Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 07287788

History | View | Annotate | Download (123 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/x86_cpu.h"
78
#include "libavutil/bswap.h"
79

    
80
unsigned swscale_version(void)
81
{
82
    return LIBSWSCALE_VERSION_INT;
83
}
84

    
85
#undef MOVNTQ
86
#undef PAVGB
87

    
88
//#undef HAVE_MMX2
89
//#define HAVE_AMD3DNOW
90
//#undef HAVE_MMX
91
//#undef ARCH_X86
92
#define DITHER1XBPP
93

    
94
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
95

    
96
#define RET 0xC3 //near return opcode for x86
97

    
98
#ifdef M_PI
99
#define PI M_PI
100
#else
101
#define PI 3.14159265358979323846
102
#endif
103

    
104
#define isSupportedIn(x)    (       \
105
           (x)==PIX_FMT_YUV420P     \
106
        || (x)==PIX_FMT_YUVA420P    \
107
        || (x)==PIX_FMT_YUYV422     \
108
        || (x)==PIX_FMT_UYVY422     \
109
        || (x)==PIX_FMT_RGB48BE     \
110
        || (x)==PIX_FMT_RGB48LE     \
111
        || (x)==PIX_FMT_RGB32       \
112
        || (x)==PIX_FMT_RGB32_1     \
113
        || (x)==PIX_FMT_BGR24       \
114
        || (x)==PIX_FMT_BGR565      \
115
        || (x)==PIX_FMT_BGR555      \
116
        || (x)==PIX_FMT_BGR32       \
117
        || (x)==PIX_FMT_BGR32_1     \
118
        || (x)==PIX_FMT_RGB24       \
119
        || (x)==PIX_FMT_RGB565      \
120
        || (x)==PIX_FMT_RGB555      \
121
        || (x)==PIX_FMT_GRAY8       \
122
        || (x)==PIX_FMT_YUV410P     \
123
        || (x)==PIX_FMT_YUV440P     \
124
        || (x)==PIX_FMT_GRAY16BE    \
125
        || (x)==PIX_FMT_GRAY16LE    \
126
        || (x)==PIX_FMT_YUV444P     \
127
        || (x)==PIX_FMT_YUV422P     \
128
        || (x)==PIX_FMT_YUV411P     \
129
        || (x)==PIX_FMT_PAL8        \
130
        || (x)==PIX_FMT_BGR8        \
131
        || (x)==PIX_FMT_RGB8        \
132
        || (x)==PIX_FMT_BGR4_BYTE   \
133
        || (x)==PIX_FMT_RGB4_BYTE   \
134
        || (x)==PIX_FMT_YUV440P     \
135
        || (x)==PIX_FMT_MONOWHITE   \
136
        || (x)==PIX_FMT_MONOBLACK   \
137
        || (x)==PIX_FMT_YUV420PLE   \
138
        || (x)==PIX_FMT_YUV422PLE   \
139
        || (x)==PIX_FMT_YUV444PLE   \
140
        || (x)==PIX_FMT_YUV420PBE   \
141
        || (x)==PIX_FMT_YUV422PBE   \
142
        || (x)==PIX_FMT_YUV444PBE   \
143
    )
144
#define isSupportedOut(x)   (       \
145
           (x)==PIX_FMT_YUV420P     \
146
        || (x)==PIX_FMT_YUVA420P    \
147
        || (x)==PIX_FMT_YUYV422     \
148
        || (x)==PIX_FMT_UYVY422     \
149
        || (x)==PIX_FMT_YUV444P     \
150
        || (x)==PIX_FMT_YUV422P     \
151
        || (x)==PIX_FMT_YUV411P     \
152
        || isRGB(x)                 \
153
        || isBGR(x)                 \
154
        || (x)==PIX_FMT_NV12        \
155
        || (x)==PIX_FMT_NV21        \
156
        || (x)==PIX_FMT_GRAY16BE    \
157
        || (x)==PIX_FMT_GRAY16LE    \
158
        || (x)==PIX_FMT_GRAY8       \
159
        || (x)==PIX_FMT_YUV410P     \
160
        || (x)==PIX_FMT_YUV440P     \
161
        || (x)==PIX_FMT_YUV420PLE   \
162
        || (x)==PIX_FMT_YUV422PLE   \
163
        || (x)==PIX_FMT_YUV444PLE   \
164
        || (x)==PIX_FMT_YUV420PBE   \
165
        || (x)==PIX_FMT_YUV422PBE   \
166
        || (x)==PIX_FMT_YUV444PBE   \
167
    )
168
#define isPacked(x)         (       \
169
           (x)==PIX_FMT_PAL8        \
170
        || (x)==PIX_FMT_YUYV422     \
171
        || (x)==PIX_FMT_UYVY422     \
172
        || isRGB(x)                 \
173
        || isBGR(x)                 \
174
    )
175
#define usePal(x)           (       \
176
           (x)==PIX_FMT_PAL8        \
177
        || (x)==PIX_FMT_BGR4_BYTE   \
178
        || (x)==PIX_FMT_RGB4_BYTE   \
179
        || (x)==PIX_FMT_BGR8        \
180
        || (x)==PIX_FMT_RGB8        \
181
    )
182

    
183
#define RGB2YUV_SHIFT 15
184
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
185
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
186
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
187
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
188
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
189
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
190
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
191
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
192
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
193

    
194
extern const int32_t ff_yuv2rgb_coeffs[8][4];
195

    
196
static const double rgb2yuv_table[8][9]={
197
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
198
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
199
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
200
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
201
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
202
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
203
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
204
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
205
};
206

    
207
/*
208
NOTES
209
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
210

211
TODO
212
more intelligent misalignment avoidance for the horizontal scaler
213
write special vertical cubic upscale version
214
optimize C code (YV12 / minmax)
215
add support for packed pixel YUV input & output
216
add support for Y8 output
217
optimize BGR24 & BGR32
218
add BGR4 output support
219
write special BGR->BGR scaler
220
*/
221

    
222
#if ARCH_X86 && CONFIG_GPL
223
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
224
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
225
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
226
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
227
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
228
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
229
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
230
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
231

    
232
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
233
        0x0103010301030103LL,
234
        0x0200020002000200LL,};
235

    
236
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
237
        0x0602060206020602LL,
238
        0x0004000400040004LL,};
239

    
240
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
241
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
242
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
243
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
244
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
245
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
246

    
247
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
248
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
249
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
250

    
251
#ifdef FAST_BGR2YV12
252
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
253
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
254
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
255
#else
256
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
257
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
258
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
259
#endif /* FAST_BGR2YV12 */
260
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
261
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
262
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
263

    
264
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
265
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
266
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
267
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
268
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
269

    
270
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
271
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
272
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
273
};
274

    
275
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
276

    
277
#endif /* ARCH_X86 && CONFIG_GPL */
278

    
279
// clipping helper table for C implementations:
280
static unsigned char clip_table[768];
281

    
282
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
283

    
284
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4[2][8])={
285
{  1,   3,   1,   3,   1,   3,   1,   3, },
286
{  2,   0,   2,   0,   2,   0,   2,   0, },
287
};
288

    
289
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8[2][8])={
290
{  6,   2,   6,   2,   6,   2,   6,   2, },
291
{  0,   4,   0,   4,   0,   4,   0,   4, },
292
};
293

    
294
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32[8][8])={
295
{ 17,   9,  23,  15,  16,   8,  22,  14, },
296
{  5,  29,   3,  27,   4,  28,   2,  26, },
297
{ 21,  13,  19,  11,  20,  12,  18,  10, },
298
{  0,  24,   6,  30,   1,  25,   7,  31, },
299
{ 16,   8,  22,  14,  17,   9,  23,  15, },
300
{  4,  28,   2,  26,   5,  29,   3,  27, },
301
{ 20,  12,  18,  10,  21,  13,  19,  11, },
302
{  1,  25,   7,  31,   0,  24,   6,  30, },
303
};
304

    
305
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73[8][8])={
306
{  0,  55,  14,  68,   3,  58,  17,  72, },
307
{ 37,  18,  50,  32,  40,  22,  54,  35, },
308
{  9,  64,   5,  59,  13,  67,   8,  63, },
309
{ 46,  27,  41,  23,  49,  31,  44,  26, },
310
{  2,  57,  16,  71,   1,  56,  15,  70, },
311
{ 39,  21,  52,  34,  38,  19,  51,  33, },
312
{ 11,  66,   7,  62,  10,  65,   6,  60, },
313
{ 48,  30,  43,  25,  47,  29,  42,  24, },
314
};
315

    
316
#if 1
317
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
318
{117,  62, 158, 103, 113,  58, 155, 100, },
319
{ 34, 199,  21, 186,  31, 196,  17, 182, },
320
{144,  89, 131,  76, 141,  86, 127,  72, },
321
{  0, 165,  41, 206,  10, 175,  52, 217, },
322
{110,  55, 151,  96, 120,  65, 162, 107, },
323
{ 28, 193,  14, 179,  38, 203,  24, 189, },
324
{138,  83, 124,  69, 148,  93, 134,  79, },
325
{  7, 172,  48, 213,   3, 168,  45, 210, },
326
};
327
#elif 1
328
// tries to correct a gamma of 1.5
329
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
330
{  0, 143,  18, 200,   2, 156,  25, 215, },
331
{ 78,  28, 125,  64,  89,  36, 138,  74, },
332
{ 10, 180,   3, 161,  16, 195,   8, 175, },
333
{109,  51,  93,  38, 121,  60, 105,  47, },
334
{  1, 152,  23, 210,   0, 147,  20, 205, },
335
{ 85,  33, 134,  71,  81,  30, 130,  67, },
336
{ 14, 190,   6, 171,  12, 185,   5, 166, },
337
{117,  57, 101,  44, 113,  54,  97,  41, },
338
};
339
#elif 1
340
// tries to correct a gamma of 2.0
341
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
342
{  0, 124,   8, 193,   0, 140,  12, 213, },
343
{ 55,  14, 104,  42,  66,  19, 119,  52, },
344
{  3, 168,   1, 145,   6, 187,   3, 162, },
345
{ 86,  31,  70,  21,  99,  39,  82,  28, },
346
{  0, 134,  11, 206,   0, 129,   9, 200, },
347
{ 62,  17, 114,  48,  58,  16, 109,  45, },
348
{  5, 181,   2, 157,   4, 175,   1, 151, },
349
{ 95,  36,  78,  26,  90,  34,  74,  24, },
350
};
351
#else
352
// tries to correct a gamma of 2.5
353
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
354
{  0, 107,   3, 187,   0, 125,   6, 212, },
355
{ 39,   7,  86,  28,  49,  11, 102,  36, },
356
{  1, 158,   0, 131,   3, 180,   1, 151, },
357
{ 68,  19,  52,  12,  81,  25,  64,  17, },
358
{  0, 119,   5, 203,   0, 113,   4, 195, },
359
{ 45,   9,  96,  33,  42,   8,  91,  30, },
360
{  2, 172,   1, 144,   2, 165,   0, 137, },
361
{ 77,  23,  60,  15,  72,  21,  56,  14, },
362
};
363
#endif
364

    
365
const char *sws_format_name(enum PixelFormat format)
366
{
367
    switch (format) {
368
        case PIX_FMT_YUV420P:
369
            return "yuv420p";
370
        case PIX_FMT_YUVA420P:
371
            return "yuva420p";
372
        case PIX_FMT_YUYV422:
373
            return "yuyv422";
374
        case PIX_FMT_RGB24:
375
            return "rgb24";
376
        case PIX_FMT_BGR24:
377
            return "bgr24";
378
        case PIX_FMT_YUV422P:
379
            return "yuv422p";
380
        case PIX_FMT_YUV444P:
381
            return "yuv444p";
382
        case PIX_FMT_RGB32:
383
            return "rgb32";
384
        case PIX_FMT_YUV410P:
385
            return "yuv410p";
386
        case PIX_FMT_YUV411P:
387
            return "yuv411p";
388
        case PIX_FMT_RGB565:
389
            return "rgb565";
390
        case PIX_FMT_RGB555:
391
            return "rgb555";
392
        case PIX_FMT_GRAY16BE:
393
            return "gray16be";
394
        case PIX_FMT_GRAY16LE:
395
            return "gray16le";
396
        case PIX_FMT_GRAY8:
397
            return "gray8";
398
        case PIX_FMT_MONOWHITE:
399
            return "mono white";
400
        case PIX_FMT_MONOBLACK:
401
            return "mono black";
402
        case PIX_FMT_PAL8:
403
            return "Palette";
404
        case PIX_FMT_YUVJ420P:
405
            return "yuvj420p";
406
        case PIX_FMT_YUVJ422P:
407
            return "yuvj422p";
408
        case PIX_FMT_YUVJ444P:
409
            return "yuvj444p";
410
        case PIX_FMT_XVMC_MPEG2_MC:
411
            return "xvmc_mpeg2_mc";
412
        case PIX_FMT_XVMC_MPEG2_IDCT:
413
            return "xvmc_mpeg2_idct";
414
        case PIX_FMT_UYVY422:
415
            return "uyvy422";
416
        case PIX_FMT_UYYVYY411:
417
            return "uyyvyy411";
418
        case PIX_FMT_RGB32_1:
419
            return "rgb32x";
420
        case PIX_FMT_BGR32_1:
421
            return "bgr32x";
422
        case PIX_FMT_BGR32:
423
            return "bgr32";
424
        case PIX_FMT_BGR565:
425
            return "bgr565";
426
        case PIX_FMT_BGR555:
427
            return "bgr555";
428
        case PIX_FMT_BGR8:
429
            return "bgr8";
430
        case PIX_FMT_BGR4:
431
            return "bgr4";
432
        case PIX_FMT_BGR4_BYTE:
433
            return "bgr4 byte";
434
        case PIX_FMT_RGB8:
435
            return "rgb8";
436
        case PIX_FMT_RGB4:
437
            return "rgb4";
438
        case PIX_FMT_RGB4_BYTE:
439
            return "rgb4 byte";
440
        case PIX_FMT_RGB48BE:
441
            return "rgb48be";
442
        case PIX_FMT_RGB48LE:
443
            return "rgb48le";
444
        case PIX_FMT_NV12:
445
            return "nv12";
446
        case PIX_FMT_NV21:
447
            return "nv21";
448
        case PIX_FMT_YUV440P:
449
            return "yuv440p";
450
        case PIX_FMT_VDPAU_H264:
451
            return "vdpau_h264";
452
        case PIX_FMT_VDPAU_MPEG1:
453
            return "vdpau_mpeg1";
454
        case PIX_FMT_VDPAU_MPEG2:
455
            return "vdpau_mpeg2";
456
        case PIX_FMT_VDPAU_WMV3:
457
            return "vdpau_wmv3";
458
        case PIX_FMT_VDPAU_VC1:
459
            return "vdpau_vc1";
460
        case PIX_FMT_YUV420PLE:
461
            return "yuv420ple";
462
        case PIX_FMT_YUV422PLE:
463
            return "yuv422ple";
464
        case PIX_FMT_YUV444PLE:
465
            return "yuv444ple";
466
        case PIX_FMT_YUV420PBE:
467
            return "yuv420pbe";
468
        case PIX_FMT_YUV422PBE:
469
            return "yuv422pbe";
470
        case PIX_FMT_YUV444PBE:
471
            return "yuv444pbe";
472
        default:
473
            return "Unknown format";
474
    }
475
}
476

    
477
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
478
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
479
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
480
{
481
    //FIXME Optimize (just quickly written not optimized..)
482
    int i;
483
    for (i=0; i<dstW; i++)
484
    {
485
        int val=1<<18;
486
        int j;
487
        for (j=0; j<lumFilterSize; j++)
488
            val += lumSrc[j][i] * lumFilter[j];
489

    
490
        dest[i]= av_clip_uint8(val>>19);
491
    }
492

    
493
    if (uDest)
494
        for (i=0; i<chrDstW; i++)
495
        {
496
            int u=1<<18;
497
            int v=1<<18;
498
            int j;
499
            for (j=0; j<chrFilterSize; j++)
500
            {
501
                u += chrSrc[j][i] * chrFilter[j];
502
                v += chrSrc[j][i + VOFW] * chrFilter[j];
503
            }
504

    
505
            uDest[i]= av_clip_uint8(u>>19);
506
            vDest[i]= av_clip_uint8(v>>19);
507
        }
508

    
509
    if (CONFIG_SWSCALE_ALPHA && aDest)
510
        for (i=0; i<dstW; i++){
511
            int val=1<<18;
512
            int j;
513
            for (j=0; j<lumFilterSize; j++)
514
                val += alpSrc[j][i] * lumFilter[j];
515

    
516
            aDest[i]= av_clip_uint8(val>>19);
517
        }
518

    
519
}
520

    
521
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
522
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
523
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
524
{
525
    //FIXME Optimize (just quickly written not optimized..)
526
    int i;
527
    for (i=0; i<dstW; i++)
528
    {
529
        int val=1<<18;
530
        int j;
531
        for (j=0; j<lumFilterSize; j++)
532
            val += lumSrc[j][i] * lumFilter[j];
533

    
534
        dest[i]= av_clip_uint8(val>>19);
535
    }
536

    
537
    if (!uDest)
538
        return;
539

    
540
    if (dstFormat == PIX_FMT_NV12)
541
        for (i=0; i<chrDstW; i++)
542
        {
543
            int u=1<<18;
544
            int v=1<<18;
545
            int j;
546
            for (j=0; j<chrFilterSize; j++)
547
            {
548
                u += chrSrc[j][i] * chrFilter[j];
549
                v += chrSrc[j][i + VOFW] * chrFilter[j];
550
            }
551

    
552
            uDest[2*i]= av_clip_uint8(u>>19);
553
            uDest[2*i+1]= av_clip_uint8(v>>19);
554
        }
555
    else
556
        for (i=0; i<chrDstW; i++)
557
        {
558
            int u=1<<18;
559
            int v=1<<18;
560
            int j;
561
            for (j=0; j<chrFilterSize; j++)
562
            {
563
                u += chrSrc[j][i] * chrFilter[j];
564
                v += chrSrc[j][i + VOFW] * chrFilter[j];
565
            }
566

    
567
            uDest[2*i]= av_clip_uint8(v>>19);
568
            uDest[2*i+1]= av_clip_uint8(u>>19);
569
        }
570
}
571

    
572
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
573
    for (i=0; i<(dstW>>1); i++){\
574
        int j;\
575
        int Y1 = 1<<18;\
576
        int Y2 = 1<<18;\
577
        int U  = 1<<18;\
578
        int V  = 1<<18;\
579
        int av_unused A1, A2;\
580
        type av_unused *r, *b, *g;\
581
        const int i2= 2*i;\
582
        \
583
        for (j=0; j<lumFilterSize; j++)\
584
        {\
585
            Y1 += lumSrc[j][i2] * lumFilter[j];\
586
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
587
        }\
588
        for (j=0; j<chrFilterSize; j++)\
589
        {\
590
            U += chrSrc[j][i] * chrFilter[j];\
591
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
592
        }\
593
        Y1>>=19;\
594
        Y2>>=19;\
595
        U >>=19;\
596
        V >>=19;\
597
        if (alpha){\
598
            A1 = 1<<18;\
599
            A2 = 1<<18;\
600
            for (j=0; j<lumFilterSize; j++){\
601
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
602
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
603
            }\
604
            A1>>=19;\
605
            A2>>=19;\
606
        }\
607

    
608
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
609
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
610
        if ((Y1|Y2|U|V)&256)\
611
        {\
612
            if (Y1>255)   Y1=255; \
613
            else if (Y1<0)Y1=0;   \
614
            if (Y2>255)   Y2=255; \
615
            else if (Y2<0)Y2=0;   \
616
            if (U>255)    U=255;  \
617
            else if (U<0) U=0;    \
618
            if (V>255)    V=255;  \
619
            else if (V<0) V=0;    \
620
        }\
621
        if (alpha && ((A1|A2)&256)){\
622
            A1=av_clip_uint8(A1);\
623
            A2=av_clip_uint8(A2);\
624
        }
625

    
626
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
627
    for (i=0; i<dstW; i++){\
628
        int j;\
629
        int Y = 0;\
630
        int U = -128<<19;\
631
        int V = -128<<19;\
632
        int av_unused A;\
633
        int R,G,B;\
634
        \
635
        for (j=0; j<lumFilterSize; j++){\
636
            Y += lumSrc[j][i     ] * lumFilter[j];\
637
        }\
638
        for (j=0; j<chrFilterSize; j++){\
639
            U += chrSrc[j][i     ] * chrFilter[j];\
640
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
641
        }\
642
        Y >>=10;\
643
        U >>=10;\
644
        V >>=10;\
645
        if (alpha){\
646
            A = rnd;\
647
            for (j=0; j<lumFilterSize; j++)\
648
                A += alpSrc[j][i     ] * lumFilter[j];\
649
            A >>=19;\
650
            if (A&256)\
651
                A = av_clip_uint8(A);\
652
        }\
653

    
654
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
655
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
656
        Y-= c->yuv2rgb_y_offset;\
657
        Y*= c->yuv2rgb_y_coeff;\
658
        Y+= rnd;\
659
        R= Y + V*c->yuv2rgb_v2r_coeff;\
660
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
661
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
662
        if ((R|G|B)&(0xC0000000)){\
663
            if (R>=(256<<22))   R=(256<<22)-1; \
664
            else if (R<0)R=0;   \
665
            if (G>=(256<<22))   G=(256<<22)-1; \
666
            else if (G<0)G=0;   \
667
            if (B>=(256<<22))   B=(256<<22)-1; \
668
            else if (B<0)B=0;   \
669
        }\
670

    
671

    
672
#define YSCALE_YUV_2_GRAY16_C \
673
    for (i=0; i<(dstW>>1); i++){\
674
        int j;\
675
        int Y1 = 1<<18;\
676
        int Y2 = 1<<18;\
677
        int U  = 1<<18;\
678
        int V  = 1<<18;\
679
        \
680
        const int i2= 2*i;\
681
        \
682
        for (j=0; j<lumFilterSize; j++)\
683
        {\
684
            Y1 += lumSrc[j][i2] * lumFilter[j];\
685
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
686
        }\
687
        Y1>>=11;\
688
        Y2>>=11;\
689
        if ((Y1|Y2|U|V)&65536)\
690
        {\
691
            if (Y1>65535)   Y1=65535; \
692
            else if (Y1<0)Y1=0;   \
693
            if (Y2>65535)   Y2=65535; \
694
            else if (Y2<0)Y2=0;   \
695
        }
696

    
697
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
698
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
699
    r = (type *)c->table_rV[V];   \
700
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
701
    b = (type *)c->table_bU[U];   \
702

    
703
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
704
    for (i=0; i<(dstW>>1); i++){ \
705
        const int i2= 2*i;       \
706
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
707
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
708
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
709
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
710
        type av_unused *r, *b, *g;                                    \
711
        int av_unused A1, A2;                                         \
712
        if (alpha){\
713
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
714
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
715
        }\
716

    
717
#define YSCALE_YUV_2_GRAY16_2_C   \
718
    for (i=0; i<(dstW>>1); i++){ \
719
        const int i2= 2*i;       \
720
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
721
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
722

    
723
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
724
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
725
    r = (type *)c->table_rV[V];\
726
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
727
    b = (type *)c->table_bU[U];\
728

    
729
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
730
    for (i=0; i<(dstW>>1); i++){\
731
        const int i2= 2*i;\
732
        int Y1= buf0[i2  ]>>7;\
733
        int Y2= buf0[i2+1]>>7;\
734
        int U= (uvbuf1[i     ])>>7;\
735
        int V= (uvbuf1[i+VOFW])>>7;\
736
        type av_unused *r, *b, *g;\
737
        int av_unused A1, A2;\
738
        if (alpha){\
739
            A1= abuf0[i2  ]>>7;\
740
            A2= abuf0[i2+1]>>7;\
741
        }\
742

    
743
#define YSCALE_YUV_2_GRAY16_1_C \
744
    for (i=0; i<(dstW>>1); i++){\
745
        const int i2= 2*i;\
746
        int Y1= buf0[i2  ]<<1;\
747
        int Y2= buf0[i2+1]<<1;\
748

    
749
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
750
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
751
    r = (type *)c->table_rV[V];\
752
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
753
    b = (type *)c->table_bU[U];\
754

    
755
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
756
    for (i=0; i<(dstW>>1); i++){\
757
        const int i2= 2*i;\
758
        int Y1= buf0[i2  ]>>7;\
759
        int Y2= buf0[i2+1]>>7;\
760
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
761
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
762
        type av_unused *r, *b, *g;\
763
        int av_unused A1, A2;\
764
        if (alpha){\
765
            A1= abuf0[i2  ]>>7;\
766
            A2= abuf0[i2+1]>>7;\
767
        }\
768

    
769
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
770
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
771
    r = (type *)c->table_rV[V];\
772
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
773
    b = (type *)c->table_bU[U];\
774

    
775
#define YSCALE_YUV_2_MONO2_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
    for (i=0; i<dstW-7; i+=8){\
779
        int acc;\
780
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
781
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
782
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
783
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
784
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
785
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
786
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
787
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
788
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
789
        dest++;\
790
    }\
791

    
792

    
793
#define YSCALE_YUV_2_MONOX_C \
794
    const uint8_t * const d128=dither_8x8_220[y&7];\
795
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
796
    int acc=0;\
797
    for (i=0; i<dstW-1; i+=2){\
798
        int j;\
799
        int Y1=1<<18;\
800
        int Y2=1<<18;\
801
\
802
        for (j=0; j<lumFilterSize; j++)\
803
        {\
804
            Y1 += lumSrc[j][i] * lumFilter[j];\
805
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
806
        }\
807
        Y1>>=19;\
808
        Y2>>=19;\
809
        if ((Y1|Y2)&256)\
810
        {\
811
            if (Y1>255)   Y1=255;\
812
            else if (Y1<0)Y1=0;\
813
            if (Y2>255)   Y2=255;\
814
            else if (Y2<0)Y2=0;\
815
        }\
816
        acc+= acc + g[Y1+d128[(i+0)&7]];\
817
        acc+= acc + g[Y2+d128[(i+1)&7]];\
818
        if ((i&7)==6){\
819
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
820
            dest++;\
821
        }\
822
    }
823

    
824

    
825
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
826
    switch(c->dstFormat)\
827
    {\
828
    case PIX_FMT_RGB48BE:\
829
    case PIX_FMT_RGB48LE:\
830
        func(uint8_t,0)\
831
            ((uint8_t*)dest)[ 0]= r[Y1];\
832
            ((uint8_t*)dest)[ 1]= r[Y1];\
833
            ((uint8_t*)dest)[ 2]= g[Y1];\
834
            ((uint8_t*)dest)[ 3]= g[Y1];\
835
            ((uint8_t*)dest)[ 4]= b[Y1];\
836
            ((uint8_t*)dest)[ 5]= b[Y1];\
837
            ((uint8_t*)dest)[ 6]= r[Y2];\
838
            ((uint8_t*)dest)[ 7]= r[Y2];\
839
            ((uint8_t*)dest)[ 8]= g[Y2];\
840
            ((uint8_t*)dest)[ 9]= g[Y2];\
841
            ((uint8_t*)dest)[10]= b[Y2];\
842
            ((uint8_t*)dest)[11]= b[Y2];\
843
            dest+=12;\
844
        }\
845
        break;\
846
    case PIX_FMT_RGBA:\
847
    case PIX_FMT_BGRA:\
848
        if (CONFIG_SMALL){\
849
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
850
            func(uint32_t,needAlpha)\
851
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
852
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
853
            }\
854
        }else{\
855
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
856
                func(uint32_t,1)\
857
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
858
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
859
                }\
860
            }else{\
861
                func(uint32_t,0)\
862
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
863
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
864
                }\
865
            }\
866
        }\
867
        break;\
868
    case PIX_FMT_ARGB:\
869
    case PIX_FMT_ABGR:\
870
        if (CONFIG_SMALL){\
871
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
872
            func(uint32_t,needAlpha)\
873
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
874
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
875
            }\
876
        }else{\
877
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
878
                func(uint32_t,1)\
879
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
880
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
881
                }\
882
            }else{\
883
                func(uint32_t,0)\
884
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
885
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
886
                }\
887
            }\
888
        }                \
889
        break;\
890
    case PIX_FMT_RGB24:\
891
        func(uint8_t,0)\
892
            ((uint8_t*)dest)[0]= r[Y1];\
893
            ((uint8_t*)dest)[1]= g[Y1];\
894
            ((uint8_t*)dest)[2]= b[Y1];\
895
            ((uint8_t*)dest)[3]= r[Y2];\
896
            ((uint8_t*)dest)[4]= g[Y2];\
897
            ((uint8_t*)dest)[5]= b[Y2];\
898
            dest+=6;\
899
        }\
900
        break;\
901
    case PIX_FMT_BGR24:\
902
        func(uint8_t,0)\
903
            ((uint8_t*)dest)[0]= b[Y1];\
904
            ((uint8_t*)dest)[1]= g[Y1];\
905
            ((uint8_t*)dest)[2]= r[Y1];\
906
            ((uint8_t*)dest)[3]= b[Y2];\
907
            ((uint8_t*)dest)[4]= g[Y2];\
908
            ((uint8_t*)dest)[5]= r[Y2];\
909
            dest+=6;\
910
        }\
911
        break;\
912
    case PIX_FMT_RGB565:\
913
    case PIX_FMT_BGR565:\
914
        {\
915
            const int dr1= dither_2x2_8[y&1    ][0];\
916
            const int dg1= dither_2x2_4[y&1    ][0];\
917
            const int db1= dither_2x2_8[(y&1)^1][0];\
918
            const int dr2= dither_2x2_8[y&1    ][1];\
919
            const int dg2= dither_2x2_4[y&1    ][1];\
920
            const int db2= dither_2x2_8[(y&1)^1][1];\
921
            func(uint16_t,0)\
922
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
923
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
924
            }\
925
        }\
926
        break;\
927
    case PIX_FMT_RGB555:\
928
    case PIX_FMT_BGR555:\
929
        {\
930
            const int dr1= dither_2x2_8[y&1    ][0];\
931
            const int dg1= dither_2x2_8[y&1    ][1];\
932
            const int db1= dither_2x2_8[(y&1)^1][0];\
933
            const int dr2= dither_2x2_8[y&1    ][1];\
934
            const int dg2= dither_2x2_8[y&1    ][0];\
935
            const int db2= dither_2x2_8[(y&1)^1][1];\
936
            func(uint16_t,0)\
937
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
938
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
939
            }\
940
        }\
941
        break;\
942
    case PIX_FMT_RGB8:\
943
    case PIX_FMT_BGR8:\
944
        {\
945
            const uint8_t * const d64= dither_8x8_73[y&7];\
946
            const uint8_t * const d32= dither_8x8_32[y&7];\
947
            func(uint8_t,0)\
948
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
949
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
950
            }\
951
        }\
952
        break;\
953
    case PIX_FMT_RGB4:\
954
    case PIX_FMT_BGR4:\
955
        {\
956
            const uint8_t * const d64= dither_8x8_73 [y&7];\
957
            const uint8_t * const d128=dither_8x8_220[y&7];\
958
            func(uint8_t,0)\
959
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
960
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
961
            }\
962
        }\
963
        break;\
964
    case PIX_FMT_RGB4_BYTE:\
965
    case PIX_FMT_BGR4_BYTE:\
966
        {\
967
            const uint8_t * const d64= dither_8x8_73 [y&7];\
968
            const uint8_t * const d128=dither_8x8_220[y&7];\
969
            func(uint8_t,0)\
970
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
971
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
972
            }\
973
        }\
974
        break;\
975
    case PIX_FMT_MONOBLACK:\
976
    case PIX_FMT_MONOWHITE:\
977
        {\
978
            func_monoblack\
979
        }\
980
        break;\
981
    case PIX_FMT_YUYV422:\
982
        func2\
983
            ((uint8_t*)dest)[2*i2+0]= Y1;\
984
            ((uint8_t*)dest)[2*i2+1]= U;\
985
            ((uint8_t*)dest)[2*i2+2]= Y2;\
986
            ((uint8_t*)dest)[2*i2+3]= V;\
987
        }                \
988
        break;\
989
    case PIX_FMT_UYVY422:\
990
        func2\
991
            ((uint8_t*)dest)[2*i2+0]= U;\
992
            ((uint8_t*)dest)[2*i2+1]= Y1;\
993
            ((uint8_t*)dest)[2*i2+2]= V;\
994
            ((uint8_t*)dest)[2*i2+3]= Y2;\
995
        }                \
996
        break;\
997
    case PIX_FMT_GRAY16BE:\
998
        func_g16\
999
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
1000
            ((uint8_t*)dest)[2*i2+1]= Y1;\
1001
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
1002
            ((uint8_t*)dest)[2*i2+3]= Y2;\
1003
        }                \
1004
        break;\
1005
    case PIX_FMT_GRAY16LE:\
1006
        func_g16\
1007
            ((uint8_t*)dest)[2*i2+0]= Y1;\
1008
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
1009
            ((uint8_t*)dest)[2*i2+2]= Y2;\
1010
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
1011
        }                \
1012
        break;\
1013
    }\
1014

    
1015

    
1016
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1017
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1018
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1019
{
1020
    int i;
1021
    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)
1022
}
1023

    
1024
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1025
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1026
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1027
{
1028
    int i;
1029
    int step= fmt_depth(c->dstFormat)/8;
1030
    int aidx= 3;
1031

    
1032
    switch(c->dstFormat){
1033
    case PIX_FMT_ARGB:
1034
        dest++;
1035
        aidx= 0;
1036
    case PIX_FMT_RGB24:
1037
        aidx--;
1038
    case PIX_FMT_RGBA:
1039
        if (CONFIG_SMALL){
1040
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1041
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1042
                dest[aidx]= needAlpha ? A : 255;
1043
                dest[0]= R>>22;
1044
                dest[1]= G>>22;
1045
                dest[2]= B>>22;
1046
                dest+= step;
1047
            }
1048
        }else{
1049
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1050
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1051
                    dest[aidx]= A;
1052
                    dest[0]= R>>22;
1053
                    dest[1]= G>>22;
1054
                    dest[2]= B>>22;
1055
                    dest+= step;
1056
                }
1057
            }else{
1058
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1059
                    dest[aidx]= 255;
1060
                    dest[0]= R>>22;
1061
                    dest[1]= G>>22;
1062
                    dest[2]= B>>22;
1063
                    dest+= step;
1064
                }
1065
            }
1066
        }
1067
        break;
1068
    case PIX_FMT_ABGR:
1069
        dest++;
1070
        aidx= 0;
1071
    case PIX_FMT_BGR24:
1072
        aidx--;
1073
    case PIX_FMT_BGRA:
1074
        if (CONFIG_SMALL){
1075
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1076
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1077
                dest[aidx]= needAlpha ? A : 255;
1078
                dest[0]= B>>22;
1079
                dest[1]= G>>22;
1080
                dest[2]= R>>22;
1081
                dest+= step;
1082
            }
1083
        }else{
1084
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1085
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1086
                    dest[aidx]= A;
1087
                    dest[0]= B>>22;
1088
                    dest[1]= G>>22;
1089
                    dest[2]= R>>22;
1090
                    dest+= step;
1091
                }
1092
            }else{
1093
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1094
                    dest[aidx]= 255;
1095
                    dest[0]= B>>22;
1096
                    dest[1]= G>>22;
1097
                    dest[2]= R>>22;
1098
                    dest+= step;
1099
                }
1100
            }
1101
        }
1102
        break;
1103
    default:
1104
        assert(0);
1105
    }
1106
}
1107

    
1108
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val){
1109
    int i;
1110
    uint8_t *ptr = plane + stride*y;
1111
    for (i=0; i<height; i++){
1112
        memset(ptr, val, width);
1113
        ptr += stride;
1114
    }
1115
}
1116

    
1117
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)
1118
{
1119
    int i;
1120
    for (i = 0; i < width; i++) {
1121
        int r = src[i*6+0];
1122
        int g = src[i*6+2];
1123
        int b = src[i*6+4];
1124

    
1125
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1126
    }
1127
}
1128

    
1129
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1130
                             uint8_t *src1, uint8_t *src2, int width)
1131
{
1132
    int i;
1133
    assert(src1==src2);
1134
    for (i = 0; i < width; i++) {
1135
        int r = src1[6*i + 0];
1136
        int g = src1[6*i + 2];
1137
        int b = src1[6*i + 4];
1138

    
1139
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1140
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1141
    }
1142
}
1143

    
1144
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1145
                                  uint8_t *src1, uint8_t *src2, int width)
1146
{
1147
    int i;
1148
    assert(src1==src2);
1149
    for (i = 0; i < width; i++) {
1150
        int r= src1[12*i + 0] + src1[12*i + 6];
1151
        int g= src1[12*i + 2] + src1[12*i + 8];
1152
        int b= src1[12*i + 4] + src1[12*i + 10];
1153

    
1154
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1155
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1156
    }
1157
}
1158

    
1159
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1160
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1161
{\
1162
    int i;\
1163
    for (i=0; i<width; i++)\
1164
    {\
1165
        int b= (((const type*)src)[i]>>shb)&maskb;\
1166
        int g= (((const type*)src)[i]>>shg)&maskg;\
1167
        int r= (((const type*)src)[i]>>shr)&maskr;\
1168
\
1169
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1170
    }\
1171
}
1172

    
1173
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1174
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1175
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1176
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1177
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1178
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1179

    
1180
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused){
1181
    int i;
1182
    for (i=0; i<width; i++){
1183
        dst[i]= src[4*i];
1184
    }
1185
}
1186

    
1187
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1188
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1189
{\
1190
    int i;\
1191
    for (i=0; i<width; i++)\
1192
    {\
1193
        int b= (((const type*)src)[i]&maskb)>>shb;\
1194
        int g= (((const type*)src)[i]&maskg)>>shg;\
1195
        int r= (((const type*)src)[i]&maskr)>>shr;\
1196
\
1197
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1198
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1199
    }\
1200
}\
1201
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1202
{\
1203
    int i;\
1204
    for (i=0; i<width; i++)\
1205
    {\
1206
        int pix0= ((const type*)src)[2*i+0];\
1207
        int pix1= ((const type*)src)[2*i+1];\
1208
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1209
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1210
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1211
        g&= maskg|(2*maskg);\
1212
\
1213
        g>>=shg;\
1214
\
1215
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1216
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1217
    }\
1218
}
1219

    
1220
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1221
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1222
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1223
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1224
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1225
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1226

    
1227
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1228
{
1229
    int i;
1230
    for (i=0; i<width; i++)
1231
    {
1232
        int d= src[i];
1233

    
1234
        dst[i]= pal[d] & 0xFF;
1235
    }
1236
}
1237

    
1238
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1239
                           const uint8_t *src1, const uint8_t *src2,
1240
                           long width, uint32_t *pal)
1241
{
1242
    int i;
1243
    assert(src1 == src2);
1244
    for (i=0; i<width; i++)
1245
    {
1246
        int p= pal[src1[i]];
1247

    
1248
        dstU[i]= p>>8;
1249
        dstV[i]= p>>16;
1250
    }
1251
}
1252

    
1253
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1254
{
1255
    int i, j;
1256
    for (i=0; i<width/8; i++){
1257
        int d= ~src[i];
1258
        for(j=0; j<8; j++)
1259
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1260
    }
1261
}
1262

    
1263
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1264
{
1265
    int i, j;
1266
    for (i=0; i<width/8; i++){
1267
        int d= src[i];
1268
        for(j=0; j<8; j++)
1269
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1270
    }
1271
}
1272

    
1273

    
1274
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1275
//Plain C versions
1276
#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1277
#define COMPILE_C
1278
#endif
1279

    
1280
#if ARCH_PPC
1281
#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1282
#define COMPILE_ALTIVEC
1283
#endif
1284
#endif //ARCH_PPC
1285

    
1286
#if ARCH_X86
1287

    
1288
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1289
#define COMPILE_MMX
1290
#endif
1291

    
1292
#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1293
#define COMPILE_MMX2
1294
#endif
1295

    
1296
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1297
#define COMPILE_3DNOW
1298
#endif
1299
#endif //ARCH_X86
1300

    
1301
#define COMPILE_TEMPLATE_MMX 0
1302
#define COMPILE_TEMPLATE_MMX2 0
1303
#define COMPILE_TEMPLATE_AMD3DNOW 0
1304
#define COMPILE_TEMPLATE_ALTIVEC 0
1305

    
1306
#ifdef COMPILE_C
1307
#define RENAME(a) a ## _C
1308
#include "swscale_template.c"
1309
#endif
1310

    
1311
#ifdef COMPILE_ALTIVEC
1312
#undef RENAME
1313
#undef COMPILE_TEMPLATE_ALTIVEC
1314
#define COMPILE_TEMPLATE_ALTIVEC 1
1315
#define RENAME(a) a ## _altivec
1316
#include "swscale_template.c"
1317
#endif
1318

    
1319
#if ARCH_X86
1320

    
1321
//MMX versions
1322
#ifdef COMPILE_MMX
1323
#undef RENAME
1324
#undef COMPILE_TEMPLATE_MMX
1325
#undef COMPILE_TEMPLATE_MMX2
1326
#undef COMPILE_TEMPLATE_AMD3DNOW
1327
#define COMPILE_TEMPLATE_MMX 1
1328
#define COMPILE_TEMPLATE_MMX2 0
1329
#define COMPILE_TEMPLATE_AMD3DNOW 0
1330
#define RENAME(a) a ## _MMX
1331
#include "swscale_template.c"
1332
#endif
1333

    
1334
//MMX2 versions
1335
#ifdef COMPILE_MMX2
1336
#undef RENAME
1337
#undef COMPILE_TEMPLATE_MMX
1338
#undef COMPILE_TEMPLATE_MMX2
1339
#undef COMPILE_TEMPLATE_AMD3DNOW
1340
#define COMPILE_TEMPLATE_MMX 1
1341
#define COMPILE_TEMPLATE_MMX2 1
1342
#define COMPILE_TEMPLATE_AMD3DNOW 0
1343
#define RENAME(a) a ## _MMX2
1344
#include "swscale_template.c"
1345
#endif
1346

    
1347
//3DNOW versions
1348
#ifdef COMPILE_3DNOW
1349
#undef RENAME
1350
#undef COMPILE_TEMPLATE_MMX
1351
#undef COMPILE_TEMPLATE_MMX2
1352
#undef COMPILE_TEMPLATE_AMD3DNOW
1353
#define COMPILE_TEMPLATE_MMX 1
1354
#define COMPILE_TEMPLATE_MMX2 0
1355
#define COMPILE_TEMPLATE_AMD3DNOW 1
1356
#define RENAME(a) a ## _3DNow
1357
#include "swscale_template.c"
1358
#endif
1359

    
1360
#endif //ARCH_X86
1361

    
1362
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1363
{
1364
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1365
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
1366
    else                return getSplineCoeff(        0.0,
1367
                                             b+ 2.0*c + 3.0*d,
1368
                                                    c + 3.0*d,
1369
                                            -b- 3.0*c - 6.0*d,
1370
                                            dist-1.0);
1371
}
1372

    
1373
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1374
                             int srcW, int dstW, int filterAlign, int one, int flags,
1375
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1376
{
1377
    int i;
1378
    int filterSize;
1379
    int filter2Size;
1380
    int minFilterSize;
1381
    int64_t *filter=NULL;
1382
    int64_t *filter2=NULL;
1383
    const int64_t fone= 1LL<<54;
1384
    int ret= -1;
1385
#if ARCH_X86
1386
    if (flags & SWS_CPU_CAPS_MMX)
1387
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1388
#endif
1389

    
1390
    // NOTE: the +1 is for the MMX scaler which reads over the end
1391
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1392

    
1393
    if (FFABS(xInc - 0x10000) <10) // unscaled
1394
    {
1395
        int i;
1396
        filterSize= 1;
1397
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1398

    
1399
        for (i=0; i<dstW; i++)
1400
        {
1401
            filter[i*filterSize]= fone;
1402
            (*filterPos)[i]=i;
1403
        }
1404

    
1405
    }
1406
    else if (flags&SWS_POINT) // lame looking point sampling mode
1407
    {
1408
        int i;
1409
        int xDstInSrc;
1410
        filterSize= 1;
1411
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1412

    
1413
        xDstInSrc= xInc/2 - 0x8000;
1414
        for (i=0; i<dstW; i++)
1415
        {
1416
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1417

    
1418
            (*filterPos)[i]= xx;
1419
            filter[i]= fone;
1420
            xDstInSrc+= xInc;
1421
        }
1422
    }
1423
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1424
    {
1425
        int i;
1426
        int xDstInSrc;
1427
        filterSize= 2;
1428
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1429

    
1430
        xDstInSrc= xInc/2 - 0x8000;
1431
        for (i=0; i<dstW; i++)
1432
        {
1433
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1434
            int j;
1435

    
1436
            (*filterPos)[i]= xx;
1437
                //bilinear upscale / linear interpolate / area averaging
1438
                for (j=0; j<filterSize; j++)
1439
                {
1440
                    int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1441
                    if (coeff<0) coeff=0;
1442
                    filter[i*filterSize + j]= coeff;
1443
                    xx++;
1444
                }
1445
            xDstInSrc+= xInc;
1446
        }
1447
    }
1448
    else
1449
    {
1450
        int xDstInSrc;
1451
        int sizeFactor;
1452

    
1453
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1454
        else if (flags&SWS_X)            sizeFactor=  8;
1455
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1456
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1457
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1458
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1459
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1460
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1461
        else {
1462
            sizeFactor= 0; //GCC warning killer
1463
            assert(0);
1464
        }
1465

    
1466
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1467
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1468

    
1469
        if (filterSize > srcW-2) filterSize=srcW-2;
1470

    
1471
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1472

    
1473
        xDstInSrc= xInc - 0x10000;
1474
        for (i=0; i<dstW; i++)
1475
        {
1476
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1477
            int j;
1478
            (*filterPos)[i]= xx;
1479
            for (j=0; j<filterSize; j++)
1480
            {
1481
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1482
                double floatd;
1483
                int64_t coeff;
1484

    
1485
                if (xInc > 1<<16)
1486
                    d= d*dstW/srcW;
1487
                floatd= d * (1.0/(1<<30));
1488

    
1489
                if (flags & SWS_BICUBIC)
1490
                {
1491
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1492
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1493
                    int64_t dd = ( d*d)>>30;
1494
                    int64_t ddd= (dd*d)>>30;
1495

    
1496
                    if      (d < 1LL<<30)
1497
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1498
                    else if (d < 1LL<<31)
1499
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1500
                    else
1501
                        coeff=0.0;
1502
                    coeff *= fone>>(30+24);
1503
                }
1504
/*                else if (flags & SWS_X)
1505
                {
1506
                    double p= param ? param*0.01 : 0.3;
1507
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1508
                    coeff*= pow(2.0, - p*d*d);
1509
                }*/
1510
                else if (flags & SWS_X)
1511
                {
1512
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1513
                    double c;
1514

    
1515
                    if (floatd<1.0)
1516
                        c = cos(floatd*PI);
1517
                    else
1518
                        c=-1.0;
1519
                    if (c<0.0)      c= -pow(-c, A);
1520
                    else            c=  pow( c, A);
1521
                    coeff= (c*0.5 + 0.5)*fone;
1522
                }
1523
                else if (flags & SWS_AREA)
1524
                {
1525
                    int64_t d2= d - (1<<29);
1526
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1527
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1528
                    else coeff=0.0;
1529
                    coeff *= fone>>(30+16);
1530
                }
1531
                else if (flags & SWS_GAUSS)
1532
                {
1533
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1534
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1535
                }
1536
                else if (flags & SWS_SINC)
1537
                {
1538
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1539
                }
1540
                else if (flags & SWS_LANCZOS)
1541
                {
1542
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1543
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1544
                    if (floatd>p) coeff=0;
1545
                }
1546
                else if (flags & SWS_BILINEAR)
1547
                {
1548
                    coeff= (1<<30) - d;
1549
                    if (coeff<0) coeff=0;
1550
                    coeff *= fone >> 30;
1551
                }
1552
                else if (flags & SWS_SPLINE)
1553
                {
1554
                    double p=-2.196152422706632;
1555
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1556
                }
1557
                else {
1558
                    coeff= 0.0; //GCC warning killer
1559
                    assert(0);
1560
                }
1561

    
1562
                filter[i*filterSize + j]= coeff;
1563
                xx++;
1564
            }
1565
            xDstInSrc+= 2*xInc;
1566
        }
1567
    }
1568

    
1569
    /* apply src & dst Filter to filter -> filter2
1570
       av_free(filter);
1571
    */
1572
    assert(filterSize>0);
1573
    filter2Size= filterSize;
1574
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1575
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1576
    assert(filter2Size>0);
1577
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1578

    
1579
    for (i=0; i<dstW; i++)
1580
    {
1581
        int j, k;
1582

    
1583
        if(srcFilter){
1584
            for (k=0; k<srcFilter->length; k++){
1585
                for (j=0; j<filterSize; j++)
1586
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1587
            }
1588
        }else{
1589
            for (j=0; j<filterSize; j++)
1590
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1591
        }
1592
        //FIXME dstFilter
1593

    
1594
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1595
    }
1596
    av_freep(&filter);
1597

    
1598
    /* try to reduce the filter-size (step1 find size and shift left) */
1599
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1600
    minFilterSize= 0;
1601
    for (i=dstW-1; i>=0; i--)
1602
    {
1603
        int min= filter2Size;
1604
        int j;
1605
        int64_t cutOff=0.0;
1606

    
1607
        /* get rid off near zero elements on the left by shifting left */
1608
        for (j=0; j<filter2Size; j++)
1609
        {
1610
            int k;
1611
            cutOff += FFABS(filter2[i*filter2Size]);
1612

    
1613
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1614

    
1615
            /* preserve monotonicity because the core can't handle the filter otherwise */
1616
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1617

    
1618
            // move filter coefficients left
1619
            for (k=1; k<filter2Size; k++)
1620
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1621
            filter2[i*filter2Size + k - 1]= 0;
1622
            (*filterPos)[i]++;
1623
        }
1624

    
1625
        cutOff=0;
1626
        /* count near zeros on the right */
1627
        for (j=filter2Size-1; j>0; j--)
1628
        {
1629
            cutOff += FFABS(filter2[i*filter2Size + j]);
1630

    
1631
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1632
            min--;
1633
        }
1634

    
1635
        if (min>minFilterSize) minFilterSize= min;
1636
    }
1637

    
1638
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1639
        // we can handle the special case 4,
1640
        // so we don't want to go to the full 8
1641
        if (minFilterSize < 5)
1642
            filterAlign = 4;
1643

    
1644
        // We really don't want to waste our time
1645
        // doing useless computation, so fall back on
1646
        // the scalar C code for very small filters.
1647
        // Vectorizing is worth it only if you have a
1648
        // decent-sized vector.
1649
        if (minFilterSize < 3)
1650
            filterAlign = 1;
1651
    }
1652

    
1653
    if (flags & SWS_CPU_CAPS_MMX) {
1654
        // special case for unscaled vertical filtering
1655
        if (minFilterSize == 1 && filterAlign == 2)
1656
            filterAlign= 1;
1657
    }
1658

    
1659
    assert(minFilterSize > 0);
1660
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1661
    assert(filterSize > 0);
1662
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1663
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1664
        goto error;
1665
    *outFilterSize= filterSize;
1666

    
1667
    if (flags&SWS_PRINT_INFO)
1668
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1669
    /* try to reduce the filter-size (step2 reduce it) */
1670
    for (i=0; i<dstW; i++)
1671
    {
1672
        int j;
1673

    
1674
        for (j=0; j<filterSize; j++)
1675
        {
1676
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1677
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1678
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1679
                filter[i*filterSize + j]= 0;
1680
        }
1681
    }
1682

    
1683

    
1684
    //FIXME try to align filterPos if possible
1685

    
1686
    //fix borders
1687
    for (i=0; i<dstW; i++)
1688
    {
1689
        int j;
1690
        if ((*filterPos)[i] < 0)
1691
        {
1692
            // move filter coefficients left to compensate for filterPos
1693
            for (j=1; j<filterSize; j++)
1694
            {
1695
                int left= FFMAX(j + (*filterPos)[i], 0);
1696
                filter[i*filterSize + left] += filter[i*filterSize + j];
1697
                filter[i*filterSize + j]=0;
1698
            }
1699
            (*filterPos)[i]= 0;
1700
        }
1701

    
1702
        if ((*filterPos)[i] + filterSize > srcW)
1703
        {
1704
            int shift= (*filterPos)[i] + filterSize - srcW;
1705
            // move filter coefficients right to compensate for filterPos
1706
            for (j=filterSize-2; j>=0; j--)
1707
            {
1708
                int right= FFMIN(j + shift, filterSize-1);
1709
                filter[i*filterSize +right] += filter[i*filterSize +j];
1710
                filter[i*filterSize +j]=0;
1711
            }
1712
            (*filterPos)[i]= srcW - filterSize;
1713
        }
1714
    }
1715

    
1716
    // Note the +1 is for the MMX scaler which reads over the end
1717
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1718
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1719

    
1720
    /* normalize & store in outFilter */
1721
    for (i=0; i<dstW; i++)
1722
    {
1723
        int j;
1724
        int64_t error=0;
1725
        int64_t sum=0;
1726

    
1727
        for (j=0; j<filterSize; j++)
1728
        {
1729
            sum+= filter[i*filterSize + j];
1730
        }
1731
        sum= (sum + one/2)/ one;
1732
        for (j=0; j<*outFilterSize; j++)
1733
        {
1734
            int64_t v= filter[i*filterSize + j] + error;
1735
            int intV= ROUNDED_DIV(v, sum);
1736
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1737
            error= v - intV*sum;
1738
        }
1739
    }
1740

    
1741
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1742
    for (i=0; i<*outFilterSize; i++)
1743
    {
1744
        int j= dstW*(*outFilterSize);
1745
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1746
    }
1747

    
1748
    ret=0;
1749
error:
1750
    av_free(filter);
1751
    av_free(filter2);
1752
    return ret;
1753
}
1754

    
1755
#ifdef COMPILE_MMX2
1756
static void initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
1757
{
1758
    uint8_t *fragmentA;
1759
    x86_reg imm8OfPShufW1A;
1760
    x86_reg imm8OfPShufW2A;
1761
    x86_reg fragmentLengthA;
1762
    uint8_t *fragmentB;
1763
    x86_reg imm8OfPShufW1B;
1764
    x86_reg imm8OfPShufW2B;
1765
    x86_reg fragmentLengthB;
1766
    int fragmentPos;
1767

    
1768
    int xpos, i;
1769

    
1770
    // create an optimized horizontal scaling routine
1771

    
1772
    //code fragment
1773

    
1774
    __asm__ volatile(
1775
        "jmp                         9f                 \n\t"
1776
    // Begin
1777
        "0:                                             \n\t"
1778
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1779
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1780
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1781
        "punpcklbw                %%mm7, %%mm1          \n\t"
1782
        "punpcklbw                %%mm7, %%mm0          \n\t"
1783
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1784
        "1:                                             \n\t"
1785
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1786
        "2:                                             \n\t"
1787
        "psubw                    %%mm1, %%mm0          \n\t"
1788
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1789
        "pmullw                   %%mm3, %%mm0          \n\t"
1790
        "psllw                       $7, %%mm1          \n\t"
1791
        "paddw                    %%mm1, %%mm0          \n\t"
1792

    
1793
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1794

    
1795
        "add                         $8, %%"REG_a"      \n\t"
1796
    // End
1797
        "9:                                             \n\t"
1798
//        "int $3                                         \n\t"
1799
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1800
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1801
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1802
        "dec                         %1                 \n\t"
1803
        "dec                         %2                 \n\t"
1804
        "sub                         %0, %1             \n\t"
1805
        "sub                         %0, %2             \n\t"
1806
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1807
        "sub                         %0, %3             \n\t"
1808

    
1809

    
1810
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1811
        "=r" (fragmentLengthA)
1812
    );
1813

    
1814
    __asm__ volatile(
1815
        "jmp                         9f                 \n\t"
1816
    // Begin
1817
        "0:                                             \n\t"
1818
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1819
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1820
        "punpcklbw                %%mm7, %%mm0          \n\t"
1821
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1822
        "1:                                             \n\t"
1823
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1824
        "2:                                             \n\t"
1825
        "psubw                    %%mm1, %%mm0          \n\t"
1826
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1827
        "pmullw                   %%mm3, %%mm0          \n\t"
1828
        "psllw                       $7, %%mm1          \n\t"
1829
        "paddw                    %%mm1, %%mm0          \n\t"
1830

    
1831
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1832

    
1833
        "add                         $8, %%"REG_a"      \n\t"
1834
    // End
1835
        "9:                                             \n\t"
1836
//        "int                       $3                   \n\t"
1837
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1838
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1839
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1840
        "dec                         %1                 \n\t"
1841
        "dec                         %2                 \n\t"
1842
        "sub                         %0, %1             \n\t"
1843
        "sub                         %0, %2             \n\t"
1844
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1845
        "sub                         %0, %3             \n\t"
1846

    
1847

    
1848
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1849
        "=r" (fragmentLengthB)
1850
    );
1851

    
1852
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1853
    fragmentPos=0;
1854

    
1855
    for (i=0; i<dstW/numSplits; i++)
1856
    {
1857
        int xx=xpos>>16;
1858

    
1859
        if ((i&3) == 0)
1860
        {
1861
            int a=0;
1862
            int b=((xpos+xInc)>>16) - xx;
1863
            int c=((xpos+xInc*2)>>16) - xx;
1864
            int d=((xpos+xInc*3)>>16) - xx;
1865
            int inc                = (d+1<4);
1866
            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
1867
            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
1868
            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
1869
            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
1870
            int maxShift= 3-(d+inc);
1871
            int shift=0;
1872

    
1873
            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1874
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1875
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1876
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1877
            filterPos[i/2]= xx;
1878

    
1879
            memcpy(filterCode + fragmentPos, fragment, fragmentLength);
1880

    
1881
            filterCode[fragmentPos + imm8OfPShufW1]=
1882
                (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
1883
            filterCode[fragmentPos + imm8OfPShufW2]=
1884
                a | (b<<2) | (c<<4) | (d<<6);
1885

    
1886
            if (i+4-inc>=dstW) shift=maxShift; //avoid overread
1887
            else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1888

    
1889
            if (shift && i>=shift)
1890
            {
1891
                filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
1892
                filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
1893
                filterPos[i/2]-=shift;
1894
            }
1895

    
1896
            fragmentPos+= fragmentLength;
1897

    
1898
            filterCode[fragmentPos]= RET;
1899
        }
1900
        xpos+=xInc;
1901
    }
1902
    filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1903
}
1904
#endif /* COMPILE_MMX2 */
1905

    
1906
static void globalInit(void){
1907
    // generating tables:
1908
    int i;
1909
    for (i=0; i<768; i++){
1910
        int c= av_clip_uint8(i-256);
1911
        clip_table[i]=c;
1912
    }
1913
}
1914

    
1915
static SwsFunc getSwsFunc(SwsContext *c)
1916
{
1917
#if CONFIG_RUNTIME_CPUDETECT
1918
    int flags = c->flags;
1919

    
1920
#if ARCH_X86 && CONFIG_GPL
1921
    // ordered per speed fastest first
1922
    if (flags & SWS_CPU_CAPS_MMX2) {
1923
        sws_init_swScale_MMX2(c);
1924
        return swScale_MMX2;
1925
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1926
        sws_init_swScale_3DNow(c);
1927
        return swScale_3DNow;
1928
    } else if (flags & SWS_CPU_CAPS_MMX) {
1929
        sws_init_swScale_MMX(c);
1930
        return swScale_MMX;
1931
    } else {
1932
        sws_init_swScale_C(c);
1933
        return swScale_C;
1934
    }
1935

    
1936
#else
1937
#if ARCH_PPC
1938
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1939
        sws_init_swScale_altivec(c);
1940
        return swScale_altivec;
1941
    } else {
1942
        sws_init_swScale_C(c);
1943
        return swScale_C;
1944
    }
1945
#endif
1946
    sws_init_swScale_C(c);
1947
    return swScale_C;
1948
#endif /* ARCH_X86 && CONFIG_GPL */
1949
#else //CONFIG_RUNTIME_CPUDETECT
1950
#if   COMPILE_TEMPLATE_MMX2
1951
    sws_init_swScale_MMX2(c);
1952
    return swScale_MMX2;
1953
#elif COMPILE_TEMPLATE_AMD3DNOW
1954
    sws_init_swScale_3DNow(c);
1955
    return swScale_3DNow;
1956
#elif COMPILE_TEMPLATE_MMX
1957
    sws_init_swScale_MMX(c);
1958
    return swScale_MMX;
1959
#elif COMPILE_TEMPLATE_ALTIVEC
1960
    sws_init_swScale_altivec(c);
1961
    return swScale_altivec;
1962
#else
1963
    sws_init_swScale_C(c);
1964
    return swScale_C;
1965
#endif
1966
#endif //!CONFIG_RUNTIME_CPUDETECT
1967
}
1968

    
1969
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1970
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1971
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1972
    /* Copy Y plane */
1973
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1974
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1975
    else
1976
    {
1977
        int i;
1978
        const uint8_t *srcPtr= src[0];
1979
        uint8_t *dstPtr= dst;
1980
        for (i=0; i<srcSliceH; i++)
1981
        {
1982
            memcpy(dstPtr, srcPtr, c->srcW);
1983
            srcPtr+= srcStride[0];
1984
            dstPtr+= dstStride[0];
1985
        }
1986
    }
1987
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1988
    if (c->dstFormat == PIX_FMT_NV12)
1989
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1990
    else
1991
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1992

    
1993
    return srcSliceH;
1994
}
1995

    
1996
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1997
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1998
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1999

    
2000
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2001

    
2002
    return srcSliceH;
2003
}
2004

    
2005
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2006
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2007
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2008

    
2009
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2010

    
2011
    return srcSliceH;
2012
}
2013

    
2014
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2015
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2016
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2017

    
2018
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2019

    
2020
    return srcSliceH;
2021
}
2022

    
2023
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2024
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2025
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2026

    
2027
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2028

    
2029
    return srcSliceH;
2030
}
2031

    
2032
static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2033
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2034
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2035
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2036
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2037

    
2038
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2039

    
2040
    if (dstParam[3])
2041
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2042

    
2043
    return srcSliceH;
2044
}
2045

    
2046
static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2047
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2048
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2049
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2050
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2051

    
2052
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2053

    
2054
    return srcSliceH;
2055
}
2056

    
2057
static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2058
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2059
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2060
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2061
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2062

    
2063
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2064

    
2065
    if (dstParam[3])
2066
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2067

    
2068
    return srcSliceH;
2069
}
2070

    
2071
static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2072
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2073
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2074
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2075
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2076

    
2077
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2078

    
2079
    return srcSliceH;
2080
}
2081

    
2082
static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2083
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
2084
    const enum PixelFormat srcFormat= c->srcFormat;
2085
    const enum PixelFormat dstFormat= c->dstFormat;
2086
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
2087
                 const uint8_t *palette)=NULL;
2088
    int i;
2089
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2090
    uint8_t *srcPtr= src[0];
2091

    
2092
    if (!usePal(srcFormat))
2093
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2094
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2095

    
2096
    switch(dstFormat){
2097
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
2098
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
2099
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
2100
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
2101
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
2102
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
2103
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2104
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2105
    }
2106

    
2107

    
2108
    for (i=0; i<srcSliceH; i++) {
2109
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
2110
        srcPtr+= srcStride[0];
2111
        dstPtr+= dstStride[0];
2112
    }
2113

    
2114
    return srcSliceH;
2115
}
2116

    
2117
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
2118
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2119
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
2120
    const enum PixelFormat srcFormat= c->srcFormat;
2121
    const enum PixelFormat dstFormat= c->dstFormat;
2122
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2123
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2124
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2125
    const int dstId= fmt_depth(dstFormat) >> 2;
2126
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2127

    
2128
    /* BGR -> BGR */
2129
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
2130
       || (isRGB(srcFormat) && isRGB(dstFormat))){
2131
        switch(srcId | (dstId<<4)){
2132
        case 0x34: conv= rgb16to15; break;
2133
        case 0x36: conv= rgb24to15; break;
2134
        case 0x38: conv= rgb32to15; break;
2135
        case 0x43: conv= rgb15to16; break;
2136
        case 0x46: conv= rgb24to16; break;
2137
        case 0x48: conv= rgb32to16; break;
2138
        case 0x63: conv= rgb15to24; break;
2139
        case 0x64: conv= rgb16to24; break;
2140
        case 0x68: conv= rgb32to24; break;
2141
        case 0x83: conv= rgb15to32; break;
2142
        case 0x84: conv= rgb16to32; break;
2143
        case 0x86: conv= rgb24to32; break;
2144
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2145
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2146
        }
2147
    }else if (  (isBGR(srcFormat) && isRGB(dstFormat))
2148
             || (isRGB(srcFormat) && isBGR(dstFormat))){
2149
        switch(srcId | (dstId<<4)){
2150
        case 0x33: conv= rgb15tobgr15; break;
2151
        case 0x34: conv= rgb16tobgr15; break;
2152
        case 0x36: conv= rgb24tobgr15; break;
2153
        case 0x38: conv= rgb32tobgr15; break;
2154
        case 0x43: conv= rgb15tobgr16; break;
2155
        case 0x44: conv= rgb16tobgr16; break;
2156
        case 0x46: conv= rgb24tobgr16; break;
2157
        case 0x48: conv= rgb32tobgr16; break;
2158
        case 0x63: conv= rgb15tobgr24; break;
2159
        case 0x64: conv= rgb16tobgr24; break;
2160
        case 0x66: conv= rgb24tobgr24; break;
2161
        case 0x68: conv= rgb32tobgr24; break;
2162
        case 0x83: conv= rgb15tobgr32; break;
2163
        case 0x84: conv= rgb16tobgr32; break;
2164
        case 0x86: conv= rgb24tobgr32; break;
2165
        case 0x88: conv= rgb32tobgr32; break;
2166
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2167
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2168
        }
2169
    }else{
2170
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2171
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2172
    }
2173

    
2174
    if(conv)
2175
    {
2176
        uint8_t *srcPtr= src[0];
2177
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2178
            srcPtr += ALT32_CORR;
2179

    
2180
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2181
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2182
        else
2183
        {
2184
            int i;
2185
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2186

    
2187
            for (i=0; i<srcSliceH; i++)
2188
            {
2189
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
2190
                srcPtr+= srcStride[0];
2191
                dstPtr+= dstStride[0];
2192
            }
2193
        }
2194
    }
2195
    return srcSliceH;
2196
}
2197

    
2198
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2199
                              int srcSliceH, uint8_t* dst[], int dstStride[]){
2200

    
2201
    rgb24toyv12(
2202
        src[0],
2203
        dst[0]+ srcSliceY    *dstStride[0],
2204
        dst[1]+(srcSliceY>>1)*dstStride[1],
2205
        dst[2]+(srcSliceY>>1)*dstStride[2],
2206
        c->srcW, srcSliceH,
2207
        dstStride[0], dstStride[1], srcStride[0]);
2208
    if (dst[3])
2209
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2210
    return srcSliceH;
2211
}
2212

    
2213
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2214
                             int srcSliceH, uint8_t* dst[], int dstStride[]){
2215
    int i;
2216

    
2217
    /* copy Y */
2218
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2219
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2220
    else{
2221
        uint8_t *srcPtr= src[0];
2222
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2223

    
2224
        for (i=0; i<srcSliceH; i++)
2225
        {
2226
            memcpy(dstPtr, srcPtr, c->srcW);
2227
            srcPtr+= srcStride[0];
2228
            dstPtr+= dstStride[0];
2229
        }
2230
    }
2231

    
2232
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){
2233
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2234
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
2235
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2236
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
2237
    }else{
2238
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2239
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
2240
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2241
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
2242
    }
2243
    if (dst[3])
2244
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2245
    return srcSliceH;
2246
}
2247

    
2248
/* unscaled copy like stuff (assumes nearly identical formats) */
2249
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2250
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2251
{
2252
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2253
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2254
    else
2255
    {
2256
        int i;
2257
        uint8_t *srcPtr= src[0];
2258
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2259
        int length=0;
2260

    
2261
        /* universal length finder */
2262
        while(length+c->srcW <= FFABS(dstStride[0])
2263
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2264
        assert(length!=0);
2265

    
2266
        for (i=0; i<srcSliceH; i++)
2267
        {
2268
            memcpy(dstPtr, srcPtr, length);
2269
            srcPtr+= srcStride[0];
2270
            dstPtr+= dstStride[0];
2271
        }
2272
    }
2273
    return srcSliceH;
2274
}
2275

    
2276
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2277
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2278
{
2279
    int plane, i, j;
2280
    for (plane=0; plane<4; plane++)
2281
    {
2282
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
2283
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2284
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
2285
        uint8_t *srcPtr= src[plane];
2286
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2287

    
2288
        if (!dst[plane]) continue;
2289
        // ignore palette for GRAY8
2290
        if (plane == 1 && !dst[2]) continue;
2291
        if (!src[plane] || (plane == 1 && !src[2])){
2292
            if(is16BPS(c->dstFormat))
2293
                length*=2;
2294
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
2295
        }else
2296
        {
2297
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)){
2298
                if (!isBE(c->srcFormat)) srcPtr++;
2299
                for (i=0; i<height; i++){
2300
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2301
                    srcPtr+= srcStride[plane];
2302
                    dstPtr+= dstStride[plane];
2303
                }
2304
            }else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)){
2305
                for (i=0; i<height; i++){
2306
                    for (j=0; j<length; j++){
2307
                        dstPtr[ j<<1   ] = srcPtr[j];
2308
                        dstPtr[(j<<1)+1] = srcPtr[j];
2309
                    }
2310
                    srcPtr+= srcStride[plane];
2311
                    dstPtr+= dstStride[plane];
2312
                }
2313
            }else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2314
                  && isBE(c->srcFormat) != isBE(c->dstFormat)){
2315

    
2316
                for (i=0; i<height; i++){
2317
                    for (j=0; j<length; j++)
2318
                        ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2319
                    srcPtr+= srcStride[plane];
2320
                    dstPtr+= dstStride[plane];
2321
                }
2322
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2323
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2324
            else
2325
            {
2326
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2327
                    length*=2;
2328
                for (i=0; i<height; i++)
2329
                {
2330
                    memcpy(dstPtr, srcPtr, length);
2331
                    srcPtr+= srcStride[plane];
2332
                    dstPtr+= dstStride[plane];
2333
                }
2334
            }
2335
        }
2336
    }
2337
    return srcSliceH;
2338
}
2339

    
2340

    
2341
static void getSubSampleFactors(int *h, int *v, int format){
2342
    switch(format){
2343
    case PIX_FMT_UYVY422:
2344
    case PIX_FMT_YUYV422:
2345
        *h=1;
2346
        *v=0;
2347
        break;
2348
    case PIX_FMT_YUV420P:
2349
    case PIX_FMT_YUV420PLE:
2350
    case PIX_FMT_YUV420PBE:
2351
    case PIX_FMT_YUVA420P:
2352
    case PIX_FMT_GRAY16BE:
2353
    case PIX_FMT_GRAY16LE:
2354
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2355
    case PIX_FMT_NV12:
2356
    case PIX_FMT_NV21:
2357
        *h=1;
2358
        *v=1;
2359
        break;
2360
    case PIX_FMT_YUV440P:
2361
        *h=0;
2362
        *v=1;
2363
        break;
2364
    case PIX_FMT_YUV410P:
2365
        *h=2;
2366
        *v=2;
2367
        break;
2368
    case PIX_FMT_YUV444P:
2369
    case PIX_FMT_YUV444PLE:
2370
    case PIX_FMT_YUV444PBE:
2371
        *h=0;
2372
        *v=0;
2373
        break;
2374
    case PIX_FMT_YUV422P:
2375
    case PIX_FMT_YUV422PLE:
2376
    case PIX_FMT_YUV422PBE:
2377
        *h=1;
2378
        *v=0;
2379
        break;
2380
    case PIX_FMT_YUV411P:
2381
        *h=2;
2382
        *v=0;
2383
        break;
2384
    default:
2385
        *h=0;
2386
        *v=0;
2387
        break;
2388
    }
2389
}
2390

    
2391
static uint16_t roundToInt16(int64_t f){
2392
    int r= (f + (1<<15))>>16;
2393
         if (r<-0x7FFF) return 0x8000;
2394
    else if (r> 0x7FFF) return 0x7FFF;
2395
    else                return r;
2396
}
2397

    
2398
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2399
    int64_t crv =  inv_table[0];
2400
    int64_t cbu =  inv_table[1];
2401
    int64_t cgu = -inv_table[2];
2402
    int64_t cgv = -inv_table[3];
2403
    int64_t cy  = 1<<16;
2404
    int64_t oy  = 0;
2405

    
2406
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2407
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2408

    
2409
    c->brightness= brightness;
2410
    c->contrast  = contrast;
2411
    c->saturation= saturation;
2412
    c->srcRange  = srcRange;
2413
    c->dstRange  = dstRange;
2414
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2415

    
2416
    c->uOffset=   0x0400040004000400LL;
2417
    c->vOffset=   0x0400040004000400LL;
2418

    
2419
    if (!srcRange){
2420
        cy= (cy*255) / 219;
2421
        oy= 16<<16;
2422
    }else{
2423
        crv= (crv*224) / 255;
2424
        cbu= (cbu*224) / 255;
2425
        cgu= (cgu*224) / 255;
2426
        cgv= (cgv*224) / 255;
2427
    }
2428

    
2429
    cy = (cy *contrast             )>>16;
2430
    crv= (crv*contrast * saturation)>>32;
2431
    cbu= (cbu*contrast * saturation)>>32;
2432
    cgu= (cgu*contrast * saturation)>>32;
2433
    cgv= (cgv*contrast * saturation)>>32;
2434

    
2435
    oy -= 256*brightness;
2436

    
2437
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2438
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2439
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2440
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2441
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2442
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2443

    
2444
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2445
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2446
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2447
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2448
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2449
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2450

    
2451
    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2452
    //FIXME factorize
2453

    
2454
#ifdef COMPILE_ALTIVEC
2455
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2456
        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
2457
#endif
2458
    return 0;
2459
}
2460

    
2461
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2462
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2463

    
2464
    *inv_table = c->srcColorspaceTable;
2465
    *table     = c->dstColorspaceTable;
2466
    *srcRange  = c->srcRange;
2467
    *dstRange  = c->dstRange;
2468
    *brightness= c->brightness;
2469
    *contrast  = c->contrast;
2470
    *saturation= c->saturation;
2471

    
2472
    return 0;
2473
}
2474

    
2475
static int handle_jpeg(enum PixelFormat *format)
2476
{
2477
    switch (*format) {
2478
        case PIX_FMT_YUVJ420P:
2479
            *format = PIX_FMT_YUV420P;
2480
            return 1;
2481
        case PIX_FMT_YUVJ422P:
2482
            *format = PIX_FMT_YUV422P;
2483
            return 1;
2484
        case PIX_FMT_YUVJ444P:
2485
            *format = PIX_FMT_YUV444P;
2486
            return 1;
2487
        case PIX_FMT_YUVJ440P:
2488
            *format = PIX_FMT_YUV440P;
2489
            return 1;
2490
        default:
2491
            return 0;
2492
    }
2493
}
2494

    
2495
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2496
                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2497
{
2498

    
2499
    SwsContext *c;
2500
    int i;
2501
    int usesVFilter, usesHFilter;
2502
    int unscaled, needsDither;
2503
    int srcRange, dstRange;
2504
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2505
#if ARCH_X86
2506
    if (flags & SWS_CPU_CAPS_MMX)
2507
        __asm__ volatile("emms\n\t"::: "memory");
2508
#endif
2509

    
2510
#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
2511
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2512
#if   COMPILE_TEMPLATE_MMX2
2513
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2514
#elif COMPILE_TEMPLATE_AMD3DNOW
2515
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2516
#elif COMPILE_TEMPLATE_MMX
2517
    flags |= SWS_CPU_CAPS_MMX;
2518
#elif COMPILE_TEMPLATE_ALTIVEC
2519
    flags |= SWS_CPU_CAPS_ALTIVEC;
2520
#elif ARCH_BFIN
2521
    flags |= SWS_CPU_CAPS_BFIN;
2522
#endif
2523
#endif /* CONFIG_RUNTIME_CPUDETECT */
2524
    if (clip_table[512] != 255) globalInit();
2525
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2526

    
2527
    unscaled = (srcW == dstW && srcH == dstH);
2528
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2529
        && (fmt_depth(dstFormat))<24
2530
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2531

    
2532
    srcRange = handle_jpeg(&srcFormat);
2533
    dstRange = handle_jpeg(&dstFormat);
2534

    
2535
    if (!isSupportedIn(srcFormat))
2536
    {
2537
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2538
        return NULL;
2539
    }
2540
    if (!isSupportedOut(dstFormat))
2541
    {
2542
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2543
        return NULL;
2544
    }
2545

    
2546
    i= flags & ( SWS_POINT
2547
                |SWS_AREA
2548
                |SWS_BILINEAR
2549
                |SWS_FAST_BILINEAR
2550
                |SWS_BICUBIC
2551
                |SWS_X
2552
                |SWS_GAUSS
2553
                |SWS_LANCZOS
2554
                |SWS_SINC
2555
                |SWS_SPLINE
2556
                |SWS_BICUBLIN);
2557
    if(!i || (i & (i-1)))
2558
    {
2559
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2560
        return NULL;
2561
    }
2562

    
2563
    /* sanity check */
2564
    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
2565
    {
2566
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2567
               srcW, srcH, dstW, dstH);
2568
        return NULL;
2569
    }
2570
    if(srcW > VOFW || dstW > VOFW){
2571
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2572
        return NULL;
2573
    }
2574

    
2575
    if (!dstFilter) dstFilter= &dummyFilter;
2576
    if (!srcFilter) srcFilter= &dummyFilter;
2577

    
2578
    c= av_mallocz(sizeof(SwsContext));
2579

    
2580
    c->av_class = &sws_context_class;
2581
    c->srcW= srcW;
2582
    c->srcH= srcH;
2583
    c->dstW= dstW;
2584
    c->dstH= dstH;
2585
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2586
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2587
    c->flags= flags;
2588
    c->dstFormat= dstFormat;
2589
    c->srcFormat= srcFormat;
2590
    c->vRounder= 4* 0x0001000100010001ULL;
2591

    
2592
    usesHFilter= usesVFilter= 0;
2593
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2594
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2595
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2596
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2597
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2598
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2599
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2600
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2601

    
2602
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2603
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2604

    
2605
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2606
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2607

    
2608
    // drop some chroma lines if the user wants it
2609
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2610
    c->chrSrcVSubSample+= c->vChrDrop;
2611

    
2612
    // drop every other pixel for chroma calculation unless user wants full chroma
2613
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2614
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2615
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2616
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2617
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2618
        c->chrSrcHSubSample=1;
2619

    
2620
    if (param){
2621
        c->param[0] = param[0];
2622
        c->param[1] = param[1];
2623
    }else{
2624
        c->param[0] =
2625
        c->param[1] = SWS_PARAM_DEFAULT;
2626
    }
2627

    
2628
    // Note the -((-x)>>y) is so that we always round toward +inf.
2629
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2630
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2631
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2632
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2633

    
2634
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2635

    
2636
    /* unscaled special cases */
2637
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2638
    {
2639
        /* yv12_to_nv12 */
2640
        if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2641
        {
2642
            c->swScale= PlanarToNV12Wrapper;
2643
        }
2644
        /* yuv2bgr */
2645
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2646
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1))
2647
        {
2648
            c->swScale= ff_yuv2rgb_get_func_ptr(c);
2649
        }
2650

    
2651
        if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))
2652
        {
2653
            c->swScale= yvu9toyv12Wrapper;
2654
        }
2655

    
2656
        /* bgr24toYV12 */
2657
        if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2658
            c->swScale= bgr24toyv12Wrapper;
2659

    
2660
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2661
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2662
           && (isBGR(dstFormat) || isRGB(dstFormat))
2663
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2664
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2665
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2666
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2667
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2668
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2669
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2670
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2671
                                             && dstFormat != PIX_FMT_RGB32_1
2672
                                             && dstFormat != PIX_FMT_BGR32_1
2673
           && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
2674
           && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
2675
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2676
             c->swScale= rgb2rgbWrapper;
2677

    
2678
        if ((usePal(srcFormat) && (
2679
                 dstFormat == PIX_FMT_RGB32   ||
2680
                 dstFormat == PIX_FMT_RGB32_1 ||
2681
                 dstFormat == PIX_FMT_RGB24   ||
2682
                 dstFormat == PIX_FMT_BGR32   ||
2683
                 dstFormat == PIX_FMT_BGR32_1 ||
2684
                 dstFormat == PIX_FMT_BGR24)))
2685
             c->swScale= pal2rgbWrapper;
2686

    
2687
        if (srcFormat == PIX_FMT_YUV422P)
2688
        {
2689
            if (dstFormat == PIX_FMT_YUYV422)
2690
                c->swScale= YUV422PToYuy2Wrapper;
2691
            else if (dstFormat == PIX_FMT_UYVY422)
2692
                c->swScale= YUV422PToUyvyWrapper;
2693
        }
2694

    
2695
        /* LQ converters if -sws 0 or -sws 4*/
2696
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2697
            /* yv12_to_yuy2 */
2698
            if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)
2699
            {
2700
                if (dstFormat == PIX_FMT_YUYV422)
2701
                    c->swScale= PlanarToYuy2Wrapper;
2702
                else if (dstFormat == PIX_FMT_UYVY422)
2703
                    c->swScale= PlanarToUyvyWrapper;
2704
            }
2705
        }
2706
        if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2707
            c->swScale= YUYV2YUV420Wrapper;
2708
        if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2709
            c->swScale= UYVY2YUV420Wrapper;
2710
        if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2711
            c->swScale= YUYV2YUV422Wrapper;
2712
        if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2713
            c->swScale= UYVY2YUV422Wrapper;
2714

    
2715
#ifdef COMPILE_ALTIVEC
2716
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2717
            !(c->flags & SWS_BITEXACT) &&
2718
            srcFormat == PIX_FMT_YUV420P) {
2719
          // unscaled YV12 -> packed YUV, we want speed
2720
          if (dstFormat == PIX_FMT_YUYV422)
2721
              c->swScale= yv12toyuy2_unscaled_altivec;
2722
          else if (dstFormat == PIX_FMT_UYVY422)
2723
              c->swScale= yv12touyvy_unscaled_altivec;
2724
        }
2725
#endif
2726

    
2727
        /* simple copy */
2728
        if (  srcFormat == dstFormat
2729
            || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2730
            || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2731
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2732
            || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2733
            || (isGray(dstFormat) && isGray(srcFormat))
2734
            || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2735
                && c->chrDstHSubSample == c->chrSrcHSubSample
2736
                && c->chrDstVSubSample == c->chrSrcVSubSample
2737
                && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
2738
                && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
2739
        {
2740
            if (isPacked(c->srcFormat))
2741
                c->swScale= packedCopy;
2742
            else /* Planar YUV or gray */
2743
                c->swScale= planarCopy;
2744
        }
2745
#if ARCH_BFIN
2746
        if (flags & SWS_CPU_CAPS_BFIN)
2747
            ff_bfin_get_unscaled_swscale (c);
2748
#endif
2749

    
2750
        if (c->swScale){
2751
            if (flags&SWS_PRINT_INFO)
2752
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2753
                                sws_format_name(srcFormat), sws_format_name(dstFormat));
2754
            return c;
2755
        }
2756
    }
2757

    
2758
    if (flags & SWS_CPU_CAPS_MMX2)
2759
    {
2760
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2761
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2762
        {
2763
            if (flags&SWS_PRINT_INFO)
2764
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2765
        }
2766
        if (usesHFilter) c->canMMX2BeUsed=0;
2767
    }
2768
    else
2769
        c->canMMX2BeUsed=0;
2770

    
2771
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2772
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2773

    
2774
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2775
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2776
    // n-2 is the last chrominance sample available
2777
    // this is not perfect, but no one should notice the difference, the more correct variant
2778
    // would be like the vertical one, but that would require some special code for the
2779
    // first and last pixel
2780
    if (flags&SWS_FAST_BILINEAR)
2781
    {
2782
        if (c->canMMX2BeUsed)
2783
        {
2784
            c->lumXInc+= 20;
2785
            c->chrXInc+= 20;
2786
        }
2787
        //we don't use the x86 asm scaler if MMX is available
2788
        else if (flags & SWS_CPU_CAPS_MMX)
2789
        {
2790
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2791
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2792
        }
2793
    }
2794

    
2795
    /* precalculate horizontal scaler filter coefficients */
2796
    {
2797
        const int filterAlign=
2798
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2799
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2800
            1;
2801

    
2802
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2803
                   srcW      ,       dstW, filterAlign, 1<<14,
2804
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2805
                   srcFilter->lumH, dstFilter->lumH, c->param);
2806
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2807
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2808
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2809
                   srcFilter->chrH, dstFilter->chrH, c->param);
2810

    
2811
#define MAX_MMX2_FILTER_CODE_SIZE 10000
2812
#if defined(COMPILE_MMX2)
2813
// can't downscale !!!
2814
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2815
        {
2816
#ifdef MAP_ANONYMOUS
2817
            c->lumMmx2FilterCode = mmap(NULL, MAX_MMX2_FILTER_CODE_SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2818
            c->chrMmx2FilterCode = mmap(NULL, MAX_MMX2_FILTER_CODE_SIZE, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2819
#elif HAVE_VIRTUALALLOC
2820
            c->lumMmx2FilterCode = VirtualAlloc(NULL, MAX_MMX2_FILTER_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2821
            c->chrMmx2FilterCode = VirtualAlloc(NULL, MAX_MMX2_FILTER_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2822
#else
2823
            c->lumMmx2FilterCode = av_malloc(MAX_MMX2_FILTER_CODE_SIZE);
2824
            c->chrMmx2FilterCode = av_malloc(MAX_MMX2_FILTER_CODE_SIZE);
2825
#endif
2826

    
2827
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2828
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2829
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2830
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2831

    
2832
            initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2833
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2834

    
2835
#ifdef MAP_ANONYMOUS
2836
            mprotect(c->lumMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE, PROT_EXEC | PROT_READ);
2837
            mprotect(c->chrMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE, PROT_EXEC | PROT_READ);
2838
#endif
2839
        }
2840
#endif /* defined(COMPILE_MMX2) */
2841
    } // initialize horizontal stuff
2842

    
2843

    
2844

    
2845
    /* precalculate vertical scaler filter coefficients */
2846
    {
2847
        const int filterAlign=
2848
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2849
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2850
            1;
2851

    
2852
        initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2853
                   srcH      ,        dstH, filterAlign, (1<<12),
2854
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2855
                   srcFilter->lumV, dstFilter->lumV, c->param);
2856
        initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2857
                   c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2858
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2859
                   srcFilter->chrV, dstFilter->chrV, c->param);
2860

    
2861
#ifdef COMPILE_ALTIVEC
2862
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2863
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2864

    
2865
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2866
            int j;
2867
            short *p = (short *)&c->vYCoeffsBank[i];
2868
            for (j=0;j<8;j++)
2869
                p[j] = c->vLumFilter[i];
2870
        }
2871

    
2872
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2873
            int j;
2874
            short *p = (short *)&c->vCCoeffsBank[i];
2875
            for (j=0;j<8;j++)
2876
                p[j] = c->vChrFilter[i];
2877
        }
2878
#endif
2879
    }
2880

    
2881
    // calculate buffer sizes so that they won't run out while handling these damn slices
2882
    c->vLumBufSize= c->vLumFilterSize;
2883
    c->vChrBufSize= c->vChrFilterSize;
2884
    for (i=0; i<dstH; i++)
2885
    {
2886
        int chrI= i*c->chrDstH / dstH;
2887
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2888
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2889

    
2890
        nextSlice>>= c->chrSrcVSubSample;
2891
        nextSlice<<= c->chrSrcVSubSample;
2892
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2893
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2894
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2895
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2896
    }
2897

    
2898
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2899
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2900
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2901
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2902
        c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2903
    //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)
2904
    /* align at 16 bytes for AltiVec */
2905
    for (i=0; i<c->vLumBufSize; i++)
2906
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2907
    for (i=0; i<c->vChrBufSize; i++)
2908
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2909
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2910
        for (i=0; i<c->vLumBufSize; i++)
2911
            c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2912

    
2913
    //try to avoid drawing green stuff between the right end and the stride end
2914
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2915

    
2916
    assert(2*VOFW == VOF);
2917

    
2918
    assert(c->chrDstH <= dstH);
2919

    
2920
    if (flags&SWS_PRINT_INFO)
2921
    {
2922
#ifdef DITHER1XBPP
2923
        const char *dither= " dithered";
2924
#else
2925
        const char *dither= "";
2926
#endif
2927
        if (flags&SWS_FAST_BILINEAR)
2928
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2929
        else if (flags&SWS_BILINEAR)
2930
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2931
        else if (flags&SWS_BICUBIC)
2932
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2933
        else if (flags&SWS_X)
2934
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2935
        else if (flags&SWS_POINT)
2936
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2937
        else if (flags&SWS_AREA)
2938
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2939
        else if (flags&SWS_BICUBLIN)
2940
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2941
        else if (flags&SWS_GAUSS)
2942
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2943
        else if (flags&SWS_SINC)
2944
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2945
        else if (flags&SWS_LANCZOS)
2946
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2947
        else if (flags&SWS_SPLINE)
2948
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2949
        else
2950
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2951

    
2952
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2953
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2954
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2955
        else
2956
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2957
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2958

    
2959
        if (flags & SWS_CPU_CAPS_MMX2)
2960
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2961
        else if (flags & SWS_CPU_CAPS_3DNOW)
2962
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2963
        else if (flags & SWS_CPU_CAPS_MMX)
2964
            av_log(c, AV_LOG_INFO, "using MMX\n");
2965
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2966
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2967
        else
2968
            av_log(c, AV_LOG_INFO, "using C\n");
2969
    }
2970

    
2971
    if (flags & SWS_PRINT_INFO)
2972
    {
2973
        if (flags & SWS_CPU_CAPS_MMX)
2974
        {
2975
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2976
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2977
            else
2978
            {
2979
                if (c->hLumFilterSize==4)
2980
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2981
                else if (c->hLumFilterSize==8)
2982
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2983
                else
2984
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2985

    
2986
                if (c->hChrFilterSize==4)
2987
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2988
                else if (c->hChrFilterSize==8)
2989
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2990
                else
2991
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2992
            }
2993
        }
2994
        else
2995
        {
2996
#if ARCH_X86
2997
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
2998
#else
2999
            if (flags & SWS_FAST_BILINEAR)
3000
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
3001
            else
3002
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
3003
#endif
3004
        }
3005
        if (isPlanarYUV(dstFormat))
3006
        {
3007
            if (c->vLumFilterSize==1)
3008
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3009
            else
3010
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3011
        }
3012
        else
3013
        {
3014
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
3015
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
3016
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3017
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
3018
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3019
            else
3020
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3021
        }
3022

    
3023
        if (dstFormat==PIX_FMT_BGR24)
3024
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
3025
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
3026
        else if (dstFormat==PIX_FMT_RGB32)
3027
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3028
        else if (dstFormat==PIX_FMT_BGR565)
3029
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3030
        else if (dstFormat==PIX_FMT_BGR555)
3031
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3032

    
3033
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
3034
    }
3035
    if (flags & SWS_PRINT_INFO)
3036
    {
3037
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3038
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
3039
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3040
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
3041
    }
3042

    
3043
    c->swScale= getSwsFunc(c);
3044
    return c;
3045
}
3046

    
3047
static void reset_ptr(uint8_t* src[], int format){
3048
    if(!isALPHA(format))
3049
        src[3]=NULL;
3050
    if(!isPlanarYUV(format)){
3051
        src[3]=src[2]=NULL;
3052
        if(   format != PIX_FMT_PAL8
3053
           && format != PIX_FMT_RGB8
3054
           && format != PIX_FMT_BGR8
3055
           && format != PIX_FMT_RGB4_BYTE
3056
           && format != PIX_FMT_BGR4_BYTE
3057
          )
3058
            src[1]= NULL;
3059
    }
3060
}
3061

    
3062
/**
3063
 * swscale wrapper, so we don't need to export the SwsContext.
3064
 * Assumes planar YUV to be in YUV order instead of YVU.
3065
 */
3066
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3067
              int srcSliceH, uint8_t* dst[], int dstStride[]){
3068
    int i;
3069
    uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
3070
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
3071

    
3072
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
3073
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
3074
        return 0;
3075
    }
3076
    if (c->sliceDir == 0) {
3077
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
3078
    }
3079

    
3080
    if (usePal(c->srcFormat)){
3081
        for (i=0; i<256; i++){
3082
            int p, r, g, b,y,u,v;
3083
            if(c->srcFormat == PIX_FMT_PAL8){
3084
                p=((uint32_t*)(src[1]))[i];
3085
                r= (p>>16)&0xFF;
3086
                g= (p>> 8)&0xFF;
3087
                b=  p     &0xFF;
3088
            }else if(c->srcFormat == PIX_FMT_RGB8){
3089
                r= (i>>5    )*36;
3090
                g= ((i>>2)&7)*36;
3091
                b= (i&3     )*85;
3092
            }else if(c->srcFormat == PIX_FMT_BGR8){
3093
                b= (i>>6    )*85;
3094
                g= ((i>>3)&7)*36;
3095
                r= (i&7     )*36;
3096
            }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
3097
                r= (i>>3    )*255;
3098
                g= ((i>>1)&3)*85;
3099
                b= (i&1     )*255;
3100
            }else {
3101
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
3102
                b= (i>>3    )*255;
3103
                g= ((i>>1)&3)*85;
3104
                r= (i&1     )*255;
3105
            }
3106
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3107
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3108
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3109
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
3110

    
3111

    
3112
            switch(c->dstFormat) {
3113
            case PIX_FMT_BGR32:
3114
#if !HAVE_BIGENDIAN
3115
            case PIX_FMT_RGB24:
3116
#endif
3117
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
3118
                break;
3119
            case PIX_FMT_BGR32_1:
3120
#if HAVE_BIGENDIAN
3121
            case PIX_FMT_BGR24:
3122
#endif
3123
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
3124
                break;
3125
            case PIX_FMT_RGB32_1:
3126
#if HAVE_BIGENDIAN
3127
            case PIX_FMT_RGB24:
3128
#endif
3129
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3130
                break;
3131
            case PIX_FMT_RGB32:
3132
#if !HAVE_BIGENDIAN
3133
            case PIX_FMT_BGR24:
3134
#endif
3135
            default:
3136
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
3137
            }
3138
        }
3139
    }
3140

    
3141
    // copy strides, so they can safely be modified
3142
    if (c->sliceDir == 1) {
3143
        // slices go from top to bottom
3144
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3145
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3146

    
3147
        reset_ptr(src2, c->srcFormat);
3148
        reset_ptr(dst2, c->dstFormat);
3149

    
3150
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
3151
    } else {
3152
        // slices go from bottom to top => we flip the image internally
3153
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3154
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3155

    
3156
        src2[0] += (srcSliceH-1)*srcStride[0];
3157
        if (!usePal(c->srcFormat))
3158
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3159
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3160
        src2[3] += (srcSliceH-1)*srcStride[3];
3161
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
3162
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3163
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3164
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
3165

    
3166
        reset_ptr(src2, c->srcFormat);
3167
        reset_ptr(dst2, c->dstFormat);
3168

    
3169
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3170
    }
3171
}
3172

    
3173
#if LIBSWSCALE_VERSION_MAJOR < 1
3174
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3175
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
3176
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3177
}
3178
#endif
3179

    
3180
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3181
                                float lumaSharpen, float chromaSharpen,
3182
                                float chromaHShift, float chromaVShift,
3183
                                int verbose)
3184
{
3185
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3186

    
3187
    if (lumaGBlur!=0.0){
3188
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3189
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3190
    }else{
3191
        filter->lumH= sws_getIdentityVec();
3192
        filter->lumV= sws_getIdentityVec();
3193
    }
3194

    
3195
    if (chromaGBlur!=0.0){
3196
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3197
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3198
    }else{
3199
        filter->chrH= sws_getIdentityVec();
3200
        filter->chrV= sws_getIdentityVec();
3201
    }
3202

    
3203
    if (chromaSharpen!=0.0){
3204
        SwsVector *id= sws_getIdentityVec();
3205
        sws_scaleVec(filter->chrH, -chromaSharpen);
3206
        sws_scaleVec(filter->chrV, -chromaSharpen);
3207
        sws_addVec(filter->chrH, id);
3208
        sws_addVec(filter->chrV, id);
3209
        sws_freeVec(id);
3210
    }
3211

    
3212
    if (lumaSharpen!=0.0){
3213
        SwsVector *id= sws_getIdentityVec();
3214
        sws_scaleVec(filter->lumH, -lumaSharpen);
3215
        sws_scaleVec(filter->lumV, -lumaSharpen);
3216
        sws_addVec(filter->lumH, id);
3217
        sws_addVec(filter->lumV, id);
3218
        sws_freeVec(id);
3219
    }
3220

    
3221
    if (chromaHShift != 0.0)
3222
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3223

    
3224
    if (chromaVShift != 0.0)
3225
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3226

    
3227
    sws_normalizeVec(filter->chrH, 1.0);
3228
    sws_normalizeVec(filter->chrV, 1.0);
3229
    sws_normalizeVec(filter->lumH, 1.0);
3230
    sws_normalizeVec(filter->lumV, 1.0);
3231

    
3232
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3233
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3234

    
3235
    return filter;
3236
}
3237

    
3238
SwsVector *sws_getGaussianVec(double variance, double quality){
3239
    const int length= (int)(variance*quality + 0.5) | 1;
3240
    int i;
3241
    double *coeff= av_malloc(length*sizeof(double));
3242
    double middle= (length-1)*0.5;
3243
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3244

    
3245
    vec->coeff= coeff;
3246
    vec->length= length;
3247

    
3248
    for (i=0; i<length; i++)
3249
    {
3250
        double dist= i-middle;
3251
        coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3252
    }
3253

    
3254
    sws_normalizeVec(vec, 1.0);
3255

    
3256
    return vec;
3257
}
3258

    
3259
SwsVector *sws_getConstVec(double c, int length){
3260
    int i;
3261
    double *coeff= av_malloc(length*sizeof(double));
3262
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3263

    
3264
    vec->coeff= coeff;
3265
    vec->length= length;
3266

    
3267
    for (i=0; i<length; i++)
3268
        coeff[i]= c;
3269

    
3270
    return vec;
3271
}
3272

    
3273

    
3274
SwsVector *sws_getIdentityVec(void){
3275
    return sws_getConstVec(1.0, 1);
3276
}
3277

    
3278
double sws_dcVec(SwsVector *a){
3279
    int i;
3280
    double sum=0;
3281

    
3282
    for (i=0; i<a->length; i++)
3283
        sum+= a->coeff[i];
3284

    
3285
    return sum;
3286
}
3287

    
3288
void sws_scaleVec(SwsVector *a, double scalar){
3289
    int i;
3290

    
3291
    for (i=0; i<a->length; i++)
3292
        a->coeff[i]*= scalar;
3293
}
3294

    
3295
void sws_normalizeVec(SwsVector *a, double height){
3296
    sws_scaleVec(a, height/sws_dcVec(a));
3297
}
3298

    
3299
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
3300
    int length= a->length + b->length - 1;
3301
    double *coeff= av_malloc(length*sizeof(double));
3302
    int i, j;
3303
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3304

    
3305
    vec->coeff= coeff;
3306
    vec->length= length;
3307

    
3308
    for (i=0; i<length; i++) coeff[i]= 0.0;
3309

    
3310
    for (i=0; i<a->length; i++)
3311
    {
3312
        for (j=0; j<b->length; j++)
3313
        {
3314
            coeff[i+j]+= a->coeff[i]*b->coeff[j];
3315
        }
3316
    }
3317

    
3318
    return vec;
3319
}
3320

    
3321
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
3322
    int length= FFMAX(a->length, b->length);
3323
    double *coeff= av_malloc(length*sizeof(double));
3324
    int i;
3325
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3326

    
3327
    vec->coeff= coeff;
3328
    vec->length= length;
3329

    
3330
    for (i=0; i<length; i++) coeff[i]= 0.0;
3331

    
3332
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3333
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3334

    
3335
    return vec;
3336
}
3337

    
3338
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
3339
    int length= FFMAX(a->length, b->length);
3340
    double *coeff= av_malloc(length*sizeof(double));
3341
    int i;
3342
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3343

    
3344
    vec->coeff= coeff;
3345
    vec->length= length;
3346

    
3347
    for (i=0; i<length; i++) coeff[i]= 0.0;
3348

    
3349
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3350
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3351

    
3352
    return vec;
3353
}
3354

    
3355
/* shift left / or right if "shift" is negative */
3356
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
3357
    int length= a->length + FFABS(shift)*2;
3358
    double *coeff= av_malloc(length*sizeof(double));
3359
    int i;
3360
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3361

    
3362
    vec->coeff= coeff;
3363
    vec->length= length;
3364

    
3365
    for (i=0; i<length; i++) coeff[i]= 0.0;
3366

    
3367
    for (i=0; i<a->length; i++)
3368
    {
3369
        coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3370
    }
3371

    
3372
    return vec;
3373
}
3374

    
3375
void sws_shiftVec(SwsVector *a, int shift){
3376
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3377
    av_free(a->coeff);
3378
    a->coeff= shifted->coeff;
3379
    a->length= shifted->length;
3380
    av_free(shifted);
3381
}
3382

    
3383
void sws_addVec(SwsVector *a, SwsVector *b){
3384
    SwsVector *sum= sws_sumVec(a, b);
3385
    av_free(a->coeff);
3386
    a->coeff= sum->coeff;
3387
    a->length= sum->length;
3388
    av_free(sum);
3389
}
3390

    
3391
void sws_subVec(SwsVector *a, SwsVector *b){
3392
    SwsVector *diff= sws_diffVec(a, b);
3393
    av_free(a->coeff);
3394
    a->coeff= diff->coeff;
3395
    a->length= diff->length;
3396
    av_free(diff);
3397
}
3398

    
3399
void sws_convVec(SwsVector *a, SwsVector *b){
3400
    SwsVector *conv= sws_getConvVec(a, b);
3401
    av_free(a->coeff);
3402
    a->coeff= conv->coeff;
3403
    a->length= conv->length;
3404
    av_free(conv);
3405
}
3406

    
3407
SwsVector *sws_cloneVec(SwsVector *a){
3408
    double *coeff= av_malloc(a->length*sizeof(double));
3409
    int i;
3410
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3411

    
3412
    vec->coeff= coeff;
3413
    vec->length= a->length;
3414

    
3415
    for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
3416

    
3417
    return vec;
3418
}
3419

    
3420
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){
3421
    int i;
3422
    double max=0;
3423
    double min=0;
3424
    double range;
3425

    
3426
    for (i=0; i<a->length; i++)
3427
        if (a->coeff[i]>max) max= a->coeff[i];
3428

    
3429
    for (i=0; i<a->length; i++)
3430
        if (a->coeff[i]<min) min= a->coeff[i];
3431

    
3432
    range= max - min;
3433

    
3434
    for (i=0; i<a->length; i++)
3435
    {
3436
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3437
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3438
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3439
        av_log(log_ctx, log_level, "|\n");
3440
    }
3441
}
3442

    
3443
#if LIBSWSCALE_VERSION_MAJOR < 1
3444
void sws_printVec(SwsVector *a){
3445
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3446
}
3447
#endif
3448

    
3449
void sws_freeVec(SwsVector *a){
3450
    if (!a) return;
3451
    av_freep(&a->coeff);
3452
    a->length=0;
3453
    av_free(a);
3454
}
3455

    
3456
void sws_freeFilter(SwsFilter *filter){
3457
    if (!filter) return;
3458

    
3459
    if (filter->lumH) sws_freeVec(filter->lumH);
3460
    if (filter->lumV) sws_freeVec(filter->lumV);
3461
    if (filter->chrH) sws_freeVec(filter->chrH);
3462
    if (filter->chrV) sws_freeVec(filter->chrV);
3463
    av_free(filter);
3464
}
3465

    
3466

    
3467
void sws_freeContext(SwsContext *c){
3468
    int i;
3469
    if (!c) return;
3470

    
3471
    if (c->lumPixBuf)
3472
    {
3473
        for (i=0; i<c->vLumBufSize; i++)
3474
            av_freep(&c->lumPixBuf[i]);
3475
        av_freep(&c->lumPixBuf);
3476
    }
3477

    
3478
    if (c->chrPixBuf)
3479
    {
3480
        for (i=0; i<c->vChrBufSize; i++)
3481
            av_freep(&c->chrPixBuf[i]);
3482
        av_freep(&c->chrPixBuf);
3483
    }
3484

    
3485
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
3486
        for (i=0; i<c->vLumBufSize; i++)
3487
            av_freep(&c->alpPixBuf[i]);
3488
        av_freep(&c->alpPixBuf);
3489
    }
3490

    
3491
    av_freep(&c->vLumFilter);
3492
    av_freep(&c->vChrFilter);
3493
    av_freep(&c->hLumFilter);
3494
    av_freep(&c->hChrFilter);
3495
#ifdef COMPILE_ALTIVEC
3496
    av_freep(&c->vYCoeffsBank);
3497
    av_freep(&c->vCCoeffsBank);
3498
#endif
3499

    
3500
    av_freep(&c->vLumFilterPos);
3501
    av_freep(&c->vChrFilterPos);
3502
    av_freep(&c->hLumFilterPos);
3503
    av_freep(&c->hChrFilterPos);
3504

    
3505
#if ARCH_X86 && CONFIG_GPL
3506
#ifdef MAP_ANONYMOUS
3507
    if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE);
3508
    if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE);
3509
#elif HAVE_VIRTUALALLOC
3510
    if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE, MEM_RELEASE);
3511
    if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, MAX_MMX2_FILTER_CODE_SIZE, MEM_RELEASE);
3512
#else
3513
    av_free(c->lumMmx2FilterCode);
3514
    av_free(c->chrMmx2FilterCode);
3515
#endif
3516
    c->lumMmx2FilterCode=NULL;
3517
    c->chrMmx2FilterCode=NULL;
3518
#endif /* ARCH_X86 && CONFIG_GPL */
3519

    
3520
    av_freep(&c->lumMmx2Filter);
3521
    av_freep(&c->chrMmx2Filter);
3522
    av_freep(&c->lumMmx2FilterPos);
3523
    av_freep(&c->chrMmx2FilterPos);
3524
    av_freep(&c->yuvTable);
3525

    
3526
    av_free(c);
3527
}
3528

    
3529
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3530
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3531
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3532
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
3533
{
3534
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3535

    
3536
    if (!param)
3537
        param = default_param;
3538

    
3539
    if (context) {
3540
        if (context->srcW != srcW || context->srcH != srcH ||
3541
            context->srcFormat != srcFormat ||
3542
            context->dstW != dstW || context->dstH != dstH ||
3543
            context->dstFormat != dstFormat || context->flags != flags ||
3544
            context->param[0] != param[0] || context->param[1] != param[1])
3545
        {
3546
            sws_freeContext(context);
3547
            context = NULL;
3548
        }
3549
    }
3550
    if (!context) {
3551
        return sws_getContext(srcW, srcH, srcFormat,
3552
                              dstW, dstH, dstFormat, flags,
3553
                              srcFilter, dstFilter, param);
3554
    }
3555
    return context;
3556
}
3557