4, 195, },
378
{ 45,   9,  96,  33,  42,   8,  91,  30, },
379
{  2, 172,   1, 144,   2, 165,   0, 137, },
380
{ 77,  23,  60,  15,  72,  21,  56,  14, },
381
};
382
#endif
383

    
384
const char *sws_format_name(enum PixelFormat format)
385
{
386
    if ((unsigned)format < PIX_FMT_NB && av_pix_fmt_descriptors[format].name)
387
        return av_pix_fmt_descriptors[format].name;
388
    else
389
        return "Unknown format";
390
}
391

    
392
static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
393
                                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
394
                                                    const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
395
                                                    int dstW, int chrDstW, int big_endian)
396
{
397
    //FIXME Optimize (just quickly written not optimized..)
398
    int i;
399

    
400
    for (i = 0; i < dstW; i++) {
401
        int val = 1 << 10;
402
        int j;
403

    
404
        for (j = 0; j < lumFilterSize; j++)
405
            val += lumSrc[j][i] * lumFilter[j];
406

    
407
        if (big_endian) {
408
            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
409
        } else {
410
            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
411
        }
412
    }
413

    
414
    if (uDest) {
415
        for (i = 0; i < chrDstW; i++) {
416
            int u = 1 << 10;
417
            int v = 1 << 10;
418
            int j;
419

    
420
            for (j = 0; j < chrFilterSize; j++) {
421
                u += chrSrc[j][i       ] * chrFilter[j];
422
                v += chrSrc[j][i + VOFW] * chrFilter[j];
423
            }
424

    
425
            if (big_endian) {
426
                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
427
                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
428
            } else {
429
                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
430
                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
431
            }
432
        }
433
    }
434

    
435
    if (CONFIG_SWSCALE_ALPHA && aDest) {
436
        for (i = 0; i < dstW; i++) {
437
            int val = 1 << 10;
438
            int j;
439

    
440
            for (j = 0; j < lumFilterSize; j++)
441
                val += alpSrc[j][i] * lumFilter[j];
442

    
443
            if (big_endian) {
444
                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
445
            } else {
446
                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
447
            }
448
        }
449
    }
450
}
451

    
452
static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
453
                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
454
                                 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
455
                                 enum PixelFormat dstFormat)
456
{
457
    if (isBE(dstFormat)) {
458
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
459
                               chrFilter, chrSrc, chrFilterSize,
460
                               alpSrc,
461
                               dest, uDest, vDest, aDest,
462
                               dstW, chrDstW, 1);
463
    } else {
464
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
465
                               chrFilter, chrSrc, chrFilterSize,
466
                               alpSrc,
467
                               dest, uDest, vDest, aDest,
468
                               dstW, chrDstW, 0);
469
    }
470
}
471

    
472
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
473
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
474
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
475
{
476
    //FIXME Optimize (just quickly written not optimized..)
477
    int i;
478
    for (i=0; i<dstW; i++) {
479
        int val=1<<18;
480
        int j;
481
        for (j=0; j<lumFilterSize; j++)
482
            val += lumSrc[j][i] * lumFilter[j];
483

    
484
        dest[i]= av_clip_uint8(val>>19);
485
    }
486

    
487
    if (uDest)
488
        for (i=0; i<chrDstW; i++) {
489
            int u=1<<18;
490
            int v=1<<18;
491
            int j;
492
            for (j=0; j<chrFilterSize; j++) {
493
                u += chrSrc[j][i] * chrFilter[j];
494
                v += chrSrc[j][i + VOFW] * chrFilter[j];
495
            }
496

    
497
            uDest[i]= av_clip_uint8(u>>19);
498
            vDest[i]= av_clip_uint8(v>>19);
499
        }
500

    
501
    if (CONFIG_SWSCALE_ALPHA && aDest)
502
        for (i=0; i<dstW; i++) {
503
            int val=1<<18;
504
            int j;
505
            for (j=0; j<lumFilterSize; j++)
506
                val += alpSrc[j][i] * lumFilter[j];
507

    
508
            aDest[i]= av_clip_uint8(val>>19);
509
        }
510

    
511
}
512

    
513
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
514
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
515
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
516
{
517
    //FIXME Optimize (just quickly written not optimized..)
518
    int i;
519
    for (i=0; i<dstW; i++) {
520
        int val=1<<18;
521
        int j;
522
        for (j=0; j<lumFilterSize; j++)
523
            val += lumSrc[j][i] * lumFilter[j];
524

    
525
        dest[i]= av_clip_uint8(val>>19);
526
    }
527

    
528
    if (!uDest)
529
        return;
530

    
531
    if (dstFormat == PIX_FMT_NV12)
532
        for (i=0; i<chrDstW; i++) {
533
            int u=1<<18;
534
            int v=1<<18;
535
            int j;
536
            for (j=0; j<chrFilterSize; j++) {
537
                u += chrSrc[j][i] * chrFilter[j];
538
                v += chrSrc[j][i + VOFW] * chrFilter[j];
539
            }
540

    
541
            uDest[2*i]= av_clip_uint8(u>>19);
542
            uDest[2*i+1]= av_clip_uint8(v>>19);
543
        }
544
    else
545
        for (i=0; i<chrDstW; i++) {
546
            int u=1<<18;
547
            int v=1<<18;
548
            int j;
549
            for (j=0; j<chrFilterSize; j++) {
550
                u += chrSrc[j][i] * chrFilter[j];
551
                v += chrSrc[j][i + VOFW] * chrFilter[j];
552
            }
553

    
554
            uDest[2*i]= av_clip_uint8(v>>19);
555
            uDest[2*i+1]= av_clip_uint8(u>>19);
556
        }
557
}
558

    
559
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
560
    for (i=0; i<(dstW>>1); i++) {\
561
        int j;\
562
        int Y1 = 1<<18;\
563
        int Y2 = 1<<18;\
564
        int U  = 1<<18;\
565
        int V  = 1<<18;\
566
        int av_unused A1, A2;\
567
        type av_unused *r, *b, *g;\
568
        const int i2= 2*i;\
569
        \
570
        for (j=0; j<lumFilterSize; j++) {\
571
            Y1 += lumSrc[j][i2] * lumFilter[j];\
572
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
573
        }\
574
        for (j=0; j<chrFilterSize; j++) {\
575
            U += chrSrc[j][i] * chrFilter[j];\
576
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
577
        }\
578
        Y1>>=19;\
579
        Y2>>=19;\
580
        U >>=19;\
581
        V >>=19;\
582
        if (alpha) {\
583
            A1 = 1<<18;\
584
            A2 = 1<<18;\
585
            for (j=0; j<lumFilterSize; j++) {\
586
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
587
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
588
            }\
589
            A1>>=19;\
590
            A2>>=19;\
591
        }\
592

    
593
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
594
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
595
        if ((Y1|Y2|U|V)&256) {\
596
            if (Y1>255)   Y1=255; \
597
            else if (Y1<0)Y1=0;   \
598
            if (Y2>255)   Y2=255; \
599
            else if (Y2<0)Y2=0;   \
600
            if (U>255)    U=255;  \
601
            else if (U<0) U=0;    \
602
            if (V>255)    V=255;  \
603
            else if (V<0) V=0;    \
604
        }\
605
        if (alpha && ((A1|A2)&256)) {\
606
            A1=av_clip_uint8(A1);\
607
            A2=av_clip_uint8(A2);\
608
        }
609

    
610
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
611
    for (i=0; i<dstW; i++) {\
612
        int j;\
613
        int Y = 0;\
614
        int U = -128<<19;\
615
        int V = -128<<19;\
616
        int av_unused A;\
617
        int R,G,B;\
618
        \
619
        for (j=0; j<lumFilterSize; j++) {\
620
            Y += lumSrc[j][i     ] * lumFilter[j];\
621
        }\
622
        for (j=0; j<chrFilterSize; j++) {\
623
            U += chrSrc[j][i     ] * chrFilter[j];\
624
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
625
        }\
626
        Y >>=10;\
627
        U >>=10;\
628
        V >>=10;\
629
        if (alpha) {\
630
            A = rnd;\
631
            for (j=0; j<lumFilterSize; j++)\
632
                A += alpSrc[j][i     ] * lumFilter[j];\
633
            A >>=19;\
634
            if (A&256)\
635
                A = av_clip_uint8(A);\
636
        }\
637

    
638
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
639
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
640
        Y-= c->yuv2rgb_y_offset;\
641
        Y*= c->yuv2rgb_y_coeff;\
642
        Y+= rnd;\
643
        R= Y + V*c->yuv2rgb_v2r_coeff;\
644
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
645
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
646
        if ((R|G|B)&(0xC0000000)) {\
647
            if (R>=(256<<22))   R=(256<<22)-1; \
648
            else if (R<0)R=0;   \
649
            if (G>=(256<<22))   G=(256<<22)-1; \
650
            else if (G<0)G=0;   \
651
            if (B>=(256<<22))   B=(256<<22)-1; \
652
            else if (B<0)B=0;   \
653
        }\
654

    
655

    
656
#define YSCALE_YUV_2_GRAY16_C \
657
    for (i=0; i<(dstW>>1); i++) {\
658
        int j;\
659
        int Y1 = 1<<18;\
660
        int Y2 = 1<<18;\
661
        int U  = 1<<18;\
662
        int V  = 1<<18;\
663
        \
664
        const int i2= 2*i;\
665
        \
666
        for (j=0; j<lumFilterSize; j++) {\
667
            Y1 += lumSrc[j][i2] * lumFilter[j];\
668
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
669
        }\
670
        Y1>>=11;\
671
        Y2>>=11;\
672
        if ((Y1|Y2|U|V)&65536) {\
673
            if (Y1>65535)   Y1=65535; \
674
            else if (Y1<0)Y1=0;   \
675
            if (Y2>65535)   Y2=65535; \
676
            else if (Y2<0)Y2=0;   \
677
        }
678

    
679
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
680
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
681
    r = (type *)c->table_rV[V];   \
682
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
683
    b = (type *)c->table_bU[U];   \
684

    
685
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
686
    for (i=0; i<(dstW>>1); i++) { \
687
        const int i2= 2*i;       \
688
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
689
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
690
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
691
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
692
        type av_unused *r, *b, *g;                                    \
693
        int av_unused A1, A2;                                         \
694
        if (alpha) {\
695
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
696
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
697
        }\
698

    
699
#define YSCALE_YUV_2_GRAY16_2_C   \
700
    for (i=0; i<(dstW>>1); i++) { \
701
        const int i2= 2*i;       \
702
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
703
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
704

    
705
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
706
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
707
    r = (type *)c->table_rV[V];\
708
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
709
    b = (type *)c->table_bU[U];\
710

    
711
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
712
    for (i=0; i<(dstW>>1); i++) {\
713
        const int i2= 2*i;\
714
        int Y1= buf0[i2  ]>>7;\
715
        int Y2= buf0[i2+1]>>7;\
716
        int U= (uvbuf1[i     ])>>7;\
717
        int V= (uvbuf1[i+VOFW])>>7;\
718
        type av_unused *r, *b, *g;\
719
        int av_unused A1, A2;\
720
        if (alpha) {\
721
            A1= abuf0[i2  ]>>7;\
722
            A2= abuf0[i2+1]>>7;\
723
        }\
724

    
725
#define YSCALE_YUV_2_GRAY16_1_C \
726
    for (i=0; i<(dstW>>1); i++) {\
727
        const int i2= 2*i;\
728
        int Y1= buf0[i2  ]<<1;\
729
        int Y2= buf0[i2+1]<<1;\
730

    
731
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
732
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
733
    r = (type *)c->table_rV[V];\
734
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
735
    b = (type *)c->table_bU[U];\
736

    
737
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
738
    for (i=0; i<(dstW>>1); i++) {\
739
        const int i2= 2*i;\
740
        int Y1= buf0[i2  ]>>7;\
741
        int Y2= buf0[i2+1]>>7;\
742
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
743
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
744
        type av_unused *r, *b, *g;\
745
        int av_unused A1, A2;\
746
        if (alpha) {\
747
            A1= abuf0[i2  ]>>7;\
748
            A2= abuf0[i2+1]>>7;\
749
        }\
750

    
751
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
752
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
753
    r = (type *)c->table_rV[V];\
754
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
755
    b = (type *)c->table_bU[U];\
756

    
757
#define YSCALE_YUV_2_MONO2_C \
758
    const uint8_t * const d128=dither_8x8_220[y&7];\
759
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
760
    for (i=0; i<dstW-7; i+=8) {\
761
        int acc;\
762
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
763
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
764
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
765
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
766
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
767
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
768
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
769
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
770
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
771
        dest++;\
772
    }\
773

    
774

    
775
#define YSCALE_YUV_2_MONOX_C \
776
    const uint8_t * const d128=dither_8x8_220[y&7];\
777
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
778
    int acc=0;\
779
    for (i=0; i<dstW-1; i+=2) {\
780
        int j;\
781
        int Y1=1<<18;\
782
        int Y2=1<<18;\
783
\
784
        for (j=0; j<lumFilterSize; j++) {\
785
            Y1 += lumSrc[j][i] * lumFilter[j];\
786
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
787
        }\
788
        Y1>>=19;\
789
        Y2>>=19;\
790
        if ((Y1|Y2)&256) {\
791
            if (Y1>255)   Y1=255;\
792
            else if (Y1<0)Y1=0;\
793
            if (Y2>255)   Y2=255;\
794
            else if (Y2<0)Y2=0;\
795
        }\
796
        acc+= acc + g[Y1+d128[(i+0)&7]];\
797
        acc+= acc + g[Y2+d128[(i+1)&7]];\
798
        if ((i&7)==6) {\
799
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
800
            dest++;\
801
        }\
802
    }
803

    
804

    
805
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
806
    switch(c->dstFormat) {\
807
    case PIX_FMT_RGB48BE:\
808
    case PIX_FMT_RGB48LE:\
809
        func(uint8_t,0)\
810
            ((uint8_t*)dest)[ 0]= r[Y1];\
811
            ((uint8_t*)dest)[ 1]= r[Y1];\
812
            ((uint8_t*)dest)[ 2]= g[Y1];\
813
            ((uint8_t*)dest)[ 3]= g[Y1];\
814
            ((uint8_t*)dest)[ 4]= b[Y1];\
815
            ((uint8_t*)dest)[ 5]= b[Y1];\
816
            ((uint8_t*)dest)[ 6]= r[Y2];\
817
            ((uint8_t*)dest)[ 7]= r[Y2];\
818
            ((uint8_t*)dest)[ 8]= g[Y2];\
819
            ((uint8_t*)dest)[ 9]= g[Y2];\
820
            ((uint8_t*)dest)[10]= b[Y2];\
821
            ((uint8_t*)dest)[11]= b[Y2];\
822
            dest+=12;\
823
        }\
824
        break;\
825
    case PIX_FMT_RGBA:\
826
    case PIX_FMT_BGRA:\
827
        if (CONFIG_SMALL) {\
828
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
829
            func(uint32_t,needAlpha)\
830
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
831
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
832
            }\
833
        } else {\
834
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
835
                func(uint32_t,1)\
836
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
837
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
838
                }\
839
            } else {\
840
                func(uint32_t,0)\
841
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
842
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
843
                }\
844
            }\
845
        }\
846
        break;\
847
    case PIX_FMT_ARGB:\
848
    case PIX_FMT_ABGR:\
849
        if (CONFIG_SMALL) {\
850
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
851
            func(uint32_t,needAlpha)\
852
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
853
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
854
            }\
855
        } else {\
856
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
857
                func(uint32_t,1)\
858
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
859
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
860
                }\
861
            } else {\
862
                func(uint32_t,0)\
863
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
864
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
865
                }\
866
            }\
867
        }                \
868
        break;\
869
    case PIX_FMT_RGB24:\
870
        func(uint8_t,0)\
871
            ((uint8_t*)dest)[0]= r[Y1];\
872
            ((uint8_t*)dest)[1]= g[Y1];\
873
            ((uint8_t*)dest)[2]= b[Y1];\
874
            ((uint8_t*)dest)[3]= r[Y2];\
875
            ((uint8_t*)dest)[4]= g[Y2];\
876
            ((uint8_t*)dest)[5]= b[Y2];\
877
            dest+=6;\
878
        }\
879
        break;\
880
    case PIX_FMT_BGR24:\
881
        func(uint8_t,0)\
882
            ((uint8_t*)dest)[0]= b[Y1];\
883
            ((uint8_t*)dest)[1]= g[Y1];\
884
            ((uint8_t*)dest)[2]= r[Y1];\
885
            ((uint8_t*)dest)[3]= b[Y2];\
886
            ((uint8_t*)dest)[4]= g[Y2];\
887
            ((uint8_t*)dest)[5]= r[Y2];\
888
            dest+=6;\
889
        }\
890
        break;\
891
    case PIX_FMT_RGB565:\
892
    case PIX_FMT_BGR565:\
893
        {\
894
            const int dr1= dither_2x2_8[y&1    ][0];\
895
            const int dg1= dither_2x2_4[y&1    ][0];\
896
            const int db1= dither_2x2_8[(y&1)^1][0];\
897
            const int dr2= dither_2x2_8[y&1    ][1];\
898
            const int dg2= dither_2x2_4[y&1    ][1];\
899
            const int db2= dither_2x2_8[(y&1)^1][1];\
900
            func(uint16_t,0)\
901
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
902
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
903
            }\
904
        }\
905
        break;\
906
    case PIX_FMT_RGB555:\
907
    case PIX_FMT_BGR555:\
908
        {\
909
            const int dr1= dither_2x2_8[y&1    ][0];\
910
            const int dg1= dither_2x2_8[y&1    ][1];\
911
            const int db1= dither_2x2_8[(y&1)^1][0];\
912
            const int dr2= dither_2x2_8[y&1    ][1];\
913
            const int dg2= dither_2x2_8[y&1    ][0];\
914
            const int db2= dither_2x2_8[(y&1)^1][1];\
915
            func(uint16_t,0)\
916
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
917
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
918
            }\
919
        }\
920
        break;\
921
    case PIX_FMT_RGB8:\
922
    case PIX_FMT_BGR8:\
923
        {\
924
            const uint8_t * const d64= dither_8x8_73[y&7];\
925
            const uint8_t * const d32= dither_8x8_32[y&7];\
926
            func(uint8_t,0)\
927
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
928
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
929
            }\
930
        }\
931
        break;\
932
    case PIX_FMT_RGB4:\
933
    case PIX_FMT_BGR4:\
934
        {\
935
            const uint8_t * const d64= dither_8x8_73 [y&7];\
936
            const uint8_t * const d128=dither_8x8_220[y&7];\
937
            func(uint8_t,0)\
938
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
939
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
940
            }\
941
        }\
942
        break;\
943
    case PIX_FMT_RGB4_BYTE:\
944
    case PIX_FMT_BGR4_BYTE:\
945
        {\
946
            const uint8_t * const d64= dither_8x8_73 [y&7];\
947
            const uint8_t * const d128=dither_8x8_220[y&7];\
948
            func(uint8_t,0)\
949
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
950
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
951
            }\
952
        }\
953
        break;\
954
    case PIX_FMT_MONOBLACK:\
955
    case PIX_FMT_MONOWHITE:\
956
        {\
957
            func_monoblack\
958
        }\
959
        break;\
960
    case PIX_FMT_YUYV422:\
961
        func2\
962
            ((uint8_t*)dest)[2*i2+0]= Y1;\
963
            ((uint8_t*)dest)[2*i2+1]= U;\
964
            ((uint8_t*)dest)[2*i2+2]= Y2;\
965
            ((uint8_t*)dest)[2*i2+3]= V;\
966
        }                \
967
        break;\
968
    case PIX_FMT_UYVY422:\
969
        func2\
970
            ((uint8_t*)dest)[2*i2+0]= U;\
971
            ((uint8_t*)dest)[2*i2+1]= Y1;\
972
            ((uint8_t*)dest)[2*i2+2]= V;\
973
            ((uint8_t*)dest)[2*i2+3]= Y2;\
974
        }                \
975
        break;\
976
    case PIX_FMT_GRAY16BE:\
977
        func_g16\
978
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
979
            ((uint8_t*)dest)[2*i2+1]= Y1;\
980
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
981
            ((uint8_t*)dest)[2*i2+3]= Y2;\
982
        }                \
983
        break;\
984
    case PIX_FMT_GRAY16LE:\
985
        func_g16\
986
            ((uint8_t*)dest)[2*i2+0]= Y1;\
987
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
988
            ((uint8_t*)dest)[2*i2+2]= Y2;\
989
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
990
        }                \
991
        break;\
992
    }\
993

    
994

    
995
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
996
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
997
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
998
{
999
    int i;
1000
    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)
1001
}
1002

    
1003
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1004
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1005
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1006
{
1007
    int i;
1008
    int step= fmt_depth(c->dstFormat)/8;
1009
    int aidx= 3;
1010

    
1011
    switch(c->dstFormat) {
1012
    case PIX_FMT_ARGB:
1013
        dest++;
1014
        aidx= 0;
1015
    case PIX_FMT_RGB24:
1016
        aidx--;
1017
    case PIX_FMT_RGBA:
1018
        if (CONFIG_SMALL) {
1019
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1020
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1021
                dest[aidx]= needAlpha ? A : 255;
1022
                dest[0]= R>>22;
1023
                dest[1]= G>>22;
1024
                dest[2]= B>>22;
1025
                dest+= step;
1026
            }
1027
        } else {
1028
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1029
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1030
                    dest[aidx]= A;
1031
                    dest[0]= R>>22;
1032
                    dest[1]= G>>22;
1033
                    dest[2]= B>>22;
1034
                    dest+= step;
1035
                }
1036
            } else {
1037
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1038
                    dest[aidx]= 255;
1039
                    dest[0]= R>>22;
1040
                    dest[1]= G>>22;
1041
                    dest[2]= B>>22;
1042
                    dest+= step;
1043
                }
1044
            }
1045
        }
1046
        break;
1047
    case PIX_FMT_ABGR:
1048
        dest++;
1049
        aidx= 0;
1050
    case PIX_FMT_BGR24:
1051
        aidx--;
1052
    case PIX_FMT_BGRA:
1053
        if (CONFIG_SMALL) {
1054
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1055
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1056
                dest[aidx]= needAlpha ? A : 255;
1057
                dest[0]= B>>22;
1058
                dest[1]= G>>22;
1059
                dest[2]= R>>22;
1060
                dest+= step;
1061
            }
1062
        } else {
1063
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1064
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1065
                    dest[aidx]= A;
1066
                    dest[0]= B>>22;
1067
                    dest[1]= G>>22;
1068
                    dest[2]= R>>22;
1069
                    dest+= step;
1070
                }
1071
            } else {
1072
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1073
                    dest[aidx]= 255;
1074
                    dest[0]= B>>22;
1075
                    dest[1]= G>>22;
1076
                    dest[2]= R>>22;
1077
                    dest+= step;
1078
                }
1079
            }
1080
        }
1081
        break;
1082
    default:
1083
        assert(0);
1084
    }
1085
}
1086

    
1087
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
1088
{
1089
    int i;
1090
    uint8_t *ptr = plane + stride*y;
1091
    for (i=0; i<height; i++) {
1092
        memset(ptr, val, width);
1093
        ptr += stride;
1094
    }
1095
}
1096

    
1097
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
1098
                            uint32_t *unused)
1099
{
1100
    int i;
1101
    for (i = 0; i < width; i++) {
1102
        int r = src[i*6+0];
1103
        int g = src[i*6+2];
1104
        int b = src[i*6+4];
1105

    
1106
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1107
    }
1108
}
1109

    
1110
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1111
                             const uint8_t *src1, const uint8_t *src2,
1112
                             int width, uint32_t *unused)
1113
{
1114
    int i;
1115
    assert(src1==src2);
1116
    for (i = 0; i < width; i++) {
1117
        int r = src1[6*i + 0];
1118
        int g = src1[6*i + 2];
1119
        int b = src1[6*i + 4];
1120

    
1121
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1122
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1123
    }
1124
}
1125

    
1126
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1127
                                  const uint8_t *src1, const uint8_t *src2,
1128
                                  int width, uint32_t *unused)
1129
{
1130
    int i;
1131
    assert(src1==src2);
1132
    for (i = 0; i < width; i++) {
1133
        int r= src1[12*i + 0] + src1[12*i + 6];
1134
        int g= src1[12*i + 2] + src1[12*i + 8];
1135
        int b= src1[12*i + 4] + src1[12*i + 10];
1136

    
1137
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1138
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1139
    }
1140
}
1141

    
1142
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1143
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1144
{\
1145
    int i;\
1146
    for (i=0; i<width; i++) {\
1147
        int b= (((const type*)src)[i]>>shb)&maskb;\
1148
        int g= (((const type*)src)[i]>>shg)&maskg;\
1149
        int r= (((const type*)src)[i]>>shr)&maskr;\
1150
\
1151
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1152
    }\
1153
}
1154

    
1155
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1156
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1157
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1158
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1159
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1160
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1161

    
1162
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1163
{
1164
    int i;
1165
    for (i=0; i<width; i++) {
1166
        dst[i]= src[4*i];
1167
    }
1168
}
1169

    
1170
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1171
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1172
{\
1173
    int i;\
1174
    for (i=0; i<width; i++) {\
1175
        int b= (((const type*)src)[i]&maskb)>>shb;\
1176
        int g= (((const type*)src)[i]&maskg)>>shg;\
1177
        int r= (((const type*)src)[i]&maskr)>>shr;\
1178
\
1179
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1180
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1181
    }\
1182
}\
1183
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1184
{\
1185
    int i;\
1186
    for (i=0; i<width; i++) {\
1187
        int pix0= ((const type*)src)[2*i+0];\
1188
        int pix1= ((const type*)src)[2*i+1];\
1189
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1190
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1191
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1192
        g&= maskg|(2*maskg);\
1193
\
1194
        g>>=shg;\
1195
\
1196
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1197
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1198
    }\
1199
}
1200

    
1201
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1202
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1203
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1204
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1205
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1206
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1207

    
1208
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1209
{
1210
    int i;
1211
    for (i=0; i<width; i++) {
1212
        int d= src[i];
1213

    
1214
        dst[i]= pal[d] & 0xFF;
1215
    }
1216
}
1217

    
1218
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1219
                           const uint8_t *src1, const uint8_t *src2,
1220
                           long width, uint32_t *pal)
1221
{
1222
    int i;
1223
    assert(src1 == src2);
1224
    for (i=0; i<width; i++) {
1225
        int p= pal[src1[i]];
1226

    
1227
        dstU[i]= p>>8;
1228
        dstV[i]= p>>16;
1229
    }
1230
}
1231

    
1232
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1233
{
1234
    int i, j;
1235
    for (i=0; i<width/8; i++) {
1236
        int d= ~src[i];
1237
        for(j=0; j<8; j++)
1238
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1239
    }
1240
}
1241

    
1242
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1243
{
1244
    int i, j;
1245
    for (i=0; i<width/8; i++) {
1246
        int d= src[i];
1247
        for(j=0; j<8; j++)
1248
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1249
    }
1250
}
1251

    
1252

    
1253
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1254
//Plain C versions
1255
#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1256
#define COMPILE_C
1257
#endif
1258

    
1259
#if ARCH_PPC
1260
#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1261
#define COMPILE_ALTIVEC
1262
#endif
1263
#endif //ARCH_PPC
1264

    
1265
#if ARCH_X86
1266

    
1267
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1268
#define COMPILE_MMX
1269
#endif
1270

    
1271
#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1272
#define COMPILE_MMX2
1273
#endif
1274

    
1275
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1276
#define COMPILE_3DNOW
1277
#endif
1278
#endif //ARCH_X86
1279

    
1280
#define COMPILE_TEMPLATE_MMX 0
1281
#define COMPILE_TEMPLATE_MMX2 0
1282
#define COMPILE_TEMPLATE_AMD3DNOW 0
1283
#define COMPILE_TEMPLATE_ALTIVEC 0
1284

    
1285
#ifdef COMPILE_C
1286
#define RENAME(a) a ## _C
1287
#include "swscale_template.c"
1288
#endif
1289

    
1290
#ifdef COMPILE_ALTIVEC
1291
#undef RENAME
1292
#undef COMPILE_TEMPLATE_ALTIVEC
1293
#define COMPILE_TEMPLATE_ALTIVEC 1
1294
#define RENAME(a) a ## _altivec
1295
#include "swscale_template.c"
1296
#endif
1297

    
1298
#if ARCH_X86
1299

    
1300
//MMX versions
1301
#ifdef COMPILE_MMX
1302
#undef RENAME
1303
#undef COMPILE_TEMPLATE_MMX
1304
#undef COMPILE_TEMPLATE_MMX2
1305
#undef COMPILE_TEMPLATE_AMD3DNOW
1306
#define COMPILE_TEMPLATE_MMX 1
1307
#define COMPILE_TEMPLATE_MMX2 0
1308
#define COMPILE_TEMPLATE_AMD3DNOW 0
1309
#define RENAME(a) a ## _MMX
1310
#include "swscale_template.c"
1311
#endif
1312

    
1313
//MMX2 versions
1314
#ifdef COMPILE_MMX2
1315
#undef RENAME
1316
#undef COMPILE_TEMPLATE_MMX
1317
#undef COMPILE_TEMPLATE_MMX2
1318
#undef COMPILE_TEMPLATE_AMD3DNOW
1319
#define COMPILE_TEMPLATE_MMX 1
1320
#define COMPILE_TEMPLATE_MMX2 1
1321
#define COMPILE_TEMPLATE_AMD3DNOW 0
1322
#define RENAME(a) a ## _MMX2
1323
#include "swscale_template.c"
1324
#endif
1325

    
1326
//3DNOW versions
1327
#ifdef COMPILE_3DNOW
1328
#undef RENAME
1329
#undef COMPILE_TEMPLATE_MMX
1330
#undef COMPILE_TEMPLATE_MMX2
1331
#undef COMPILE_TEMPLATE_AMD3DNOW
1332
#define COMPILE_TEMPLATE_MMX 1
1333
#define COMPILE_TEMPLATE_MMX2 0
1334
#define COMPILE_TEMPLATE_AMD3DNOW 1
1335
#define RENAME(a) a ## _3DNow
1336
#include "swscale_template.c"
1337
#endif
1338

    
1339
#endif //ARCH_X86
1340

    
1341
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1342
{
1343
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1344
    if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
1345
    else           return getSplineCoeff(        0.0,
1346
                                          b+ 2.0*c + 3.0*d,
1347
                                                 c + 3.0*d,
1348
                                         -b- 3.0*c - 6.0*d,
1349
                                         dist-1.0);
1350
}
1351

    
1352
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1353
                             int srcW, int dstW, int filterAlign, int one, int flags,
1354
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1355
{
1356
    int i;
1357
    int filterSize;
1358
    int filter2Size;
1359
    int minFilterSize;
1360
    int64_t *filter=NULL;
1361
    int64_t *filter2=NULL;
1362
    const int64_t fone= 1LL<<54;
1363
    int ret= -1;
1364
#if ARCH_X86
1365
    if (flags & SWS_CPU_CAPS_MMX)
1366
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1367
#endif
1368

    
1369
    // NOTE: the +1 is for the MMX scaler which reads over the end
1370
    FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW+1)*sizeof(int16_t), fail);
1371

    
1372
    if (FFABS(xInc - 0x10000) <10) { // unscaled
1373
        int i;
1374
        filterSize= 1;
1375
        FF_ALLOCZ_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
1376

    
1377
        for (i=0; i<dstW; i++) {
1378
            filter[i*filterSize]= fone;
1379
            (*filterPos)[i]=i;
1380
        }
1381

    
1382
    } else if (flags&SWS_POINT) { // lame looking point sampling mode
1383
        int i;
1384
        int xDstInSrc;
1385
        filterSize= 1;
1386
        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
1387

    
1388
        xDstInSrc= xInc/2 - 0x8000;
1389
        for (i=0; i<dstW; i++) {
1390
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1391

    
1392
            (*filterPos)[i]= xx;
1393
            filter[i]= fone;
1394
            xDstInSrc+= xInc;
1395
        }
1396
    } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
1397
        int i;
1398
        int xDstInSrc;
1399
        filterSize= 2;
1400
        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
1401

    
1402
        xDstInSrc= xInc/2 - 0x8000;
1403
        for (i=0; i<dstW; i++) {
1404
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1405
            int j;
1406

    
1407
            (*filterPos)[i]= xx;
1408
            //bilinear upscale / linear interpolate / area averaging
1409
            for (j=0; j<filterSize; j++) {
1410
                int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1411
                if (coeff<0) coeff=0;
1412
                filter[i*filterSize + j]= coeff;
1413
                xx++;
1414
            }
1415
            xDstInSrc+= xInc;
1416
        }
1417
    } else {
1418
        int xDstInSrc;
1419
        int sizeFactor;
1420

    
1421
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1422
        else if (flags&SWS_X)            sizeFactor=  8;
1423
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1424
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1425
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1426
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1427
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1428
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1429
        else {
1430
            sizeFactor= 0; //GCC warning killer
1431
            assert(0);
1432
        }
1433

    
1434
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1435
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1436

    
1437
        if (filterSize > srcW-2) filterSize=srcW-2;
1438

    
1439
        FF_ALLOC_OR_GOTO(NULL, filter, dstW*sizeof(*filter)*filterSize, fail);
1440

    
1441
        xDstInSrc= xInc - 0x10000;
1442
        for (i=0; i<dstW; i++) {
1443
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1444
            int j;
1445
            (*filterPos)[i]= xx;
1446
            for (j=0; j<filterSize; j++) {
1447
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1448
                double floatd;
1449
                int64_t coeff;
1450

    
1451
                if (xInc > 1<<16)
1452
                    d= d*dstW/srcW;
1453
                floatd= d * (1.0/(1<<30));
1454

    
1455
                if (flags & SWS_BICUBIC) {
1456
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1457
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1458
                    int64_t dd = ( d*d)>>30;
1459
                    int64_t ddd= (dd*d)>>30;
1460

    
1461
                    if      (d < 1LL<<30)
1462
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1463
                    else if (d < 1LL<<31)
1464
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1465
                    else
1466
                        coeff=0.0;
1467
                    coeff *= fone>>(30+24);
1468
                }
1469
/*                else if (flags & SWS_X) {
1470
                    double p= param ? param*0.01 : 0.3;
1471
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1472
                    coeff*= pow(2.0, - p*d*d);
1473
                }*/
1474
                else if (flags & SWS_X) {
1475
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1476
                    double c;
1477

    
1478
                    if (floatd<1.0)
1479
                        c = cos(floatd*PI);
1480
                    else
1481
                        c=-1.0;
1482
                    if (c<0.0)      c= -pow(-c, A);
1483
                    else            c=  pow( c, A);
1484
                    coeff= (c*0.5 + 0.5)*fone;
1485
                } else if (flags & SWS_AREA) {
1486
                    int64_t d2= d - (1<<29);
1487
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1488
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1489
                    else coeff=0.0;
1490
                    coeff *= fone>>(30+16);
1491
                } else if (flags & SWS_GAUSS) {
1492
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1493
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1494
                } else if (flags & SWS_SINC) {
1495
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1496
                } else if (flags & SWS_LANCZOS) {
1497
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1498
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1499
                    if (floatd>p) coeff=0;
1500
                } else if (flags & SWS_BILINEAR) {
1501
                    coeff= (1<<30) - d;
1502
                    if (coeff<0) coeff=0;
1503
                    coeff *= fone >> 30;
1504
                } else if (flags & SWS_SPLINE) {
1505
                    double p=-2.196152422706632;
1506
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1507
                } else {
1508
                    coeff= 0.0; //GCC warning killer
1509
                    assert(0);
1510
                }
1511

    
1512
                filter[i*filterSize + j]= coeff;
1513
                xx++;
1514
            }
1515
            xDstInSrc+= 2*xInc;
1516
        }
1517
    }
1518

    
1519
    /* apply src & dst Filter to filter -> filter2
1520
       av_free(filter);
1521
    */
1522
    assert(filterSize>0);
1523
    filter2Size= filterSize;
1524
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1525
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1526
    assert(filter2Size>0);
1527
    FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size*dstW*sizeof(*filter2), fail);
1528

    
1529
    for (i=0; i<dstW; i++) {
1530
        int j, k;
1531

    
1532
        if(srcFilter) {
1533
            for (k=0; k<srcFilter->length; k++) {
1534
                for (j=0; j<filterSize; j++)
1535
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1536
            }
1537
        } else {
1538
            for (j=0; j<filterSize; j++)
1539
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1540
        }
1541
        //FIXME dstFilter
1542

    
1543
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1544
    }
1545
    av_freep(&filter);
1546

    
1547
    /* try to reduce the filter-size (step1 find size and shift left) */
1548
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1549
    minFilterSize= 0;
1550
    for (i=dstW-1; i>=0; i--) {
1551
        int min= filter2Size;
1552
        int j;
1553
        int64_t cutOff=0.0;
1554

    
1555
        /* get rid off near zero elements on the left by shifting left */
1556
        for (j=0; j<filter2Size; j++) {
1557
            int k;
1558
            cutOff += FFABS(filter2[i*filter2Size]);
1559

    
1560
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1561

    
1562
            /* preserve monotonicity because the core can't handle the filter otherwise */
1563
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1564

    
1565
            // move filter coefficients left
1566
            for (k=1; k<filter2Size; k++)
1567
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1568
            filter2[i*filter2Size + k - 1]= 0;
1569
            (*filterPos)[i]++;
1570
        }
1571

    
1572
        cutOff=0;
1573
        /* count near zeros on the right */
1574
        for (j=filter2Size-1; j>0; j--) {
1575
            cutOff += FFABS(filter2[i*filter2Size + j]);
1576

    
1577
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1578
            min--;
1579
        }
1580

    
1581
        if (min>minFilterSize) minFilterSize= min;
1582
    }
1583

    
1584
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1585
        // we can handle the special case 4,
1586
        // so we don't want to go to the full 8
1587
        if (minFilterSize < 5)
1588
            filterAlign = 4;
1589

    
1590
        // We really don't want to waste our time
1591
        // doing useless computation, so fall back on
1592
        // the scalar C code for very small filters.
1593
        // Vectorizing is worth it only if you have a
1594
        // decent-sized vector.
1595
        if (minFilterSize < 3)
1596
            filterAlign = 1;
1597
    }
1598

    
1599
    if (flags & SWS_CPU_CAPS_MMX) {
1600
        // special case for unscaled vertical filtering
1601
        if (minFilterSize == 1 && filterAlign == 2)
1602
            filterAlign= 1;
1603
    }
1604

    
1605
    assert(minFilterSize > 0);
1606
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1607
    assert(filterSize > 0);
1608
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1609
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1610
        goto fail;
1611
    *outFilterSize= filterSize;
1612

    
1613
    if (flags&SWS_PRINT_INFO)
1614
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1615
    /* try to reduce the filter-size (step2 reduce it) */
1616
    for (i=0; i<dstW; i++) {
1617
        int j;
1618

    
1619
        for (j=0; j<filterSize; j++) {
1620
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1621
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1622
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1623
                filter[i*filterSize + j]= 0;
1624
        }
1625
    }
1626

    
1627

    
1628
    //FIXME try to align filterPos if possible
1629

    
1630
    //fix borders
1631
    for (i=0; i<dstW; i++) {
1632
        int j;
1633
        if ((*filterPos)[i] < 0) {
1634
            // move filter coefficients left to compensate for filterPos
1635
            for (j=1; j<filterSize; j++) {
1636
                int left= FFMAX(j + (*filterPos)[i], 0);
1637
                filter[i*filterSize + left] += filter[i*filterSize + j];
1638
                filter[i*filterSize + j]=0;
1639
            }
1640
            (*filterPos)[i]= 0;
1641
        }
1642

    
1643
        if ((*filterPos)[i] + filterSize > srcW) {
1644
            int shift= (*filterPos)[i] + filterSize - srcW;
1645
            // move filter coefficients right to compensate for filterPos
1646
            for (j=filterSize-2; j>=0; j--) {
1647
                int right= FFMIN(j + shift, filterSize-1);
1648
                filter[i*filterSize +right] += filter[i*filterSize +j];
1649
                filter[i*filterSize +j]=0;
1650
            }
1651
            (*filterPos)[i]= srcW - filterSize;
1652
        }
1653
    }
1654

    
1655
    // Note the +1 is for the MMX scaler which reads over the end
1656
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1657
    FF_ALLOCZ_OR_GOTO(NULL, *outFilter, *outFilterSize*(dstW+1)*sizeof(int16_t), fail);
1658

    
1659
    /* normalize & store in outFilter */
1660
    for (i=0; i<dstW; i++) {
1661
        int j;
1662
        int64_t error=0;
1663
        int64_t sum=0;
1664

    
1665
        for (j=0; j<filterSize; j++) {
1666
            sum+= filter[i*filterSize + j];
1667
        }
1668
        sum= (sum + one/2)/ one;
1669
        for (j=0; j<*outFilterSize; j++) {
1670
            int64_t v= filter[i*filterSize + j] + error;
1671
            int intV= ROUNDED_DIV(v, sum);
1672
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1673
            error= v - intV*sum;
1674
        }
1675
    }
1676

    
1677
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1678
    for (i=0; i<*outFilterSize; i++) {
1679
        int j= dstW*(*outFilterSize);
1680
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1681
    }
1682

    
1683
    ret=0;
1684
fail:
1685
    av_free(filter);
1686
    av_free(filter2);
1687
    return ret;
1688
}
1689

    
1690
#ifdef COMPILE_MMX2
1691
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
1692
{
1693
    uint8_t *fragmentA;
1694
    x86_reg imm8OfPShufW1A;
1695
    x86_reg imm8OfPShufW2A;
1696
    x86_reg fragmentLengthA;
1697
    uint8_t *fragmentB;
1698
    x86_reg imm8OfPShufW1B;
1699
    x86_reg imm8OfPShufW2B;
1700
    x86_reg fragmentLengthB;
1701
    int fragmentPos;
1702

    
1703
    int xpos, i;
1704

    
1705
    // create an optimized horizontal scaling routine
1706
    /* This scaler is made of runtime-generated MMX2 code using specially
1707
     * tuned pshufw instructions. For every four output pixels, if four
1708
     * input pixels are enough for the fast bilinear scaling, then a chunk
1709
     * of fragmentB is used. If five input pixels are needed, then a chunk
1710
     * of fragmentA is used.
1711
     */
1712

    
1713
    //code fragment
1714

    
1715
    __asm__ volatile(
1716
        "jmp                         9f                 \n\t"
1717
    // Begin
1718
        "0:                                             \n\t"
1719
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1720
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1721
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1722
        "punpcklbw                %%mm7, %%mm1          \n\t"
1723
        "punpcklbw                %%mm7, %%mm0          \n\t"
1724
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1725
        "1:                                             \n\t"
1726
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1727
        "2:                                             \n\t"
1728
        "psubw                    %%mm1, %%mm0          \n\t"
1729
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1730
        "pmullw                   %%mm3, %%mm0          \n\t"
1731
        "psllw                       $7, %%mm1          \n\t"
1732
        "paddw                    %%mm1, %%mm0          \n\t"
1733

    
1734
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1735

    
1736
        "add                         $8, %%"REG_a"      \n\t"
1737
    // End
1738
        "9:                                             \n\t"
1739
//        "int $3                                         \n\t"
1740
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1741
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1742
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1743
        "dec                         %1                 \n\t"
1744
        "dec                         %2                 \n\t"
1745
        "sub                         %0, %1             \n\t"
1746
        "sub                         %0, %2             \n\t"
1747
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1748
        "sub                         %0, %3             \n\t"
1749

    
1750

    
1751
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1752
        "=r" (fragmentLengthA)
1753
    );
1754

    
1755
    __asm__ volatile(
1756
        "jmp                         9f                 \n\t"
1757
    // Begin
1758
        "0:                                             \n\t"
1759
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1760
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1761
        "punpcklbw                %%mm7, %%mm0          \n\t"
1762
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1763
        "1:                                             \n\t"
1764
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1765
        "2:                                             \n\t"
1766
        "psubw                    %%mm1, %%mm0          \n\t"
1767
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1768
        "pmullw                   %%mm3, %%mm0          \n\t"
1769
        "psllw                       $7, %%mm1          \n\t"
1770
        "paddw                    %%mm1, %%mm0          \n\t"
1771

    
1772
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1773

    
1774
        "add                         $8, %%"REG_a"      \n\t"
1775
    // End
1776
        "9:                                             \n\t"
1777
//        "int                       $3                   \n\t"
1778
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1779
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1780
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1781
        "dec                         %1                 \n\t"
1782
        "dec                         %2                 \n\t"
1783
        "sub                         %0, %1             \n\t"
1784
        "sub                         %0, %2             \n\t"
1785
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1786
        "sub                         %0, %3             \n\t"
1787

    
1788

    
1789
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1790
        "=r" (fragmentLengthB)
1791
    );
1792

    
1793
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1794
    fragmentPos=0;
1795

    
1796
    for (i=0; i<dstW/numSplits; i++) {
1797
        int xx=xpos>>16;
1798

    
1799
        if ((i&3) == 0) {
1800
            int a=0;
1801
            int b=((xpos+xInc)>>16) - xx;
1802
            int c=((xpos+xInc*2)>>16) - xx;
1803
            int d=((xpos+xInc*3)>>16) - xx;
1804
            int inc                = (d+1<4);
1805
            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
1806
            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
1807
            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
1808
            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
1809
            int maxShift= 3-(d+inc);
1810
            int shift=0;
1811

    
1812
            if (filterCode) {
1813
                filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1814
                filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1815
                filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1816
                filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1817
                filterPos[i/2]= xx;
1818

    
1819
                memcpy(filterCode + fragmentPos, fragment, fragmentLength);
1820

    
1821
                filterCode[fragmentPos + imm8OfPShufW1]=
1822
                    (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
1823
                filterCode[fragmentPos + imm8OfPShufW2]=
1824
                    a | (b<<2) | (c<<4) | (d<<6);
1825

    
1826
                if (i+4-inc>=dstW) shift=maxShift; //avoid overread
1827
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1828

    
1829
                if (shift && i>=shift) {
1830
                    filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
1831
                    filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
1832
                    filterPos[i/2]-=shift;
1833
                }
1834
            }
1835

    
1836
            fragmentPos+= fragmentLength;
1837

    
1838
            if (filterCode)
1839
                filterCode[fragmentPos]= RET;
1840
        }
1841
        xpos+=xInc;
1842
    }
1843
    if (filterCode)
1844
        filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1845

    
1846
    return fragmentPos + 1;
1847
}
1848
#endif /* COMPILE_MMX2 */
1849

    
1850
static SwsFunc getSwsFunc(SwsContext *c)
1851
{
1852
#if CONFIG_RUNTIME_CPUDETECT
1853
    int flags = c->flags;
1854

    
1855
#if ARCH_X86 && CONFIG_GPL
1856
    // ordered per speed fastest first
1857
    if (flags & SWS_CPU_CAPS_MMX2) {
1858
        sws_init_swScale_MMX2(c);
1859
        return swScale_MMX2;
1860
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1861
        sws_init_swScale_3DNow(c);
1862
        return swScale_3DNow;
1863
    } else if (flags & SWS_CPU_CAPS_MMX) {
1864
        sws_init_swScale_MMX(c);
1865
        return swScale_MMX;
1866
    }
1867
#else
1868
#if ARCH_PPC
1869
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1870
        sws_init_swScale_altivec(c);
1871
        return swScale_altivec;
1872
    }
1873
#endif
1874
#endif /* ARCH_X86 && CONFIG_GPL */
1875
#else //CONFIG_RUNTIME_CPUDETECT
1876
#if   COMPILE_TEMPLATE_MMX2
1877
    sws_init_swScale_MMX2(c);
1878
    return swScale_MMX2;
1879
#elif COMPILE_TEMPLATE_AMD3DNOW
1880
    sws_init_swScale_3DNow(c);
1881
    return swScale_3DNow;
1882
#elif COMPILE_TEMPLATE_MMX
1883
    sws_init_swScale_MMX(c);
1884
    return swScale_MMX;
1885
#elif COMPILE_TEMPLATE_ALTIVEC
1886
    sws_init_swScale_altivec(c);
1887
    return swScale_altivec;
1888
#endif
1889
#endif //!CONFIG_RUNTIME_CPUDETECT
1890

    
1891
    sws_init_swScale_C(c);
1892
    return swScale_C;
1893
}
1894

    
1895
static int PlanarToNV12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1896
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1897
{
1898
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1899
    /* Copy Y plane */
1900
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1901
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1902
    else {
1903
        int i;
1904
        const uint8_t *srcPtr= src[0];
1905
        uint8_t *dstPtr= dst;
1906
        for (i=0; i<srcSliceH; i++) {
1907
            memcpy(dstPtr, srcPtr, c->srcW);
1908
            srcPtr+= srcStride[0];
1909
            dstPtr+= dstStride[0];
1910
        }
1911
    }
1912
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1913
    if (c->dstFormat == PIX_FMT_NV12)
1914
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1915
    else
1916
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1917

    
1918
    return srcSliceH;
1919
}
1920

    
1921
static int PlanarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1922
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1923
{
1924
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1925

    
1926
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1927

    
1928
    return srcSliceH;
1929
}
1930

    
1931
static int PlanarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1932
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1933
{
1934
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1935

    
1936
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1937

    
1938
    return srcSliceH;
1939
}
1940

    
1941
static int YUV422PToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1942
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1943
{
1944
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1945

    
1946
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1947

    
1948
    return srcSliceH;
1949
}
1950

    
1951
static int YUV422PToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1952
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
1953
{
1954
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1955

    
1956
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1957

    
1958
    return srcSliceH;
1959
}
1960

    
1961
static int YUYV2YUV420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1962
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1963
{
1964
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1965
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1966
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1967

    
1968
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1969

    
1970
    if (dstParam[3])
1971
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1972

    
1973
    return srcSliceH;
1974
}
1975

    
1976
static int YUYV2YUV422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1977
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1978
{
1979
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1980
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1981
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1982

    
1983
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1984

    
1985
    return srcSliceH;
1986
}
1987

    
1988
static int UYVY2YUV420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1989
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1990
{
1991
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1992
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1993
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1994

    
1995
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1996

    
1997
    if (dstParam[3])
1998
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1999

    
2000
    return srcSliceH;
2001
}
2002

    
2003
static int UYVY2YUV422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2004
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2005
{
2006
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2007
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2008
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2009

    
2010
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2011

    
2012
    return srcSliceH;
2013
}
2014

    
2015
static int pal2rgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2016
                          int srcSliceH, uint8_t* dst[], int dstStride[])
2017
{
2018
    const enum PixelFormat srcFormat= c->srcFormat;
2019
    const enum PixelFormat dstFormat= c->dstFormat;
2020
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
2021
                 const uint8_t *palette)=NULL;
2022
    int i;
2023
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2024
    const uint8_t *srcPtr= src[0];
2025

    
2026
    if (!usePal(srcFormat))
2027
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2028
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2029

    
2030
    switch(dstFormat) {
2031
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
2032
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
2033
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
2034
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
2035
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
2036
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
2037
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2038
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2039
    }
2040

    
2041

    
2042
    for (i=0; i<srcSliceH; i++) {
2043
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
2044
        srcPtr+= srcStride[0];
2045
        dstPtr+= dstStride[0];
2046
    }
2047

    
2048
    return srcSliceH;
2049
}
2050

    
2051
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
2052
static int rgb2rgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2053
                          int srcSliceH, uint8_t* dst[], int dstStride[])
2054
{
2055
    const enum PixelFormat srcFormat= c->srcFormat;
2056
    const enum PixelFormat dstFormat= c->dstFormat;
2057
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2058
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2059
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2060
    const int dstId= fmt_depth(dstFormat) >> 2;
2061
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2062

    
2063
    /* BGR -> BGR */
2064
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
2065
       || (isRGB(srcFormat) && isRGB(dstFormat))) {
2066
        switch(srcId | (dstId<<4)) {
2067
        case 0x34: conv= rgb16to15; break;
2068
        case 0x36: conv= rgb24to15; break;
2069
        case 0x38: conv= rgb32to15; break;
2070
        case 0x43: conv= rgb15to16; break;
2071
        case 0x46: conv= rgb24to16; break;
2072
        case 0x48: conv= rgb32to16; break;
2073
        case 0x63: conv= rgb15to24; break;
2074
        case 0x64: conv= rgb16to24; break;
2075
        case 0x68: conv= rgb32to24; break;
2076
        case 0x83: conv= rgb15to32; break;
2077
        case 0x84: conv= rgb16to32; break;
2078
        case 0x86: conv= rgb24to32; break;
2079
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2080
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2081
        }
2082
    } else if (  (isBGR(srcFormat) && isRGB(dstFormat))
2083
             || (isRGB(srcFormat) && isBGR(dstFormat))) {
2084
        switch(srcId | (dstId<<4)) {
2085
        case 0x33: conv= rgb15tobgr15; break;
2086
        case 0x34: conv= rgb16tobgr15; break;
2087
        case 0x36: conv= rgb24tobgr15; break;
2088
        case 0x38: conv= rgb32tobgr15; break;
2089
        case 0x43: conv= rgb15tobgr16; break;
2090
        case 0x44: conv= rgb16tobgr16; break;
2091
        case 0x46: conv= rgb24tobgr16; break;
2092
        case 0x48: conv= rgb32tobgr16; break;
2093
        case 0x63: conv= rgb15tobgr24; break;
2094
        case 0x64: conv= rgb16tobgr24; break;
2095
        case 0x66: conv= rgb24tobgr24; break;
2096
        case 0x68: conv= rgb32tobgr24; break;
2097
        case 0x83: conv= rgb15tobgr32; break;
2098
        case 0x84: conv= rgb16tobgr32; break;
2099
        case 0x86: conv= rgb24tobgr32; break;
2100
        case 0x88: conv= rgb32tobgr32; break;
2101
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2102
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2103
        }
2104
    } else {
2105
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2106
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2107
    }
2108

    
2109
    if(conv) {
2110
        const uint8_t *srcPtr= src[0];
2111
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2112
            srcPtr += ALT32_CORR;
2113

    
2114
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2115
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2116
        else {
2117
            int i;
2118
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2119

    
2120
            for (i=0; i<srcSliceH; i++) {
2121
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
2122
                srcPtr+= srcStride[0];
2123
                dstPtr+= dstStride[0];
2124
            }
2125
        }
2126
    }
2127
    return srcSliceH;
2128
}
2129

    
2130
static int bgr24toyv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2131
                              int srcSliceH, uint8_t* dst[], int dstStride[])
2132
{
2133

    
2134
    rgb24toyv12(
2135
        src[0],
2136
        dst[0]+ srcSliceY    *dstStride[0],
2137
        dst[1]+(srcSliceY>>1)*dstStride[1],
2138
        dst[2]+(srcSliceY>>1)*dstStride[2],
2139
        c->srcW, srcSliceH,
2140
        dstStride[0], dstStride[1], srcStride[0]);
2141
    if (dst[3])
2142
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2143
    return srcSliceH;
2144
}
2145

    
2146
static int yvu9toyv12Wrapper(SwsContext *c, const const uint8_t* src[], int srcStride[], int srcSliceY,
2147
                             int srcSliceH, uint8_t* dst[], int dstStride[])
2148
{
2149
    int i;
2150

    
2151
    /* copy Y */
2152
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2153
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2154
    else {
2155
        const uint8_t *srcPtr= src[0];
2156
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2157

    
2158
        for (i=0; i<srcSliceH; i++) {
2159
            memcpy(dstPtr, srcPtr, c->srcW);
2160
            srcPtr+= srcStride[0];
2161
            dstPtr+= dstStride[0];
2162
        }
2163
    }
2164

    
2165
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
2166
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2167
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
2168
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2169
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
2170
    } else {
2171
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2172
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
2173
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2174
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
2175
    }
2176
    if (dst[3])
2177
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2178
    return srcSliceH;
2179
}
2180

    
2181
/* unscaled copy like stuff (assumes nearly identical formats) */
2182
static int packedCopy(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2183
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2184
{
2185
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2186
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2187
    else {
2188
        int i;
2189
        const uint8_t *srcPtr= src[0];
2190
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2191
        int length=0;
2192

    
2193
        /* universal length finder */
2194
        while(length+c->srcW <= FFABS(dstStride[0])
2195
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2196
        assert(length!=0);
2197

    
2198
        for (i=0; i<srcSliceH; i++) {
2199
            memcpy(dstPtr, srcPtr, length);
2200
            srcPtr+= srcStride[0];
2201
            dstPtr+= dstStride[0];
2202
        }
2203
    }
2204
    return srcSliceH;
2205
}
2206

    
2207
static int planarCopy(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2208
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2209
{
2210
    int plane, i, j;
2211
    for (plane=0; plane<4; plane++) {
2212
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
2213
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2214
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
2215
        const uint8_t *srcPtr= src[plane];
2216
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2217

    
2218
        if (!dst[plane]) continue;
2219
        // ignore palette for GRAY8
2220
        if (plane == 1 && !dst[2]) continue;
2221
        if (!src[plane] || (plane == 1 && !src[2])) {
2222
            if(is16BPS(c->dstFormat))
2223
                length*=2;
2224
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
2225
        } else {
2226
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
2227
                if (!isBE(c->srcFormat)) srcPtr++;
2228
                for (i=0; i<height; i++) {
2229
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2230
                    srcPtr+= srcStride[plane];
2231
                    dstPtr+= dstStride[plane];
2232
                }
2233
            } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
2234
                for (i=0; i<height; i++) {
2235
                    for (j=0; j<length; j++) {
2236
                        dstPtr[ j<<1   ] = srcPtr[j];
2237
                        dstPtr[(j<<1)+1] = srcPtr[j];
2238
                    }
2239
                    srcPtr+= srcStride[plane];
2240
                    dstPtr+= dstStride[plane];
2241
                }
2242
            } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2243
                  && isBE(c->srcFormat) != isBE(c->dstFormat)) {
2244

    
2245
                for (i=0; i<height; i++) {
2246
                    for (j=0; j<length; j++)
2247
                        ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2248
                    srcPtr+= srcStride[plane];
2249
                    dstPtr+= dstStride[plane];
2250
                }
2251
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2252
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2253
            else {
2254
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2255
                    length*=2;
2256
                for (i=0; i<height; i++) {
2257
                    memcpy(dstPtr, srcPtr, length);
2258
                    srcPtr+= srcStride[plane];
2259
                    dstPtr+= dstStride[plane];
2260
                }
2261
            }
2262
        }
2263
    }
2264
    return srcSliceH;
2265
}
2266

    
2267

    
2268
static void getSubSampleFactors(int *h, int *v, enum PixelFormat format)
2269
{
2270
    *h = av_pix_fmt_descriptors[format].log2_chroma_w;
2271
    *v = av_pix_fmt_descriptors[format].log2_chroma_h;
2272
}
2273

    
2274
static uint16_t roundToInt16(int64_t f)
2275
{
2276
    int r= (f + (1<<15))>>16;
2277
         if (r<-0x7FFF) return 0x8000;
2278
    else if (r> 0x7FFF) return 0x7FFF;
2279
    else                return r;
2280
}
2281

    
2282
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)
2283
{
2284
    int64_t crv =  inv_table[0];
2285
    int64_t cbu =  inv_table[1];
2286
    int64_t cgu = -inv_table[2];
2287
    int64_t cgv = -inv_table[3];
2288
    int64_t cy  = 1<<16;
2289
    int64_t oy  = 0;
2290

    
2291
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2292
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2293

    
2294
    c->brightness= brightness;
2295
    c->contrast  = contrast;
2296
    c->saturation= saturation;
2297
    c->srcRange  = srcRange;
2298
    c->dstRange  = dstRange;
2299
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2300

    
2301
    c->uOffset=   0x0400040004000400LL;
2302
    c->vOffset=   0x0400040004000400LL;
2303

    
2304
    if (!srcRange) {
2305
        cy= (cy*255) / 219;
2306
        oy= 16<<16;
2307
    } else {
2308
        crv= (crv*224) / 255;
2309
        cbu= (cbu*224) / 255;
2310
        cgu= (cgu*224) / 255;
2311
        cgv= (cgv*224) / 255;
2312
    }
2313

    
2314
    cy = (cy *contrast             )>>16;
2315
    crv= (crv*contrast * saturation)>>32;
2316
    cbu= (cbu*contrast * saturation)>>32;
2317
    cgu= (cgu*contrast * saturation)>>32;
2318
    cgv= (cgv*contrast * saturation)>>32;
2319

    
2320
    oy -= 256*brightness;
2321

    
2322
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2323
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2324
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2325
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2326
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2327
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2328

    
2329
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2330
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2331
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2332
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2333
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2334
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2335

    
2336
    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2337
    //FIXME factorize
2338

    
2339
#ifdef COMPILE_ALTIVEC
2340
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2341
        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
2342
#endif
2343
    return 0;
2344
}
2345

    
2346
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)
2347
{
2348
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2349

    
2350
    *inv_table = c->srcColorspaceTable;
2351
    *table     = c->dstColorspaceTable;
2352
    *srcRange  = c->srcRange;
2353
    *dstRange  = c->dstRange;
2354
    *brightness= c->brightness;
2355
    *contrast  = c->contrast;
2356
    *saturation= c->saturation;
2357

    
2358
    return 0;
2359
}
2360

    
2361
static int handle_jpeg(enum PixelFormat *format)
2362
{
2363
    switch (*format) {
2364
    case PIX_FMT_YUVJ420P:
2365
        *format = PIX_FMT_YUV420P;
2366
        return 1;
2367
    case PIX_FMT_YUVJ422P:
2368
        *format = PIX_FMT_YUV422P;
2369
        return 1;
2370
    case PIX_FMT_YUVJ444P:
2371
        *format = PIX_FMT_YUV444P;
2372
        return 1;
2373
    case PIX_FMT_YUVJ440P:
2374
        *format = PIX_FMT_YUV440P;
2375
        return 1;
2376
    default:
2377
        return 0;
2378
    }
2379
}
2380

    
2381
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2382
                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2383
{
2384

    
2385
    SwsContext *c;
2386
    int i;
2387
    int usesVFilter, usesHFilter;
2388
    int unscaled, needsDither;
2389
    int srcRange, dstRange;
2390
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2391
#if ARCH_X86
2392
    if (flags & SWS_CPU_CAPS_MMX)
2393
        __asm__ volatile("emms\n\t"::: "memory");
2394
#endif
2395

    
2396
#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
2397
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2398
#if   COMPILE_TEMPLATE_MMX2
2399
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2400
#elif COMPILE_TEMPLATE_AMD3DNOW
2401
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2402
#elif COMPILE_TEMPLATE_MMX
2403
    flags |= SWS_CPU_CAPS_MMX;
2404
#elif COMPILE_TEMPLATE_ALTIVEC
2405
    flags |= SWS_CPU_CAPS_ALTIVEC;
2406
#elif ARCH_BFIN
2407
    flags |= SWS_CPU_CAPS_BFIN;
2408
#endif
2409
#endif /* CONFIG_RUNTIME_CPUDETECT */
2410
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2411

    
2412
    unscaled = (srcW == dstW && srcH == dstH);
2413
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2414
        && (fmt_depth(dstFormat))<24
2415
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2416

    
2417
    srcRange = handle_jpeg(&srcFormat);
2418
    dstRange = handle_jpeg(&dstFormat);
2419

    
2420
    if (!isSupportedIn(srcFormat)) {
2421
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2422
        return NULL;
2423
    }
2424
    if (!isSupportedOut(dstFormat)) {
2425
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2426
        return NULL;
2427
    }
2428

    
2429
    i= flags & ( SWS_POINT
2430
                |SWS_AREA
2431
                |SWS_BILINEAR
2432
                |SWS_FAST_BILINEAR
2433
                |SWS_BICUBIC
2434
                |SWS_X
2435
                |SWS_GAUSS
2436
                |SWS_LANCZOS
2437
                |SWS_SINC
2438
                |SWS_SPLINE
2439
                |SWS_BICUBLIN);
2440
    if(!i || (i & (i-1))) {
2441
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2442
        return NULL;
2443
    }
2444

    
2445
    /* sanity check */
2446
    if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
2447
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2448
               srcW, srcH, dstW, dstH);
2449
        return NULL;
2450
    }
2451
    if(srcW > VOFW || dstW > VOFW) {
2452
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2453
        return NULL;
2454
    }
2455

    
2456
    if (!dstFilter) dstFilter= &dummyFilter;
2457
    if (!srcFilter) srcFilter= &dummyFilter;
2458

    
2459
    FF_ALLOCZ_OR_GOTO(NULL, c, sizeof(SwsContext), fail);
2460

    
2461
    c->av_class = &sws_context_class;
2462
    c->srcW= srcW;
2463
    c->srcH= srcH;
2464
    c->dstW= dstW;
2465
    c->dstH= dstH;
2466
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2467
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2468
    c->flags= flags;
2469
    c->dstFormat= dstFormat;
2470
    c->srcFormat= srcFormat;
2471
    c->vRounder= 4* 0x0001000100010001ULL;
2472

    
2473
    usesHFilter= usesVFilter= 0;
2474
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2475
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2476
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2477
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2478
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2479
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2480
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2481
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2482

    
2483
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2484
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2485

    
2486
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2487
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2488

    
2489
    // drop some chroma lines if the user wants it
2490
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2491
    c->chrSrcVSubSample+= c->vChrDrop;
2492

    
2493
    // drop every other pixel for chroma calculation unless user wants full chroma
2494
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2495
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2496
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2497
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2498
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2499
        c->chrSrcHSubSample=1;
2500

    
2501
    if (param) {
2502
        c->param[0] = param[0];
2503
        c->param[1] = param[1];
2504
    } else {
2505
        c->param[0] =
2506
        c->param[1] = SWS_PARAM_DEFAULT;
2507
    }
2508

    
2509
    // Note the -((-x)>>y) is so that we always round toward +inf.
2510
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2511
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2512
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2513
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2514

    
2515
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2516

    
2517
    /* unscaled special cases */
2518
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {
2519
        /* yv12_to_nv12 */
2520
        if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
2521
            c->swScale= PlanarToNV12Wrapper;
2522
        }
2523
        /* yuv2bgr */
2524
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2525
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
2526
            c->swScale= ff_yuv2rgb_get_func_ptr(c);
2527
        }
2528

    
2529
        if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
2530
            c->swScale= yvu9toyv12Wrapper;
2531
        }
2532

    
2533
        /* bgr24toYV12 */
2534
        if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2535
            c->swScale= bgr24toyv12Wrapper;
2536

    
2537
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2538
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2539
           && (isBGR(dstFormat) || isRGB(dstFormat))
2540
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2541
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2542
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2543
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2544
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2545
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2546
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2547
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2548
                                             && dstFormat != PIX_FMT_RGB32_1
2549
                                             && dstFormat != PIX_FMT_BGR32_1
2550
           && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
2551
           && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
2552
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2553
             c->swScale= rgb2rgbWrapper;
2554

    
2555
        if ((usePal(srcFormat) && (
2556
                 dstFormat == PIX_FMT_RGB32   ||
2557
                 dstFormat == PIX_FMT_RGB32_1 ||
2558
                 dstFormat == PIX_FMT_RGB24   ||
2559
                 dstFormat == PIX_FMT_BGR32   ||
2560
                 dstFormat == PIX_FMT_BGR32_1 ||
2561
                 dstFormat == PIX_FMT_BGR24)))
2562
             c->swScale= pal2rgbWrapper;
2563

    
2564
        if (srcFormat == PIX_FMT_YUV422P) {
2565
            if (dstFormat == PIX_FMT_YUYV422)
2566
                c->swScale= YUV422PToYuy2Wrapper;
2567
            else if (dstFormat == PIX_FMT_UYVY422)
2568
                c->swScale= YUV422PToUyvyWrapper;
2569
        }
2570

    
2571
        /* LQ converters if -sws 0 or -sws 4*/
2572
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
2573
            /* yv12_to_yuy2 */
2574
            if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
2575
                if (dstFormat == PIX_FMT_YUYV422)
2576
                    c->swScale= PlanarToYuy2Wrapper;
2577
                else if (dstFormat == PIX_FMT_UYVY422)
2578
                    c->swScale= PlanarToUyvyWrapper;
2579
            }
2580
        }
2581
        if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2582
            c->swScale= YUYV2YUV420Wrapper;
2583
        if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2584
            c->swScale= UYVY2YUV420Wrapper;
2585
        if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2586
            c->swScale= YUYV2YUV422Wrapper;
2587
        if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2588
            c->swScale= UYVY2YUV422Wrapper;
2589

    
2590
#ifdef COMPILE_ALTIVEC
2591
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2592
            !(c->flags & SWS_BITEXACT) &&
2593
            srcFormat == PIX_FMT_YUV420P) {
2594
          // unscaled YV12 -> packed YUV, we want speed
2595
          if (dstFormat == PIX_FMT_YUYV422)
2596
              c->swScale= yv12toyuy2_unscaled_altivec;
2597
          else if (dstFormat == PIX_FMT_UYVY422)
2598
              c->swScale= yv12touyvy_unscaled_altivec;
2599
        }
2600
#endif
2601

    
2602
        /* simple copy */
2603
        if (  srcFormat == dstFormat
2604
            || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2605
            || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2606
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2607
            || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2608
            || (isGray(dstFormat) && isGray(srcFormat))
2609
            || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2610
                && c->chrDstHSubSample == c->chrSrcHSubSample
2611
                && c->chrDstVSubSample == c->chrSrcVSubSample
2612
                && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
2613
                && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
2614
        {
2615
            if (isPacked(c->srcFormat))
2616
                c->swScale= packedCopy;
2617
            else /* Planar YUV or gray */
2618
                c->swScale= planarCopy;
2619
        }
2620
#if ARCH_BFIN
2621
        if (flags & SWS_CPU_CAPS_BFIN)
2622
            ff_bfin_get_unscaled_swscale (c);
2623
#endif
2624

    
2625
        if (c->swScale) {
2626
            if (flags&SWS_PRINT_INFO)
2627
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2628
                       sws_format_name(srcFormat), sws_format_name(dstFormat));
2629
            return c;
2630
        }
2631
    }
2632

    
2633
    if (flags & SWS_CPU_CAPS_MMX2) {
2634
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2635
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
2636
            if (flags&SWS_PRINT_INFO)
2637
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2638
        }
2639
        if (usesHFilter) c->canMMX2BeUsed=0;
2640
    }
2641
    else
2642
        c->canMMX2BeUsed=0;
2643

    
2644
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2645
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2646

    
2647
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2648
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2649
    // n-2 is the last chrominance sample available
2650
    // this is not perfect, but no one should notice the difference, the more correct variant
2651
    // would be like the vertical one, but that would require some special code for the
2652
    // first and last pixel
2653
    if (flags&SWS_FAST_BILINEAR) {
2654
        if (c->canMMX2BeUsed) {
2655
            c->lumXInc+= 20;
2656
            c->chrXInc+= 20;
2657
        }
2658
        //we don't use the x86 asm scaler if MMX is available
2659
        else if (flags & SWS_CPU_CAPS_MMX) {
2660
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2661
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2662
        }
2663
    }
2664

    
2665
    /* precalculate horizontal scaler filter coefficients */
2666
    {
2667
#if defined(COMPILE_MMX2)
2668
// can't downscale !!!
2669
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
2670
            c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);
2671
            c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
2672

    
2673
#ifdef MAP_ANONYMOUS
2674
            c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2675
            c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2676
#elif HAVE_VIRTUALALLOC
2677
            c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2678
            c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2679
#else
2680
            c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
2681
            c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
2682
#endif
2683

    
2684
            FF_ALLOCZ_OR_GOTO(c, c->hLumFilter   , (dstW        /8+8)*sizeof(int16_t), fail);
2685
            FF_ALLOCZ_OR_GOTO(c, c->hChrFilter   , (c->chrDstW  /4+8)*sizeof(int16_t), fail);
2686
            FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW      /2/8+8)*sizeof(int32_t), fail);
2687
            FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW/2/4+8)*sizeof(int32_t), fail);
2688

    
2689
            initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->hLumFilter, c->hLumFilterPos, 8);
2690
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->hChrFilter, c->hChrFilterPos, 4);
2691

    
2692
#ifdef MAP_ANONYMOUS
2693
            mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
2694
            mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
2695
#endif
2696
        } else
2697
#endif /* defined(COMPILE_MMX2) */
2698
        {
2699
            const int filterAlign=
2700
                (flags & SWS_CPU_CAPS_MMX) ? 4 :
2701
                (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2702
                1;
2703

    
2704
            if (initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2705
                           srcW      ,       dstW, filterAlign, 1<<14,
2706
                           (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2707
                           srcFilter->lumH, dstFilter->lumH, c->param) < 0)
2708
                goto fail;
2709
            if (initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2710
                           c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2711
                           (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2712
                           srcFilter->chrH, dstFilter->chrH, c->param) < 0)
2713
                goto fail;
2714
        }
2715
    } // initialize horizontal stuff
2716

    
2717

    
2718

    
2719
    /* precalculate vertical scaler filter coefficients */
2720
    {
2721
        const int filterAlign=
2722
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2723
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2724
            1;
2725

    
2726
        if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2727
                       srcH      ,        dstH, filterAlign, (1<<12),
2728
                       (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2729
                       srcFilter->lumV, dstFilter->lumV, c->param) < 0)
2730
            goto fail;
2731
        if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2732
                       c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2733
                       (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2734
                       srcFilter->chrV, dstFilter->chrV, c->param) < 0)
2735
            goto fail;
2736

    
2737
#ifdef COMPILE_ALTIVEC
2738
        FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof (vector signed short)*c->vLumFilterSize*c->dstH, fail);
2739
        FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH, fail);
2740

    
2741
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2742
            int j;
2743
            short *p = (short *)&c->vYCoeffsBank[i];
2744
            for (j=0;j<8;j++)
2745
                p[j] = c->vLumFilter[i];
2746
        }
2747

    
2748
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2749
            int j;
2750
            short *p = (short *)&c->vCCoeffsBank[i];
2751
            for (j=0;j<8;j++)
2752
                p[j] = c->vChrFilter[i];
2753
        }
2754
#endif
2755
    }
2756

    
2757
    // calculate buffer sizes so that they won't run out while handling these damn slices
2758
    c->vLumBufSize= c->vLumFilterSize;
2759
    c->vChrBufSize= c->vChrFilterSize;
2760
    for (i=0; i<dstH; i++) {
2761
        int chrI= i*c->chrDstH / dstH;
2762
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2763
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2764

    
2765
        nextSlice>>= c->chrSrcVSubSample;
2766
        nextSlice<<= c->chrSrcVSubSample;
2767
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2768
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2769
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2770
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2771
    }
2772

    
2773
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2774
    // allocate several megabytes to handle all possible cases)
2775
    FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
2776
    FF_ALLOC_OR_GOTO(c, c->chrPixBuf, c->vChrBufSize*2*sizeof(int16_t*), fail);
2777
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2778
        FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize*2*sizeof(int16_t*), fail);
2779
    //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000)
2780
    /* align at 16 bytes for AltiVec */
2781
    for (i=0; i<c->vLumBufSize; i++) {
2782
        FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i+c->vLumBufSize], VOF+1, fail);
2783
        c->lumPixBuf[i] = c->lumPixBuf[i+c->vLumBufSize];
2784
    }
2785
    for (i=0; i<c->vChrBufSize; i++) {
2786
        FF_ALLOC_OR_GOTO(c, c->chrPixBuf[i+c->vChrBufSize], (VOF+1)*2, fail);
2787
        c->chrPixBuf[i] = c->chrPixBuf[i+c->vChrBufSize];
2788
    }
2789
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2790
        for (i=0; i<c->vLumBufSize; i++) {
2791
            FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i+c->vLumBufSize], VOF+1, fail);
2792
            c->alpPixBuf[i] = c->alpPixBuf[i+c->vLumBufSize];
2793
        }
2794

    
2795
    //try to avoid drawing green stuff between the right end and the stride end
2796
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2797

    
2798
    assert(2*VOFW == VOF);
2799

    
2800
    assert(c->chrDstH <= dstH);
2801

    
2802
    if (flags&SWS_PRINT_INFO) {
2803
        if (flags&SWS_FAST_BILINEAR)
2804
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2805
        else if (flags&SWS_BILINEAR)
2806
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2807
        else if (flags&SWS_BICUBIC)
2808
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2809
        else if (flags&SWS_X)
2810
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2811
        else if (flags&SWS_POINT)
2812
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2813
        else if (flags&SWS_AREA)
2814
            av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
2815
        else if (flags&SWS_BICUBLIN)
2816
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2817
        else if (flags&SWS_GAUSS)
2818
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2819
        else if (flags&SWS_SINC)
2820
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2821
        else if (flags&SWS_LANCZOS)
2822
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2823
        else if (flags&SWS_SPLINE)
2824
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2825
        else
2826
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2827

    
2828
        av_log(c, AV_LOG_INFO, "from %s to %s%s ",
2829
               sws_format_name(srcFormat),
2830
#ifdef DITHER1XBPP
2831
               dstFormat == PIX_FMT_BGR555 || dstFormat == PIX_FMT_BGR565 ? "dithered " : "",
2832
#else
2833
               "",
2834
#endif
2835
               sws_format_name(dstFormat));
2836

    
2837
        if (flags & SWS_CPU_CAPS_MMX2)
2838
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2839
        else if (flags & SWS_CPU_CAPS_3DNOW)
2840
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2841
        else if (flags & SWS_CPU_CAPS_MMX)
2842
            av_log(c, AV_LOG_INFO, "using MMX\n");
2843
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2844
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2845
        else
2846
            av_log(c, AV_LOG_INFO, "using C\n");
2847
    }
2848

    
2849
    if (flags & SWS_PRINT_INFO) {
2850
        if (flags & SWS_CPU_CAPS_MMX) {
2851
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2852
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2853
            else {
2854
                if (c->hLumFilterSize==4)
2855
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2856
                else if (c->hLumFilterSize==8)
2857
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2858
                else
2859
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2860

    
2861
                if (c->hChrFilterSize==4)
2862
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2863
                else if (c->hChrFilterSize==8)
2864
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2865
                else
2866
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2867
            }
2868
        } else {
2869
#if ARCH_X86
2870
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
2871
#else
2872
            if (flags & SWS_FAST_BILINEAR)
2873
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
2874
            else
2875
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
2876
#endif
2877
        }
2878
        if (isPlanarYUV(dstFormat)) {
2879
            if (c->vLumFilterSize==1)
2880
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2881
            else
2882
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2883
        } else {
2884
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
2885
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2886
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2887
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
2888
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2889
            else
2890
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2891
        }
2892

    
2893
        if (dstFormat==PIX_FMT_BGR24)
2894
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
2895
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2896
        else if (dstFormat==PIX_FMT_RGB32)
2897
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2898
        else if (dstFormat==PIX_FMT_BGR565)
2899
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2900
        else if (dstFormat==PIX_FMT_BGR555)
2901
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2902

    
2903
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2904
    }
2905
    if (flags & SWS_PRINT_INFO) {
2906
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2907
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2908
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2909
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2910
    }
2911

    
2912
    c->swScale= getSwsFunc(c);
2913
    return c;
2914

    
2915
fail:
2916
    sws_freeContext(c);
2917
    return NULL;
2918
}
2919

    
2920
static void reset_ptr(const uint8_t* src[], int format)
2921
{
2922
    if(!isALPHA(format))
2923
        src[3]=NULL;
2924
    if(!isPlanarYUV(format)) {
2925
        src[3]=src[2]=NULL;
2926

    
2927
        if (!usePal(format))
2928
            src[1]= NULL;
2929
    }
2930
}
2931

    
2932
/**
2933
 * swscale wrapper, so we don't need to export the SwsContext.
2934
 * Assumes planar YUV to be in YUV order instead of YVU.
2935
 */
2936
int sws_scale(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
2937
              int srcSliceH, uint8_t* dst[], int dstStride[])
2938
{
2939
    int i;
2940
    const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
2941
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
2942

    
2943
    // do not mess up sliceDir if we have a "trailing" 0-size slice
2944
    if (srcSliceH == 0)
2945
        return 0;
2946

    
2947
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2948
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2949
        return 0;
2950
    }
2951
    if (c->sliceDir == 0) {
2952
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2953
    }
2954

    
2955
    if (usePal(c->srcFormat)) {
2956
        for (i=0; i<256; i++) {
2957
            int p, r, g, b,y,u,v;
2958
            if(c->srcFormat == PIX_FMT_PAL8) {
2959
                p=((uint32_t*)(src[1]))[i];
2960
                r= (p>>16)&0xFF;
2961
                g= (p>> 8)&0xFF;
2962
                b=  p     &0xFF;
2963
            } else if(c->srcFormat == PIX_FMT_RGB8) {
2964
                r= (i>>5    )*36;
2965
                g= ((i>>2)&7)*36;
2966
                b= (i&3     )*85;
2967
            } else if(c->srcFormat == PIX_FMT_BGR8) {
2968
                b= (i>>6    )*85;
2969
                g= ((i>>3)&7)*36;
2970
                r= (i&7     )*36;
2971
            } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
2972
                r= (i>>3    )*255;
2973
                g= ((i>>1)&3)*85;
2974
                b= (i&1     )*255;
2975
            } else {
2976
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2977
                b= (i>>3    )*255;
2978
                g= ((i>>1)&3)*85;
2979
                r= (i&1     )*255;
2980
            }
2981
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2982
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2983
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2984
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
2985

    
2986

    
2987
            switch(c->dstFormat) {
2988
            case PIX_FMT_BGR32:
2989
#if !HAVE_BIGENDIAN
2990
            case PIX_FMT_RGB24:
2991
#endif
2992
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
2993
                break;
2994
            case PIX_FMT_BGR32_1:
2995
#if HAVE_BIGENDIAN
2996
            case PIX_FMT_BGR24:
2997
#endif
2998
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2999
                break;
3000
            case PIX_FMT_RGB32_1:
3001
#if HAVE_BIGENDIAN
3002
            case PIX_FMT_RGB24:
3003
#endif
3004
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3005
                break;
3006
            case PIX_FMT_RGB32:
3007
#if !HAVE_BIGENDIAN
3008
            case PIX_FMT_BGR24:
3009
#endif
3010
            default:
3011
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
3012
            }
3013
        }
3014
    }
3015

    
3016
    // copy strides, so they can safely be modified
3017
    if (c->sliceDir == 1) {
3018
        // slices go from top to bottom
3019
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3020
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3021

    
3022
        reset_ptr(src2, c->srcFormat);
3023
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
3024

    
3025
        /* reset slice direction at end of frame */
3026
        if (srcSliceY + srcSliceH == c->srcH)
3027
            c->sliceDir = 0;
3028

    
3029
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
3030
    } else {
3031
        // slices go from bottom to top => we flip the image internally
3032
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3033
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3034

    
3035
        src2[0] += (srcSliceH-1)*srcStride[0];
3036
        if (!usePal(c->srcFormat))
3037
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3038
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3039
        src2[3] += (srcSliceH-1)*srcStride[3];
3040
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
3041
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3042
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3043
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
3044

    
3045
        reset_ptr(src2, c->srcFormat);
3046
        reset_ptr((const uint8_t**)dst2, c->dstFormat);
3047

    
3048
        /* reset slice direction at end of frame */
3049
        if (!srcSliceY)
3050
            c->sliceDir = 0;
3051

    
3052
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3053
    }
3054
}
3055

    
3056
#if LIBSWSCALE_VERSION_MAJOR < 1
3057
int sws_scale_ordered(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
3058
                      int srcSliceH, uint8_t* dst[], int dstStride[])
3059
{
3060
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3061
}
3062
#endif
3063

    
3064
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3065
                                float lumaSharpen, float chromaSharpen,
3066
                                float chromaHShift, float chromaVShift,
3067
                                int verbose)
3068
{
3069
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3070
    if (!filter)
3071
        return NULL;
3072

    
3073
    if (lumaGBlur!=0.0) {
3074
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3075
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3076
    } else {
3077
        filter->lumH= sws_getIdentityVec();
3078
        filter->lumV= sws_getIdentityVec();
3079
    }
3080

    
3081
    if (chromaGBlur!=0.0) {
3082
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3083
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3084
    } else {
3085
        filter->chrH= sws_getIdentityVec();
3086
        filter->chrV= sws_getIdentityVec();
3087
    }
3088

    
3089
    if (chromaSharpen!=0.0) {
3090
        SwsVector *id= sws_getIdentityVec();
3091
        sws_scaleVec(filter->chrH, -chromaSharpen);
3092
        sws_scaleVec(filter->chrV, -chromaSharpen);
3093
        sws_addVec(filter->chrH, id);
3094
        sws_addVec(filter->chrV, id);
3095
        sws_freeVec(id);
3096
    }
3097

    
3098
    if (lumaSharpen!=0.0) {
3099
        SwsVector *id= sws_getIdentityVec();
3100
        sws_scaleVec(filter->lumH, -lumaSharpen);
3101
        sws_scaleVec(filter->lumV, -lumaSharpen);
3102
        sws_addVec(filter->lumH, id);
3103
        sws_addVec(filter->lumV, id);
3104
        sws_freeVec(id);
3105
    }
3106

    
3107
    if (chromaHShift != 0.0)
3108
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3109

    
3110
    if (chromaVShift != 0.0)
3111
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3112

    
3113
    sws_normalizeVec(filter->chrH, 1.0);
3114
    sws_normalizeVec(filter->chrV, 1.0);
3115
    sws_normalizeVec(filter->lumH, 1.0);
3116
    sws_normalizeVec(filter->lumV, 1.0);
3117

    
3118
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3119
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3120

    
3121
    return filter;
3122
}
3123

    
3124
SwsVector *sws_allocVec(int length)
3125
{
3126
    SwsVector *vec = av_malloc(sizeof(SwsVector));
3127
    if (!vec)
3128
        return NULL;
3129
    vec->length = length;
3130
    vec->coeff  = av_malloc(sizeof(double) * length);
3131
    if (!vec->coeff)
3132
        av_freep(&vec);
3133
    return vec;
3134
}
3135

    
3136
SwsVector *sws_getGaussianVec(double variance, double quality)
3137
{
3138
    const int length= (int)(variance*quality + 0.5) | 1;
3139
    int i;
3140
    double middle= (length-1)*0.5;
3141
    SwsVector *vec= sws_allocVec(length);
3142

    
3143
    if (!vec)
3144
        return NULL;
3145

    
3146
    for (i=0; i<length; i++) {
3147
        double dist= i-middle;
3148
        vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3149
    }
3150

    
3151
    sws_normalizeVec(vec, 1.0);
3152

    
3153
    return vec;
3154
}
3155

    
3156
SwsVector *sws_getConstVec(double c, int length)
3157
{
3158
    int i;
3159
    SwsVector *vec= sws_allocVec(length);
3160

    
3161
    if (!vec)
3162
        return NULL;
3163

    
3164
    for (i=0; i<length; i++)
3165
        vec->coeff[i]= c;
3166

    
3167
    return vec;
3168
}
3169

    
3170

    
3171
SwsVector *sws_getIdentityVec(void)
3172
{
3173
    return sws_getConstVec(1.0, 1);
3174
}
3175

    
3176
double sws_dcVec(SwsVector *a)
3177
{
3178
    int i;
3179
    double sum=0;
3180

    
3181
    for (i=0; i<a->length; i++)
3182
        sum+= a->coeff[i];
3183

    
3184
    return sum;
3185
}
3186

    
3187
void sws_scaleVec(SwsVector *a, double scalar)
3188
{
3189
    int i;
3190

    
3191
    for (i=0; i<a->length; i++)
3192
        a->coeff[i]*= scalar;
3193
}
3194

    
3195
void sws_normalizeVec(SwsVector *a, double height)
3196
{
3197
    sws_scaleVec(a, height/sws_dcVec(a));
3198
}
3199

    
3200
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
3201
{
3202
    int length= a->length + b->length - 1;
3203
    int i, j;
3204
    SwsVector *vec= sws_getConstVec(0.0, length);
3205

    
3206
    if (!vec)
3207
        return NULL;
3208

    
3209
    for (i=0; i<a->length; i++) {
3210
        for (j=0; j<b->length; j++) {
3211
            vec->coeff[i+j]+= a->coeff[i]*b->coeff[j];
3212
        }
3213
    }
3214

    
3215
    return vec;
3216
}
3217

    
3218
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
3219
{
3220
    int length= FFMAX(a->length, b->length);
3221
    int i;
3222
    SwsVector *vec= sws_getConstVec(0.0, length);
3223

    
3224
    if (!vec)
3225
        return NULL;
3226

    
3227
    for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3228
    for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3229

    
3230
    return vec;
3231
}
3232

    
3233
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
3234
{
3235
    int length= FFMAX(a->length, b->length);
3236
    int i;
3237
    SwsVector *vec= sws_getConstVec(0.0, length);
3238

    
3239
    if (!vec)
3240
        return NULL;
3241

    
3242
    for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3243
    for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3244

    
3245
    return vec;
3246
}
3247

    
3248
/* shift left / or right if "shift" is negative */
3249
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
3250
{
3251
    int length= a->length + FFABS(shift)*2;
3252
    int i;
3253
    SwsVector *vec= sws_getConstVec(0.0, length);
3254

    
3255
    if (!vec)
3256
        return NULL;
3257

    
3258
    for (i=0; i<a->length; i++) {
3259
        vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3260
    }
3261

    
3262
    return vec;
3263
}
3264

    
3265
void sws_shiftVec(SwsVector *a, int shift)
3266
{
3267
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3268
    av_free(a->coeff);
3269
    a->coeff= shifted->coeff;
3270
    a->length= shifted->length;
3271
    av_free(shifted);
3272
}
3273

    
3274
void sws_addVec(SwsVector *a, SwsVector *b)
3275
{
3276
    SwsVector *sum= sws_sumVec(a, b);
3277
    av_free(a->coeff);
3278
    a->coeff= sum->coeff;
3279
    a->length= sum->length;
3280
    av_free(sum);
3281
}
3282

    
3283
void sws_subVec(SwsVector *a, SwsVector *b)
3284
{
3285
    SwsVector *diff= sws_diffVec(a, b);
3286
    av_free(a->coeff);
3287
    a->coeff= diff->coeff;
3288
    a->length= diff->length;
3289
    av_free(diff);
3290
}
3291

    
3292
void sws_convVec(SwsVector *a, SwsVector *b)
3293
{
3294
    SwsVector *conv= sws_getConvVec(a, b);
3295
    av_free(a->coeff);
3296
    a->coeff= conv->coeff;
3297
    a->length= conv->length;
3298
    av_free(conv);
3299
}
3300

    
3301
SwsVector *sws_cloneVec(SwsVector *a)
3302
{
3303
    int i;
3304
    SwsVector *vec= sws_allocVec(a->length);
3305

    
3306
    if (!vec)
3307
        return NULL;
3308

    
3309
    for (i=0; i<a->length; i++) vec->coeff[i]= a->coeff[i];
3310

    
3311
    return vec;
3312
}
3313

    
3314
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
3315
{
3316
    int i;
3317
    double max=0;
3318
    double min=0;
3319
    double range;
3320

    
3321
    for (i=0; i<a->length; i++)
3322
        if (a->coeff[i]>max) max= a->coeff[i];
3323

    
3324
    for (i=0; i<a->length; i++)
3325
        if (a->coeff[i]<min) min= a->coeff[i];
3326

    
3327
    range= max - min;
3328

    
3329
    for (i=0; i<a->length; i++) {
3330
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3331
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3332
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3333
        av_log(log_ctx, log_level, "|\n");
3334
    }
3335
}
3336

    
3337
#if LIBSWSCALE_VERSION_MAJOR < 1
3338
void sws_printVec(SwsVector *a)
3339
{
3340
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3341
}
3342
#endif
3343

    
3344
void sws_freeVec(SwsVector *a)
3345
{
3346
    if (!a) return;
3347
    av_freep(&a->coeff);
3348
    a->length=0;
3349
    av_free(a);
3350
}
3351

    
3352
void sws_freeFilter(SwsFilter *filter)
3353
{
3354
    if (!filter) return;
3355

    
3356
    if (filter->lumH) sws_freeVec(filter->lumH);
3357
    if (filter->lumV) sws_freeVec(filter->lumV);
3358
    if (filter->chrH) sws_freeVec(filter->chrH);
3359
    if (filter->chrV) sws_freeVec(filter->chrV);
3360
    av_free(filter);
3361
}
3362

    
3363

    
3364
void sws_freeContext(SwsContext *c)
3365
{
3366
    int i;
3367
    if (!c) return;
3368

    
3369
    if (c->lumPixBuf) {
3370
        for (i=0; i<c->vLumBufSize; i++)
3371
            av_freep(&c->lumPixBuf[i]);
3372
        av_freep(&c->lumPixBuf);
3373
    }
3374

    
3375
    if (c->chrPixBuf) {
3376
        for (i=0; i<c->vChrBufSize; i++)
3377
            av_freep(&c->chrPixBuf[i]);
3378
        av_freep(&c->chrPixBuf);
3379
    }
3380

    
3381
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
3382
        for (i=0; i<c->vLumBufSize; i++)
3383
            av_freep(&c->alpPixBuf[i]);
3384
        av_freep(&c->alpPixBuf);
3385
    }
3386

    
3387
    av_freep(&c->vLumFilter);
3388
    av_freep(&c->vChrFilter);
3389
    av_freep(&c->hLumFilter);
3390
    av_freep(&c->hChrFilter);
3391
#ifdef COMPILE_ALTIVEC
3392
    av_freep(&c->vYCoeffsBank);
3393
    av_freep(&c->vCCoeffsBank);
3394
#endif
3395

    
3396
    av_freep(&c->vLumFilterPos);
3397
    av_freep(&c->vChrFilterPos);
3398
    av_freep(&c->hLumFilterPos);
3399
    av_freep(&c->hChrFilterPos);
3400

    
3401
#if ARCH_X86 && CONFIG_GPL
3402
#ifdef MAP_ANONYMOUS
3403
    if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
3404
    if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
3405
#elif HAVE_VIRTUALALLOC
3406
    if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);
3407
    if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);
3408
#else
3409
    av_free(c->lumMmx2FilterCode);
3410
    av_free(c->chrMmx2FilterCode);
3411
#endif
3412
    c->lumMmx2FilterCode=NULL;
3413
    c->chrMmx2FilterCode=NULL;
3414
#endif /* ARCH_X86 && CONFIG_GPL */
3415

    
3416
    av_freep(&c->yuvTable);
3417

    
3418
    av_free(c);
3419
}
3420

    
3421
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3422
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3423
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3424
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
3425
{
3426
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3427

    
3428
    if (!param)
3429
        param = default_param;
3430

    
3431
    if (context) {
3432
        if (context->srcW != srcW || context->srcH != srcH ||
3433
            context->srcFormat != srcFormat ||
3434
            context->dstW != dstW || context->dstH != dstH ||
3435
            context->dstFormat != dstFormat || context->flags != flags ||
3436
            context->param[0] != param[0] || context->param[1] != param[1])
3437
        {
3438
            sws_freeContext(context);
3439
            context = NULL;
3440
        }
3441
    }
3442
    if (!context) {
3443
        return sws_getContext(srcW, srcH, srcFormat,
3444
                              dstW, dstH, dstFormat, flags,
3445
                              srcFilter, dstFilter, param);
3446
    }
3447
    return context;
3448
}
3449