Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 469fd9b0

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
#if 0
306
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_64[8][8])={
307
{  0,  48,  12,  60,   3,  51,  15,  63, },
308
{ 32,  16,  44,  28,  35,  19,  47,  31, },
309
{  8,  56,   4,  52,  11,  59,   7,  55, },
310
{ 40,  24,  36,  20,  43,  27,  39,  23, },
311
{  2,  50,  14,  62,   1,  49,  13,  61, },
312
{ 34,  18,  46,  30,  33,  17,  45,  29, },
313
{ 10,  58,   6,  54,   9,  57,   5,  53, },
314
{ 42,  26,  38,  22,  41,  25,  37,  21, },
315
};
316
#endif
317

    
318
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73[8][8])={
319
{  0,  55,  14,  68,   3,  58,  17,  72, },
320
{ 37,  18,  50,  32,  40,  22,  54,  35, },
321
{  9,  64,   5,  59,  13,  67,   8,  63, },
322
{ 46,  27,  41,  23,  49,  31,  44,  26, },
323
{  2,  57,  16,  71,   1,  56,  15,  70, },
324
{ 39,  21,  52,  34,  38,  19,  51,  33, },
325
{ 11,  66,   7,  62,  10,  65,   6,  60, },
326
{ 48,  30,  43,  25,  47,  29,  42,  24, },
327
};
328

    
329
#if 0
330
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_128[8][8])={
331
{ 68,  36,  92,  60,  66,  34,  90,  58, },
332
{ 20, 116,  12, 108,  18, 114,  10, 106, },
333
{ 84,  52,  76,  44,  82,  50,  74,  42, },
334
{  0,  96,  24, 120,   6, 102,  30, 126, },
335
{ 64,  32,  88,  56,  70,  38,  94,  62, },
336
{ 16, 112,   8, 104,  22, 118,  14, 110, },
337
{ 80,  48,  72,  40,  86,  54,  78,  46, },
338
{  4, 100,  28, 124,   2,  98,  26, 122, },
339
};
340
#endif
341

    
342
#if 1
343
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
344
{117,  62, 158, 103, 113,  58, 155, 100, },
345
{ 34, 199,  21, 186,  31, 196,  17, 182, },
346
{144,  89, 131,  76, 141,  86, 127,  72, },
347
{  0, 165,  41, 206,  10, 175,  52, 217, },
348
{110,  55, 151,  96, 120,  65, 162, 107, },
349
{ 28, 193,  14, 179,  38, 203,  24, 189, },
350
{138,  83, 124,  69, 148,  93, 134,  79, },
351
{  7, 172,  48, 213,   3, 168,  45, 210, },
352
};
353
#elif 1
354
// tries to correct a gamma of 1.5
355
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
356
{  0, 143,  18, 200,   2, 156,  25, 215, },
357
{ 78,  28, 125,  64,  89,  36, 138,  74, },
358
{ 10, 180,   3, 161,  16, 195,   8, 175, },
359
{109,  51,  93,  38, 121,  60, 105,  47, },
360
{  1, 152,  23, 210,   0, 147,  20, 205, },
361
{ 85,  33, 134,  71,  81,  30, 130,  67, },
362
{ 14, 190,   6, 171,  12, 185,   5, 166, },
363
{117,  57, 101,  44, 113,  54,  97,  41, },
364
};
365
#elif 1
366
// tries to correct a gamma of 2.0
367
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
368
{  0, 124,   8, 193,   0, 140,  12, 213, },
369
{ 55,  14, 104,  42,  66,  19, 119,  52, },
370
{  3, 168,   1, 145,   6, 187,   3, 162, },
371
{ 86,  31,  70,  21,  99,  39,  82,  28, },
372
{  0, 134,  11, 206,   0, 129,   9, 200, },
373
{ 62,  17, 114,  48,  58,  16, 109,  45, },
374
{  5, 181,   2, 157,   4, 175,   1, 151, },
375
{ 95,  36,  78,  26,  90,  34,  74,  24, },
376
};
377
#else
378
// tries to correct a gamma of 2.5
379
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
380
{  0, 107,   3, 187,   0, 125,   6, 212, },
381
{ 39,   7,  86,  28,  49,  11, 102,  36, },
382
{  1, 158,   0, 131,   3, 180,   1, 151, },
383
{ 68,  19,  52,  12,  81,  25,  64,  17, },
384
{  0, 119,   5, 203,   0, 113,   4, 195, },
385
{ 45,   9,  96,  33,  42,   8,  91,  30, },
386
{  2, 172,   1, 144,   2, 165,   0, 137, },
387
{ 77,  23,  60,  15,  72,  21,  56,  14, },
388
};
389
#endif
390

    
391
const char *sws_format_name(enum PixelFormat format)
392
{
393
    switch (format) {
394
        case PIX_FMT_YUV420P:
395
            return "yuv420p";
396
        case PIX_FMT_YUVA420P:
397
            return "yuva420p";
398
        case PIX_FMT_YUYV422:
399
            return "yuyv422";
400
        case PIX_FMT_RGB24:
401
            return "rgb24";
402
        case PIX_FMT_BGR24:
403
            return "bgr24";
404
        case PIX_FMT_YUV422P:
405
            return "yuv422p";
406
        case PIX_FMT_YUV444P:
407
            return "yuv444p";
408
        case PIX_FMT_RGB32:
409
            return "rgb32";
410
        case PIX_FMT_YUV410P:
411
            return "yuv410p";
412
        case PIX_FMT_YUV411P:
413
            return "yuv411p";
414
        case PIX_FMT_RGB565:
415
            return "rgb565";
416
        case PIX_FMT_RGB555:
417
            return "rgb555";
418
        case PIX_FMT_GRAY16BE:
419
            return "gray16be";
420
        case PIX_FMT_GRAY16LE:
421
            return "gray16le";
422
        case PIX_FMT_GRAY8:
423
            return "gray8";
424
        case PIX_FMT_MONOWHITE:
425
            return "mono white";
426
        case PIX_FMT_MONOBLACK:
427
            return "mono black";
428
        case PIX_FMT_PAL8:
429
            return "Palette";
430
        case PIX_FMT_YUVJ420P:
431
            return "yuvj420p";
432
        case PIX_FMT_YUVJ422P:
433
            return "yuvj422p";
434
        case PIX_FMT_YUVJ444P:
435
            return "yuvj444p";
436
        case PIX_FMT_XVMC_MPEG2_MC:
437
            return "xvmc_mpeg2_mc";
438
        case PIX_FMT_XVMC_MPEG2_IDCT:
439
            return "xvmc_mpeg2_idct";
440
        case PIX_FMT_UYVY422:
441
            return "uyvy422";
442
        case PIX_FMT_UYYVYY411:
443
            return "uyyvyy411";
444
        case PIX_FMT_RGB32_1:
445
            return "rgb32x";
446
        case PIX_FMT_BGR32_1:
447
            return "bgr32x";
448
        case PIX_FMT_BGR32:
449
            return "bgr32";
450
        case PIX_FMT_BGR565:
451
            return "bgr565";
452
        case PIX_FMT_BGR555:
453
            return "bgr555";
454
        case PIX_FMT_BGR8:
455
            return "bgr8";
456
        case PIX_FMT_BGR4:
457
            return "bgr4";
458
        case PIX_FMT_BGR4_BYTE:
459
            return "bgr4 byte";
460
        case PIX_FMT_RGB8:
461
            return "rgb8";
462
        case PIX_FMT_RGB4:
463
            return "rgb4";
464
        case PIX_FMT_RGB4_BYTE:
465
            return "rgb4 byte";
466
        case PIX_FMT_RGB48BE:
467
            return "rgb48be";
468
        case PIX_FMT_RGB48LE:
469
            return "rgb48le";
470
        case PIX_FMT_NV12:
471
            return "nv12";
472
        case PIX_FMT_NV21:
473
            return "nv21";
474
        case PIX_FMT_YUV440P:
475
            return "yuv440p";
476
        case PIX_FMT_VDPAU_H264:
477
            return "vdpau_h264";
478
        case PIX_FMT_VDPAU_MPEG1:
479
            return "vdpau_mpeg1";
480
        case PIX_FMT_VDPAU_MPEG2:
481
            return "vdpau_mpeg2";
482
        case PIX_FMT_VDPAU_WMV3:
483
            return "vdpau_wmv3";
484
        case PIX_FMT_VDPAU_VC1:
485
            return "vdpau_vc1";
486
        case PIX_FMT_YUV420PLE:
487
            return "yuv420ple";
488
        case PIX_FMT_YUV422PLE:
489
            return "yuv422ple";
490
        case PIX_FMT_YUV444PLE:
491
            return "yuv444ple";
492
        case PIX_FMT_YUV420PBE:
493
            return "yuv420pbe";
494
        case PIX_FMT_YUV422PBE:
495
            return "yuv422pbe";
496
        case PIX_FMT_YUV444PBE:
497
            return "yuv444pbe";
498
        default:
499
            return "Unknown format";
500
    }
501
}
502

    
503
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
504
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
505
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
506
{
507
    //FIXME Optimize (just quickly written not optimized..)
508
    int i;
509
    for (i=0; i<dstW; i++)
510
    {
511
        int val=1<<18;
512
        int j;
513
        for (j=0; j<lumFilterSize; j++)
514
            val += lumSrc[j][i] * lumFilter[j];
515

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

    
519
    if (uDest)
520
        for (i=0; i<chrDstW; i++)
521
        {
522
            int u=1<<18;
523
            int v=1<<18;
524
            int j;
525
            for (j=0; j<chrFilterSize; j++)
526
            {
527
                u += chrSrc[j][i] * chrFilter[j];
528
                v += chrSrc[j][i + VOFW] * chrFilter[j];
529
            }
530

    
531
            uDest[i]= av_clip_uint8(u>>19);
532
            vDest[i]= av_clip_uint8(v>>19);
533
        }
534

    
535
    if (CONFIG_SWSCALE_ALPHA && aDest)
536
        for (i=0; i<dstW; i++){
537
            int val=1<<18;
538
            int j;
539
            for (j=0; j<lumFilterSize; j++)
540
                val += alpSrc[j][i] * lumFilter[j];
541

    
542
            aDest[i]= av_clip_uint8(val>>19);
543
        }
544

    
545
}
546

    
547
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
548
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
549
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
550
{
551
    //FIXME Optimize (just quickly written not optimized..)
552
    int i;
553
    for (i=0; i<dstW; i++)
554
    {
555
        int val=1<<18;
556
        int j;
557
        for (j=0; j<lumFilterSize; j++)
558
            val += lumSrc[j][i] * lumFilter[j];
559

    
560
        dest[i]= av_clip_uint8(val>>19);
561
    }
562

    
563
    if (!uDest)
564
        return;
565

    
566
    if (dstFormat == PIX_FMT_NV12)
567
        for (i=0; i<chrDstW; i++)
568
        {
569
            int u=1<<18;
570
            int v=1<<18;
571
            int j;
572
            for (j=0; j<chrFilterSize; j++)
573
            {
574
                u += chrSrc[j][i] * chrFilter[j];
575
                v += chrSrc[j][i + VOFW] * chrFilter[j];
576
            }
577

    
578
            uDest[2*i]= av_clip_uint8(u>>19);
579
            uDest[2*i+1]= av_clip_uint8(v>>19);
580
        }
581
    else
582
        for (i=0; i<chrDstW; i++)
583
        {
584
            int u=1<<18;
585
            int v=1<<18;
586
            int j;
587
            for (j=0; j<chrFilterSize; j++)
588
            {
589
                u += chrSrc[j][i] * chrFilter[j];
590
                v += chrSrc[j][i + VOFW] * chrFilter[j];
591
            }
592

    
593
            uDest[2*i]= av_clip_uint8(v>>19);
594
            uDest[2*i+1]= av_clip_uint8(u>>19);
595
        }
596
}
597

    
598
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
599
    for (i=0; i<(dstW>>1); i++){\
600
        int j;\
601
        int Y1 = 1<<18;\
602
        int Y2 = 1<<18;\
603
        int U  = 1<<18;\
604
        int V  = 1<<18;\
605
        int av_unused A1, A2;\
606
        type av_unused *r, *b, *g;\
607
        const int i2= 2*i;\
608
        \
609
        for (j=0; j<lumFilterSize; j++)\
610
        {\
611
            Y1 += lumSrc[j][i2] * lumFilter[j];\
612
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
613
        }\
614
        for (j=0; j<chrFilterSize; j++)\
615
        {\
616
            U += chrSrc[j][i] * chrFilter[j];\
617
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
618
        }\
619
        Y1>>=19;\
620
        Y2>>=19;\
621
        U >>=19;\
622
        V >>=19;\
623
        if (alpha){\
624
            A1 = 1<<18;\
625
            A2 = 1<<18;\
626
            for (j=0; j<lumFilterSize; j++){\
627
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
628
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
629
            }\
630
            A1>>=19;\
631
            A2>>=19;\
632
        }\
633

    
634
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
635
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
636
        if ((Y1|Y2|U|V)&256)\
637
        {\
638
            if (Y1>255)   Y1=255; \
639
            else if (Y1<0)Y1=0;   \
640
            if (Y2>255)   Y2=255; \
641
            else if (Y2<0)Y2=0;   \
642
            if (U>255)    U=255;  \
643
            else if (U<0) U=0;    \
644
            if (V>255)    V=255;  \
645
            else if (V<0) V=0;    \
646
        }\
647
        if (alpha && ((A1|A2)&256)){\
648
            A1=av_clip_uint8(A1);\
649
            A2=av_clip_uint8(A2);\
650
        }
651

    
652
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
653
    for (i=0; i<dstW; i++){\
654
        int j;\
655
        int Y = 0;\
656
        int U = -128<<19;\
657
        int V = -128<<19;\
658
        int av_unused A;\
659
        int R,G,B;\
660
        \
661
        for (j=0; j<lumFilterSize; j++){\
662
            Y += lumSrc[j][i     ] * lumFilter[j];\
663
        }\
664
        for (j=0; j<chrFilterSize; j++){\
665
            U += chrSrc[j][i     ] * chrFilter[j];\
666
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
667
        }\
668
        Y >>=10;\
669
        U >>=10;\
670
        V >>=10;\
671
        if (alpha){\
672
            A = rnd;\
673
            for (j=0; j<lumFilterSize; j++)\
674
                A += alpSrc[j][i     ] * lumFilter[j];\
675
            A >>=19;\
676
            if (A&256)\
677
                A = av_clip_uint8(A);\
678
        }\
679

    
680
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
681
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
682
        Y-= c->yuv2rgb_y_offset;\
683
        Y*= c->yuv2rgb_y_coeff;\
684
        Y+= rnd;\
685
        R= Y + V*c->yuv2rgb_v2r_coeff;\
686
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
687
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
688
        if ((R|G|B)&(0xC0000000)){\
689
            if (R>=(256<<22))   R=(256<<22)-1; \
690
            else if (R<0)R=0;   \
691
            if (G>=(256<<22))   G=(256<<22)-1; \
692
            else if (G<0)G=0;   \
693
            if (B>=(256<<22))   B=(256<<22)-1; \
694
            else if (B<0)B=0;   \
695
        }\
696

    
697

    
698
#define YSCALE_YUV_2_GRAY16_C \
699
    for (i=0; i<(dstW>>1); i++){\
700
        int j;\
701
        int Y1 = 1<<18;\
702
        int Y2 = 1<<18;\
703
        int U  = 1<<18;\
704
        int V  = 1<<18;\
705
        \
706
        const int i2= 2*i;\
707
        \
708
        for (j=0; j<lumFilterSize; j++)\
709
        {\
710
            Y1 += lumSrc[j][i2] * lumFilter[j];\
711
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
712
        }\
713
        Y1>>=11;\
714
        Y2>>=11;\
715
        if ((Y1|Y2|U|V)&65536)\
716
        {\
717
            if (Y1>65535)   Y1=65535; \
718
            else if (Y1<0)Y1=0;   \
719
            if (Y2>65535)   Y2=65535; \
720
            else if (Y2<0)Y2=0;   \
721
        }
722

    
723
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
724
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
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_PACKED2_C(type,alpha)   \
730
    for (i=0; i<(dstW>>1); i++){ \
731
        const int i2= 2*i;       \
732
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
733
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
734
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
735
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
736
        type av_unused *r, *b, *g;                                    \
737
        int av_unused A1, A2;                                         \
738
        if (alpha){\
739
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
740
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
741
        }\
742

    
743
#define YSCALE_YUV_2_GRAY16_2_C   \
744
    for (i=0; i<(dstW>>1); i++){ \
745
        const int i2= 2*i;       \
746
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
747
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
748

    
749
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
750
    YSCALE_YUV_2_PACKED2_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_PACKED1_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= (uvbuf1[i     ])>>7;\
761
        int V= (uvbuf1[i+VOFW])>>7;\
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_GRAY16_1_C \
770
    for (i=0; i<(dstW>>1); i++){\
771
        const int i2= 2*i;\
772
        int Y1= buf0[i2  ]<<1;\
773
        int Y2= buf0[i2+1]<<1;\
774

    
775
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
776
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
777
    r = (type *)c->table_rV[V];\
778
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
779
    b = (type *)c->table_bU[U];\
780

    
781
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
782
    for (i=0; i<(dstW>>1); i++){\
783
        const int i2= 2*i;\
784
        int Y1= buf0[i2  ]>>7;\
785
        int Y2= buf0[i2+1]>>7;\
786
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
787
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
788
        type av_unused *r, *b, *g;\
789
        int av_unused A1, A2;\
790
        if (alpha){\
791
            A1= abuf0[i2  ]>>7;\
792
            A2= abuf0[i2+1]>>7;\
793
        }\
794

    
795
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
796
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
797
    r = (type *)c->table_rV[V];\
798
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
799
    b = (type *)c->table_bU[U];\
800

    
801
#define YSCALE_YUV_2_MONO2_C \
802
    const uint8_t * const d128=dither_8x8_220[y&7];\
803
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
804
    for (i=0; i<dstW-7; i+=8){\
805
        int acc;\
806
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
807
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
808
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
809
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
810
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
811
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
812
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
813
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
814
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
815
        dest++;\
816
    }\
817

    
818

    
819
#define YSCALE_YUV_2_MONOX_C \
820
    const uint8_t * const d128=dither_8x8_220[y&7];\
821
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
822
    int acc=0;\
823
    for (i=0; i<dstW-1; i+=2){\
824
        int j;\
825
        int Y1=1<<18;\
826
        int Y2=1<<18;\
827
\
828
        for (j=0; j<lumFilterSize; j++)\
829
        {\
830
            Y1 += lumSrc[j][i] * lumFilter[j];\
831
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
832
        }\
833
        Y1>>=19;\
834
        Y2>>=19;\
835
        if ((Y1|Y2)&256)\
836
        {\
837
            if (Y1>255)   Y1=255;\
838
            else if (Y1<0)Y1=0;\
839
            if (Y2>255)   Y2=255;\
840
            else if (Y2<0)Y2=0;\
841
        }\
842
        acc+= acc + g[Y1+d128[(i+0)&7]];\
843
        acc+= acc + g[Y2+d128[(i+1)&7]];\
844
        if ((i&7)==6){\
845
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
846
            dest++;\
847
        }\
848
    }
849

    
850

    
851
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
852
    switch(c->dstFormat)\
853
    {\
854
    case PIX_FMT_RGB48BE:\
855
    case PIX_FMT_RGB48LE:\
856
        func(uint8_t,0)\
857
            ((uint8_t*)dest)[ 0]= r[Y1];\
858
            ((uint8_t*)dest)[ 1]= r[Y1];\
859
            ((uint8_t*)dest)[ 2]= g[Y1];\
860
            ((uint8_t*)dest)[ 3]= g[Y1];\
861
            ((uint8_t*)dest)[ 4]= b[Y1];\
862
            ((uint8_t*)dest)[ 5]= b[Y1];\
863
            ((uint8_t*)dest)[ 6]= r[Y2];\
864
            ((uint8_t*)dest)[ 7]= r[Y2];\
865
            ((uint8_t*)dest)[ 8]= g[Y2];\
866
            ((uint8_t*)dest)[ 9]= g[Y2];\
867
            ((uint8_t*)dest)[10]= b[Y2];\
868
            ((uint8_t*)dest)[11]= b[Y2];\
869
            dest+=12;\
870
        }\
871
        break;\
872
    case PIX_FMT_RGBA:\
873
    case PIX_FMT_BGRA:\
874
        if (CONFIG_SMALL){\
875
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
876
            func(uint32_t,needAlpha)\
877
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
878
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
879
            }\
880
        }else{\
881
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
882
                func(uint32_t,1)\
883
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
884
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
885
                }\
886
            }else{\
887
                func(uint32_t,0)\
888
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
889
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
890
                }\
891
            }\
892
        }\
893
        break;\
894
    case PIX_FMT_ARGB:\
895
    case PIX_FMT_ABGR:\
896
        if (CONFIG_SMALL){\
897
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
898
            func(uint32_t,needAlpha)\
899
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
900
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
901
            }\
902
        }else{\
903
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
904
                func(uint32_t,1)\
905
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
906
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
907
                }\
908
            }else{\
909
                func(uint32_t,0)\
910
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
911
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
912
                }\
913
            }\
914
        }                \
915
        break;\
916
    case PIX_FMT_RGB24:\
917
        func(uint8_t,0)\
918
            ((uint8_t*)dest)[0]= r[Y1];\
919
            ((uint8_t*)dest)[1]= g[Y1];\
920
            ((uint8_t*)dest)[2]= b[Y1];\
921
            ((uint8_t*)dest)[3]= r[Y2];\
922
            ((uint8_t*)dest)[4]= g[Y2];\
923
            ((uint8_t*)dest)[5]= b[Y2];\
924
            dest+=6;\
925
        }\
926
        break;\
927
    case PIX_FMT_BGR24:\
928
        func(uint8_t,0)\
929
            ((uint8_t*)dest)[0]= b[Y1];\
930
            ((uint8_t*)dest)[1]= g[Y1];\
931
            ((uint8_t*)dest)[2]= r[Y1];\
932
            ((uint8_t*)dest)[3]= b[Y2];\
933
            ((uint8_t*)dest)[4]= g[Y2];\
934
            ((uint8_t*)dest)[5]= r[Y2];\
935
            dest+=6;\
936
        }\
937
        break;\
938
    case PIX_FMT_RGB565:\
939
    case PIX_FMT_BGR565:\
940
        {\
941
            const int dr1= dither_2x2_8[y&1    ][0];\
942
            const int dg1= dither_2x2_4[y&1    ][0];\
943
            const int db1= dither_2x2_8[(y&1)^1][0];\
944
            const int dr2= dither_2x2_8[y&1    ][1];\
945
            const int dg2= dither_2x2_4[y&1    ][1];\
946
            const int db2= dither_2x2_8[(y&1)^1][1];\
947
            func(uint16_t,0)\
948
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
949
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
950
            }\
951
        }\
952
        break;\
953
    case PIX_FMT_RGB555:\
954
    case PIX_FMT_BGR555:\
955
        {\
956
            const int dr1= dither_2x2_8[y&1    ][0];\
957
            const int dg1= dither_2x2_8[y&1    ][1];\
958
            const int db1= dither_2x2_8[(y&1)^1][0];\
959
            const int dr2= dither_2x2_8[y&1    ][1];\
960
            const int dg2= dither_2x2_8[y&1    ][0];\
961
            const int db2= dither_2x2_8[(y&1)^1][1];\
962
            func(uint16_t,0)\
963
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
964
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
965
            }\
966
        }\
967
        break;\
968
    case PIX_FMT_RGB8:\
969
    case PIX_FMT_BGR8:\
970
        {\
971
            const uint8_t * const d64= dither_8x8_73[y&7];\
972
            const uint8_t * const d32= dither_8x8_32[y&7];\
973
            func(uint8_t,0)\
974
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
975
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
976
            }\
977
        }\
978
        break;\
979
    case PIX_FMT_RGB4:\
980
    case PIX_FMT_BGR4:\
981
        {\
982
            const uint8_t * const d64= dither_8x8_73 [y&7];\
983
            const uint8_t * const d128=dither_8x8_220[y&7];\
984
            func(uint8_t,0)\
985
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
986
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
987
            }\
988
        }\
989
        break;\
990
    case PIX_FMT_RGB4_BYTE:\
991
    case PIX_FMT_BGR4_BYTE:\
992
        {\
993
            const uint8_t * const d64= dither_8x8_73 [y&7];\
994
            const uint8_t * const d128=dither_8x8_220[y&7];\
995
            func(uint8_t,0)\
996
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
997
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
998
            }\
999
        }\
1000
        break;\
1001
    case PIX_FMT_MONOBLACK:\
1002
    case PIX_FMT_MONOWHITE:\
1003
        {\
1004
            func_monoblack\
1005
        }\
1006
        break;\
1007
    case PIX_FMT_YUYV422:\
1008
        func2\
1009
            ((uint8_t*)dest)[2*i2+0]= Y1;\
1010
            ((uint8_t*)dest)[2*i2+1]= U;\
1011
            ((uint8_t*)dest)[2*i2+2]= Y2;\
1012
            ((uint8_t*)dest)[2*i2+3]= V;\
1013
        }                \
1014
        break;\
1015
    case PIX_FMT_UYVY422:\
1016
        func2\
1017
            ((uint8_t*)dest)[2*i2+0]= U;\
1018
            ((uint8_t*)dest)[2*i2+1]= Y1;\
1019
            ((uint8_t*)dest)[2*i2+2]= V;\
1020
            ((uint8_t*)dest)[2*i2+3]= Y2;\
1021
        }                \
1022
        break;\
1023
    case PIX_FMT_GRAY16BE:\
1024
        func_g16\
1025
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
1026
            ((uint8_t*)dest)[2*i2+1]= Y1;\
1027
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
1028
            ((uint8_t*)dest)[2*i2+3]= Y2;\
1029
        }                \
1030
        break;\
1031
    case PIX_FMT_GRAY16LE:\
1032
        func_g16\
1033
            ((uint8_t*)dest)[2*i2+0]= Y1;\
1034
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
1035
            ((uint8_t*)dest)[2*i2+2]= Y2;\
1036
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
1037
        }                \
1038
        break;\
1039
    }\
1040

    
1041

    
1042
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1043
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1044
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1045
{
1046
    int i;
1047
    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)
1048
}
1049

    
1050
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1051
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1052
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1053
{
1054
    int i;
1055
    int step= fmt_depth(c->dstFormat)/8;
1056
    int aidx= 3;
1057

    
1058
    switch(c->dstFormat){
1059
    case PIX_FMT_ARGB:
1060
        dest++;
1061
        aidx= 0;
1062
    case PIX_FMT_RGB24:
1063
        aidx--;
1064
    case PIX_FMT_RGBA:
1065
        if (CONFIG_SMALL){
1066
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1067
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1068
                dest[aidx]= needAlpha ? A : 255;
1069
                dest[0]= R>>22;
1070
                dest[1]= G>>22;
1071
                dest[2]= B>>22;
1072
                dest+= step;
1073
            }
1074
        }else{
1075
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1076
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1077
                    dest[aidx]= A;
1078
                    dest[0]= R>>22;
1079
                    dest[1]= G>>22;
1080
                    dest[2]= B>>22;
1081
                    dest+= step;
1082
                }
1083
            }else{
1084
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1085
                    dest[aidx]= 255;
1086
                    dest[0]= R>>22;
1087
                    dest[1]= G>>22;
1088
                    dest[2]= B>>22;
1089
                    dest+= step;
1090
                }
1091
            }
1092
        }
1093
        break;
1094
    case PIX_FMT_ABGR:
1095
        dest++;
1096
        aidx= 0;
1097
    case PIX_FMT_BGR24:
1098
        aidx--;
1099
    case PIX_FMT_BGRA:
1100
        if (CONFIG_SMALL){
1101
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1102
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1103
                dest[aidx]= needAlpha ? A : 255;
1104
                dest[0]= B>>22;
1105
                dest[1]= G>>22;
1106
                dest[2]= R>>22;
1107
                dest+= step;
1108
            }
1109
        }else{
1110
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1111
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1112
                    dest[aidx]= A;
1113
                    dest[0]= B>>22;
1114
                    dest[1]= G>>22;
1115
                    dest[2]= R>>22;
1116
                    dest+= step;
1117
                }
1118
            }else{
1119
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1120
                    dest[aidx]= 255;
1121
                    dest[0]= B>>22;
1122
                    dest[1]= G>>22;
1123
                    dest[2]= R>>22;
1124
                    dest+= step;
1125
                }
1126
            }
1127
        }
1128
        break;
1129
    default:
1130
        assert(0);
1131
    }
1132
}
1133

    
1134
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val){
1135
    int i;
1136
    uint8_t *ptr = plane + stride*y;
1137
    for (i=0; i<height; i++){
1138
        memset(ptr, val, width);
1139
        ptr += stride;
1140
    }
1141
}
1142

    
1143
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)
1144
{
1145
    int i;
1146
    for (i = 0; i < width; i++) {
1147
        int r = src[i*6+0];
1148
        int g = src[i*6+2];
1149
        int b = src[i*6+4];
1150

    
1151
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1152
    }
1153
}
1154

    
1155
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1156
                             uint8_t *src1, uint8_t *src2, int width)
1157
{
1158
    int i;
1159
    assert(src1==src2);
1160
    for (i = 0; i < width; i++) {
1161
        int r = src1[6*i + 0];
1162
        int g = src1[6*i + 2];
1163
        int b = src1[6*i + 4];
1164

    
1165
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1166
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1167
    }
1168
}
1169

    
1170
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1171
                                  uint8_t *src1, uint8_t *src2, int width)
1172
{
1173
    int i;
1174
    assert(src1==src2);
1175
    for (i = 0; i < width; i++) {
1176
        int r= src1[12*i + 0] + src1[12*i + 6];
1177
        int g= src1[12*i + 2] + src1[12*i + 8];
1178
        int b= src1[12*i + 4] + src1[12*i + 10];
1179

    
1180
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1181
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1182
    }
1183
}
1184

    
1185
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1186
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1187
{\
1188
    int i;\
1189
    for (i=0; i<width; i++)\
1190
    {\
1191
        int b= (((const type*)src)[i]>>shb)&maskb;\
1192
        int g= (((const type*)src)[i]>>shg)&maskg;\
1193
        int r= (((const type*)src)[i]>>shr)&maskr;\
1194
\
1195
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1196
    }\
1197
}
1198

    
1199
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1200
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1201
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1202
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1203
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1204
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1205

    
1206
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused){
1207
    int i;
1208
    for (i=0; i<width; i++){
1209
        dst[i]= src[4*i];
1210
    }
1211
}
1212

    
1213
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1214
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1215
{\
1216
    int i;\
1217
    for (i=0; i<width; i++)\
1218
    {\
1219
        int b= (((const type*)src)[i]&maskb)>>shb;\
1220
        int g= (((const type*)src)[i]&maskg)>>shg;\
1221
        int r= (((const type*)src)[i]&maskr)>>shr;\
1222
\
1223
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1224
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1225
    }\
1226
}\
1227
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1228
{\
1229
    int i;\
1230
    for (i=0; i<width; i++)\
1231
    {\
1232
        int pix0= ((const type*)src)[2*i+0];\
1233
        int pix1= ((const type*)src)[2*i+1];\
1234
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1235
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1236
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1237
        g&= maskg|(2*maskg);\
1238
\
1239
        g>>=shg;\
1240
\
1241
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1242
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1243
    }\
1244
}
1245

    
1246
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1247
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1248
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1249
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1250
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1251
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1252

    
1253
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1254
{
1255
    int i;
1256
    for (i=0; i<width; i++)
1257
    {
1258
        int d= src[i];
1259

    
1260
        dst[i]= pal[d] & 0xFF;
1261
    }
1262
}
1263

    
1264
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1265
                           const uint8_t *src1, const uint8_t *src2,
1266
                           long width, uint32_t *pal)
1267
{
1268
    int i;
1269
    assert(src1 == src2);
1270
    for (i=0; i<width; i++)
1271
    {
1272
        int p= pal[src1[i]];
1273

    
1274
        dstU[i]= p>>8;
1275
        dstV[i]= p>>16;
1276
    }
1277
}
1278

    
1279
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1280
{
1281
    int i, j;
1282
    for (i=0; i<width/8; i++){
1283
        int d= ~src[i];
1284
        for(j=0; j<8; j++)
1285
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1286
    }
1287
}
1288

    
1289
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1290
{
1291
    int i, j;
1292
    for (i=0; i<width/8; i++){
1293
        int d= src[i];
1294
        for(j=0; j<8; j++)
1295
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1296
    }
1297
}
1298

    
1299

    
1300
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1301
//Plain C versions
1302
#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1303
#define COMPILE_C
1304
#endif
1305

    
1306
#if ARCH_PPC
1307
#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1308
#define COMPILE_ALTIVEC
1309
#endif
1310
#endif //ARCH_PPC
1311

    
1312
#if ARCH_X86
1313

    
1314
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1315
#define COMPILE_MMX
1316
#endif
1317

    
1318
#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1319
#define COMPILE_MMX2
1320
#endif
1321

    
1322
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1323
#define COMPILE_3DNOW
1324
#endif
1325
#endif //ARCH_X86
1326

    
1327
#define COMPILE_TEMPLATE_MMX 0
1328
#define COMPILE_TEMPLATE_MMX2 0
1329
#define COMPILE_TEMPLATE_AMD3DNOW 0
1330
#define COMPILE_TEMPLATE_ALTIVEC 0
1331

    
1332
#ifdef COMPILE_C
1333
#define RENAME(a) a ## _C
1334
#include "swscale_template.c"
1335
#endif
1336

    
1337
#ifdef COMPILE_ALTIVEC
1338
#undef RENAME
1339
#undef COMPILE_TEMPLATE_ALTIVEC
1340
#define COMPILE_TEMPLATE_ALTIVEC 1
1341
#define RENAME(a) a ## _altivec
1342
#include "swscale_template.c"
1343
#endif
1344

    
1345
#if ARCH_X86
1346

    
1347
//MMX versions
1348
#ifdef COMPILE_MMX
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 0
1356
#define RENAME(a) a ## _MMX
1357
#include "swscale_template.c"
1358
#endif
1359

    
1360
//MMX2 versions
1361
#ifdef COMPILE_MMX2
1362
#undef RENAME
1363
#undef COMPILE_TEMPLATE_MMX
1364
#undef COMPILE_TEMPLATE_MMX2
1365
#undef COMPILE_TEMPLATE_AMD3DNOW
1366
#define COMPILE_TEMPLATE_MMX 1
1367
#define COMPILE_TEMPLATE_MMX2 1
1368
#define COMPILE_TEMPLATE_AMD3DNOW 0
1369
#define RENAME(a) a ## _MMX2
1370
#include "swscale_template.c"
1371
#endif
1372

    
1373
//3DNOW versions
1374
#ifdef COMPILE_3DNOW
1375
#undef RENAME
1376
#undef COMPILE_TEMPLATE_MMX
1377
#undef COMPILE_TEMPLATE_MMX2
1378
#undef COMPILE_TEMPLATE_AMD3DNOW
1379
#define COMPILE_TEMPLATE_MMX 1
1380
#define COMPILE_TEMPLATE_MMX2 0
1381
#define COMPILE_TEMPLATE_AMD3DNOW 1
1382
#define RENAME(a) a ## _3DNow
1383
#include "swscale_template.c"
1384
#endif
1385

    
1386
#endif //ARCH_X86
1387

    
1388
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1389
{
1390
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1391
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
1392
    else                return getSplineCoeff(        0.0,
1393
                                             b+ 2.0*c + 3.0*d,
1394
                                                    c + 3.0*d,
1395
                                            -b- 3.0*c - 6.0*d,
1396
                                            dist-1.0);
1397
}
1398

    
1399
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1400
                             int srcW, int dstW, int filterAlign, int one, int flags,
1401
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1402
{
1403
    int i;
1404
    int filterSize;
1405
    int filter2Size;
1406
    int minFilterSize;
1407
    int64_t *filter=NULL;
1408
    int64_t *filter2=NULL;
1409
    const int64_t fone= 1LL<<54;
1410
    int ret= -1;
1411
#if ARCH_X86
1412
    if (flags & SWS_CPU_CAPS_MMX)
1413
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1414
#endif
1415

    
1416
    // NOTE: the +1 is for the MMX scaler which reads over the end
1417
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1418

    
1419
    if (FFABS(xInc - 0x10000) <10) // unscaled
1420
    {
1421
        int i;
1422
        filterSize= 1;
1423
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1424

    
1425
        for (i=0; i<dstW; i++)
1426
        {
1427
            filter[i*filterSize]= fone;
1428
            (*filterPos)[i]=i;
1429
        }
1430

    
1431
    }
1432
    else if (flags&SWS_POINT) // lame looking point sampling mode
1433
    {
1434
        int i;
1435
        int xDstInSrc;
1436
        filterSize= 1;
1437
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1438

    
1439
        xDstInSrc= xInc/2 - 0x8000;
1440
        for (i=0; i<dstW; i++)
1441
        {
1442
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1443

    
1444
            (*filterPos)[i]= xx;
1445
            filter[i]= fone;
1446
            xDstInSrc+= xInc;
1447
        }
1448
    }
1449
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1450
    {
1451
        int i;
1452
        int xDstInSrc;
1453
        filterSize= 2;
1454
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1455

    
1456
        xDstInSrc= xInc/2 - 0x8000;
1457
        for (i=0; i<dstW; i++)
1458
        {
1459
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1460
            int j;
1461

    
1462
            (*filterPos)[i]= xx;
1463
                //bilinear upscale / linear interpolate / area averaging
1464
                for (j=0; j<filterSize; j++)
1465
                {
1466
                    int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1467
                    if (coeff<0) coeff=0;
1468
                    filter[i*filterSize + j]= coeff;
1469
                    xx++;
1470
                }
1471
            xDstInSrc+= xInc;
1472
        }
1473
    }
1474
    else
1475
    {
1476
        int xDstInSrc;
1477
        int sizeFactor;
1478

    
1479
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1480
        else if (flags&SWS_X)            sizeFactor=  8;
1481
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1482
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1483
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1484
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1485
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1486
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1487
        else {
1488
            sizeFactor= 0; //GCC warning killer
1489
            assert(0);
1490
        }
1491

    
1492
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1493
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1494

    
1495
        if (filterSize > srcW-2) filterSize=srcW-2;
1496

    
1497
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1498

    
1499
        xDstInSrc= xInc - 0x10000;
1500
        for (i=0; i<dstW; i++)
1501
        {
1502
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1503
            int j;
1504
            (*filterPos)[i]= xx;
1505
            for (j=0; j<filterSize; j++)
1506
            {
1507
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1508
                double floatd;
1509
                int64_t coeff;
1510

    
1511
                if (xInc > 1<<16)
1512
                    d= d*dstW/srcW;
1513
                floatd= d * (1.0/(1<<30));
1514

    
1515
                if (flags & SWS_BICUBIC)
1516
                {
1517
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1518
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1519
                    int64_t dd = ( d*d)>>30;
1520
                    int64_t ddd= (dd*d)>>30;
1521

    
1522
                    if      (d < 1LL<<30)
1523
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1524
                    else if (d < 1LL<<31)
1525
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1526
                    else
1527
                        coeff=0.0;
1528
                    coeff *= fone>>(30+24);
1529
                }
1530
/*                else if (flags & SWS_X)
1531
                {
1532
                    double p= param ? param*0.01 : 0.3;
1533
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1534
                    coeff*= pow(2.0, - p*d*d);
1535
                }*/
1536
                else if (flags & SWS_X)
1537
                {
1538
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1539
                    double c;
1540

    
1541
                    if (floatd<1.0)
1542
                        c = cos(floatd*PI);
1543
                    else
1544
                        c=-1.0;
1545
                    if (c<0.0)      c= -pow(-c, A);
1546
                    else            c=  pow( c, A);
1547
                    coeff= (c*0.5 + 0.5)*fone;
1548
                }
1549
                else if (flags & SWS_AREA)
1550
                {
1551
                    int64_t d2= d - (1<<29);
1552
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1553
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1554
                    else coeff=0.0;
1555
                    coeff *= fone>>(30+16);
1556
                }
1557
                else if (flags & SWS_GAUSS)
1558
                {
1559
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1560
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1561
                }
1562
                else if (flags & SWS_SINC)
1563
                {
1564
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1565
                }
1566
                else if (flags & SWS_LANCZOS)
1567
                {
1568
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1569
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1570
                    if (floatd>p) coeff=0;
1571
                }
1572
                else if (flags & SWS_BILINEAR)
1573
                {
1574
                    coeff= (1<<30) - d;
1575
                    if (coeff<0) coeff=0;
1576
                    coeff *= fone >> 30;
1577
                }
1578
                else if (flags & SWS_SPLINE)
1579
                {
1580
                    double p=-2.196152422706632;
1581
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1582
                }
1583
                else {
1584
                    coeff= 0.0; //GCC warning killer
1585
                    assert(0);
1586
                }
1587

    
1588
                filter[i*filterSize + j]= coeff;
1589
                xx++;
1590
            }
1591
            xDstInSrc+= 2*xInc;
1592
        }
1593
    }
1594

    
1595
    /* apply src & dst Filter to filter -> filter2
1596
       av_free(filter);
1597
    */
1598
    assert(filterSize>0);
1599
    filter2Size= filterSize;
1600
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1601
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1602
    assert(filter2Size>0);
1603
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1604

    
1605
    for (i=0; i<dstW; i++)
1606
    {
1607
        int j, k;
1608

    
1609
        if(srcFilter){
1610
            for (k=0; k<srcFilter->length; k++){
1611
                for (j=0; j<filterSize; j++)
1612
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1613
            }
1614
        }else{
1615
            for (j=0; j<filterSize; j++)
1616
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1617
        }
1618
        //FIXME dstFilter
1619

    
1620
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1621
    }
1622
    av_freep(&filter);
1623

    
1624
    /* try to reduce the filter-size (step1 find size and shift left) */
1625
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1626
    minFilterSize= 0;
1627
    for (i=dstW-1; i>=0; i--)
1628
    {
1629
        int min= filter2Size;
1630
        int j;
1631
        int64_t cutOff=0.0;
1632

    
1633
        /* get rid off near zero elements on the left by shifting left */
1634
        for (j=0; j<filter2Size; j++)
1635
        {
1636
            int k;
1637
            cutOff += FFABS(filter2[i*filter2Size]);
1638

    
1639
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1640

    
1641
            /* preserve monotonicity because the core can't handle the filter otherwise */
1642
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1643

    
1644
            // move filter coefficients left
1645
            for (k=1; k<filter2Size; k++)
1646
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1647
            filter2[i*filter2Size + k - 1]= 0;
1648
            (*filterPos)[i]++;
1649
        }
1650

    
1651
        cutOff=0;
1652
        /* count near zeros on the right */
1653
        for (j=filter2Size-1; j>0; j--)
1654
        {
1655
            cutOff += FFABS(filter2[i*filter2Size + j]);
1656

    
1657
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1658
            min--;
1659
        }
1660

    
1661
        if (min>minFilterSize) minFilterSize= min;
1662
    }
1663

    
1664
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1665
        // we can handle the special case 4,
1666
        // so we don't want to go to the full 8
1667
        if (minFilterSize < 5)
1668
            filterAlign = 4;
1669

    
1670
        // We really don't want to waste our time
1671
        // doing useless computation, so fall back on
1672
        // the scalar C code for very small filters.
1673
        // Vectorizing is worth it only if you have a
1674
        // decent-sized vector.
1675
        if (minFilterSize < 3)
1676
            filterAlign = 1;
1677
    }
1678

    
1679
    if (flags & SWS_CPU_CAPS_MMX) {
1680
        // special case for unscaled vertical filtering
1681
        if (minFilterSize == 1 && filterAlign == 2)
1682
            filterAlign= 1;
1683
    }
1684

    
1685
    assert(minFilterSize > 0);
1686
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1687
    assert(filterSize > 0);
1688
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1689
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1690
        goto error;
1691
    *outFilterSize= filterSize;
1692

    
1693
    if (flags&SWS_PRINT_INFO)
1694
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1695
    /* try to reduce the filter-size (step2 reduce it) */
1696
    for (i=0; i<dstW; i++)
1697
    {
1698
        int j;
1699

    
1700
        for (j=0; j<filterSize; j++)
1701
        {
1702
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1703
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1704
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1705
                filter[i*filterSize + j]= 0;
1706
        }
1707
    }
1708

    
1709

    
1710
    //FIXME try to align filterPos if possible
1711

    
1712
    //fix borders
1713
    for (i=0; i<dstW; i++)
1714
    {
1715
        int j;
1716
        if ((*filterPos)[i] < 0)
1717
        {
1718
            // move filter coefficients left to compensate for filterPos
1719
            for (j=1; j<filterSize; j++)
1720
            {
1721
                int left= FFMAX(j + (*filterPos)[i], 0);
1722
                filter[i*filterSize + left] += filter[i*filterSize + j];
1723
                filter[i*filterSize + j]=0;
1724
            }
1725
            (*filterPos)[i]= 0;
1726
        }
1727

    
1728
        if ((*filterPos)[i] + filterSize > srcW)
1729
        {
1730
            int shift= (*filterPos)[i] + filterSize - srcW;
1731
            // move filter coefficients right to compensate for filterPos
1732
            for (j=filterSize-2; j>=0; j--)
1733
            {
1734
                int right= FFMIN(j + shift, filterSize-1);
1735
                filter[i*filterSize +right] += filter[i*filterSize +j];
1736
                filter[i*filterSize +j]=0;
1737
            }
1738
            (*filterPos)[i]= srcW - filterSize;
1739
        }
1740
    }
1741

    
1742
    // Note the +1 is for the MMX scaler which reads over the end
1743
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1744
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1745

    
1746
    /* normalize & store in outFilter */
1747
    for (i=0; i<dstW; i++)
1748
    {
1749
        int j;
1750
        int64_t error=0;
1751
        int64_t sum=0;
1752

    
1753
        for (j=0; j<filterSize; j++)
1754
        {
1755
            sum+= filter[i*filterSize + j];
1756
        }
1757
        sum= (sum + one/2)/ one;
1758
        for (j=0; j<*outFilterSize; j++)
1759
        {
1760
            int64_t v= filter[i*filterSize + j] + error;
1761
            int intV= ROUNDED_DIV(v, sum);
1762
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1763
            error= v - intV*sum;
1764
        }
1765
    }
1766

    
1767
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1768
    for (i=0; i<*outFilterSize; i++)
1769
    {
1770
        int j= dstW*(*outFilterSize);
1771
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1772
    }
1773

    
1774
    ret=0;
1775
error:
1776
    av_free(filter);
1777
    av_free(filter2);
1778
    return ret;
1779
}
1780

    
1781
#ifdef COMPILE_MMX2
1782
static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1783
{
1784
    uint8_t *fragmentA;
1785
    x86_reg imm8OfPShufW1A;
1786
    x86_reg imm8OfPShufW2A;
1787
    x86_reg fragmentLengthA;
1788
    uint8_t *fragmentB;
1789
    x86_reg imm8OfPShufW1B;
1790
    x86_reg imm8OfPShufW2B;
1791
    x86_reg fragmentLengthB;
1792
    int fragmentPos;
1793

    
1794
    int xpos, i;
1795

    
1796
    // create an optimized horizontal scaling routine
1797

    
1798
    //code fragment
1799

    
1800
    __asm__ volatile(
1801
        "jmp                         9f                 \n\t"
1802
    // Begin
1803
        "0:                                             \n\t"
1804
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1805
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1806
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1807
        "punpcklbw                %%mm7, %%mm1          \n\t"
1808
        "punpcklbw                %%mm7, %%mm0          \n\t"
1809
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1810
        "1:                                             \n\t"
1811
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1812
        "2:                                             \n\t"
1813
        "psubw                    %%mm1, %%mm0          \n\t"
1814
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1815
        "pmullw                   %%mm3, %%mm0          \n\t"
1816
        "psllw                       $7, %%mm1          \n\t"
1817
        "paddw                    %%mm1, %%mm0          \n\t"
1818

    
1819
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1820

    
1821
        "add                         $8, %%"REG_a"      \n\t"
1822
    // End
1823
        "9:                                             \n\t"
1824
//        "int $3                                         \n\t"
1825
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1826
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1827
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1828
        "dec                         %1                 \n\t"
1829
        "dec                         %2                 \n\t"
1830
        "sub                         %0, %1             \n\t"
1831
        "sub                         %0, %2             \n\t"
1832
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1833
        "sub                         %0, %3             \n\t"
1834

    
1835

    
1836
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1837
        "=r" (fragmentLengthA)
1838
    );
1839

    
1840
    __asm__ volatile(
1841
        "jmp                         9f                 \n\t"
1842
    // Begin
1843
        "0:                                             \n\t"
1844
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1845
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1846
        "punpcklbw                %%mm7, %%mm0          \n\t"
1847
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1848
        "1:                                             \n\t"
1849
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1850
        "2:                                             \n\t"
1851
        "psubw                    %%mm1, %%mm0          \n\t"
1852
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1853
        "pmullw                   %%mm3, %%mm0          \n\t"
1854
        "psllw                       $7, %%mm1          \n\t"
1855
        "paddw                    %%mm1, %%mm0          \n\t"
1856

    
1857
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1858

    
1859
        "add                         $8, %%"REG_a"      \n\t"
1860
    // End
1861
        "9:                                             \n\t"
1862
//        "int                       $3                   \n\t"
1863
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1864
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1865
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1866
        "dec                         %1                 \n\t"
1867
        "dec                         %2                 \n\t"
1868
        "sub                         %0, %1             \n\t"
1869
        "sub                         %0, %2             \n\t"
1870
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1871
        "sub                         %0, %3             \n\t"
1872

    
1873

    
1874
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1875
        "=r" (fragmentLengthB)
1876
    );
1877

    
1878
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1879
    fragmentPos=0;
1880

    
1881
    for (i=0; i<dstW/numSplits; i++)
1882
    {
1883
        int xx=xpos>>16;
1884

    
1885
        if ((i&3) == 0)
1886
        {
1887
            int a=0;
1888
            int b=((xpos+xInc)>>16) - xx;
1889
            int c=((xpos+xInc*2)>>16) - xx;
1890
            int d=((xpos+xInc*3)>>16) - xx;
1891
            int inc                = (d+1<4);
1892
            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
1893
            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
1894
            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
1895
            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
1896
            int maxShift= 3-(d+inc);
1897
            int shift=0;
1898

    
1899
            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1900
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1901
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1902
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1903
            filterPos[i/2]= xx;
1904

    
1905
            memcpy(funnyCode + fragmentPos, fragment, fragmentLength);
1906

    
1907
            funnyCode[fragmentPos + imm8OfPShufW1]=
1908
                (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
1909
            funnyCode[fragmentPos + imm8OfPShufW2]=
1910
                a | (b<<2) | (c<<4) | (d<<6);
1911

    
1912
            if (i+4-inc>=dstW) shift=maxShift; //avoid overread
1913
            else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1914

    
1915
            if (shift && i>=shift)
1916
            {
1917
                funnyCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
1918
                funnyCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
1919
                filterPos[i/2]-=shift;
1920
            }
1921

    
1922
            fragmentPos+= fragmentLength;
1923

    
1924
            funnyCode[fragmentPos]= RET;
1925
        }
1926
        xpos+=xInc;
1927
    }
1928
    filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1929
}
1930
#endif /* COMPILE_MMX2 */
1931

    
1932
static void globalInit(void){
1933
    // generating tables:
1934
    int i;
1935
    for (i=0; i<768; i++){
1936
        int c= av_clip_uint8(i-256);
1937
        clip_table[i]=c;
1938
    }
1939
}
1940

    
1941
static SwsFunc getSwsFunc(SwsContext *c)
1942
{
1943
#if CONFIG_RUNTIME_CPUDETECT
1944
    int flags = c->flags;
1945

    
1946
#if ARCH_X86 && CONFIG_GPL
1947
    // ordered per speed fastest first
1948
    if (flags & SWS_CPU_CAPS_MMX2) {
1949
        sws_init_swScale_MMX2(c);
1950
        return swScale_MMX2;
1951
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1952
        sws_init_swScale_3DNow(c);
1953
        return swScale_3DNow;
1954
    } else if (flags & SWS_CPU_CAPS_MMX) {
1955
        sws_init_swScale_MMX(c);
1956
        return swScale_MMX;
1957
    } else {
1958
        sws_init_swScale_C(c);
1959
        return swScale_C;
1960
    }
1961

    
1962
#else
1963
#if ARCH_PPC
1964
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1965
        sws_init_swScale_altivec(c);
1966
        return swScale_altivec;
1967
    } else {
1968
        sws_init_swScale_C(c);
1969
        return swScale_C;
1970
    }
1971
#endif
1972
    sws_init_swScale_C(c);
1973
    return swScale_C;
1974
#endif /* ARCH_X86 && CONFIG_GPL */
1975
#else //CONFIG_RUNTIME_CPUDETECT
1976
#if   COMPILE_TEMPLATE_MMX2
1977
    sws_init_swScale_MMX2(c);
1978
    return swScale_MMX2;
1979
#elif COMPILE_TEMPLATE_AMD3DNOW
1980
    sws_init_swScale_3DNow(c);
1981
    return swScale_3DNow;
1982
#elif COMPILE_TEMPLATE_MMX
1983
    sws_init_swScale_MMX(c);
1984
    return swScale_MMX;
1985
#elif COMPILE_TEMPLATE_ALTIVEC
1986
    sws_init_swScale_altivec(c);
1987
    return swScale_altivec;
1988
#else
1989
    sws_init_swScale_C(c);
1990
    return swScale_C;
1991
#endif
1992
#endif //!CONFIG_RUNTIME_CPUDETECT
1993
}
1994

    
1995
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1996
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1997
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1998
    /* Copy Y plane */
1999
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2000
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
2001
    else
2002
    {
2003
        int i;
2004
        const uint8_t *srcPtr= src[0];
2005
        uint8_t *dstPtr= dst;
2006
        for (i=0; i<srcSliceH; i++)
2007
        {
2008
            memcpy(dstPtr, srcPtr, c->srcW);
2009
            srcPtr+= srcStride[0];
2010
            dstPtr+= dstStride[0];
2011
        }
2012
    }
2013
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
2014
    if (c->dstFormat == PIX_FMT_NV12)
2015
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
2016
    else
2017
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
2018

    
2019
    return srcSliceH;
2020
}
2021

    
2022
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2023
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2024
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2025

    
2026
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2027

    
2028
    return srcSliceH;
2029
}
2030

    
2031
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2032
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2033
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2034

    
2035
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2036

    
2037
    return srcSliceH;
2038
}
2039

    
2040
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2041
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2042
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2043

    
2044
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2045

    
2046
    return srcSliceH;
2047
}
2048

    
2049
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2050
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2051
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2052

    
2053
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2054

    
2055
    return srcSliceH;
2056
}
2057

    
2058
static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2059
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2060
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2061
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2062
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2063

    
2064
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2065

    
2066
    if (dstParam[3])
2067
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2068

    
2069
    return srcSliceH;
2070
}
2071

    
2072
static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2073
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2074
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2075
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2076
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2077

    
2078
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2079

    
2080
    return srcSliceH;
2081
}
2082

    
2083
static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2084
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2085
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2086
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2087
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2088

    
2089
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2090

    
2091
    if (dstParam[3])
2092
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2093

    
2094
    return srcSliceH;
2095
}
2096

    
2097
static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2098
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
2099
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2100
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2101
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2102

    
2103
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2104

    
2105
    return srcSliceH;
2106
}
2107

    
2108
static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2109
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
2110
    const enum PixelFormat srcFormat= c->srcFormat;
2111
    const enum PixelFormat dstFormat= c->dstFormat;
2112
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
2113
                 const uint8_t *palette)=NULL;
2114
    int i;
2115
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2116
    uint8_t *srcPtr= src[0];
2117

    
2118
    if (!usePal(srcFormat))
2119
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2120
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2121

    
2122
    switch(dstFormat){
2123
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
2124
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
2125
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
2126
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
2127
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
2128
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
2129
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2130
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2131
    }
2132

    
2133

    
2134
    for (i=0; i<srcSliceH; i++) {
2135
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
2136
        srcPtr+= srcStride[0];
2137
        dstPtr+= dstStride[0];
2138
    }
2139

    
2140
    return srcSliceH;
2141
}
2142

    
2143
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
2144
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2145
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
2146
    const enum PixelFormat srcFormat= c->srcFormat;
2147
    const enum PixelFormat dstFormat= c->dstFormat;
2148
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2149
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2150
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2151
    const int dstId= fmt_depth(dstFormat) >> 2;
2152
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2153

    
2154
    /* BGR -> BGR */
2155
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
2156
       || (isRGB(srcFormat) && isRGB(dstFormat))){
2157
        switch(srcId | (dstId<<4)){
2158
        case 0x34: conv= rgb16to15; break;
2159
        case 0x36: conv= rgb24to15; break;
2160
        case 0x38: conv= rgb32to15; break;
2161
        case 0x43: conv= rgb15to16; break;
2162
        case 0x46: conv= rgb24to16; break;
2163
        case 0x48: conv= rgb32to16; break;
2164
        case 0x63: conv= rgb15to24; break;
2165
        case 0x64: conv= rgb16to24; break;
2166
        case 0x68: conv= rgb32to24; break;
2167
        case 0x83: conv= rgb15to32; break;
2168
        case 0x84: conv= rgb16to32; break;
2169
        case 0x86: conv= rgb24to32; break;
2170
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2171
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2172
        }
2173
    }else if (  (isBGR(srcFormat) && isRGB(dstFormat))
2174
             || (isRGB(srcFormat) && isBGR(dstFormat))){
2175
        switch(srcId | (dstId<<4)){
2176
        case 0x33: conv= rgb15tobgr15; break;
2177
        case 0x34: conv= rgb16tobgr15; break;
2178
        case 0x36: conv= rgb24tobgr15; break;
2179
        case 0x38: conv= rgb32tobgr15; break;
2180
        case 0x43: conv= rgb15tobgr16; break;
2181
        case 0x44: conv= rgb16tobgr16; break;
2182
        case 0x46: conv= rgb24tobgr16; break;
2183
        case 0x48: conv= rgb32tobgr16; break;
2184
        case 0x63: conv= rgb15tobgr24; break;
2185
        case 0x64: conv= rgb16tobgr24; break;
2186
        case 0x66: conv= rgb24tobgr24; break;
2187
        case 0x68: conv= rgb32tobgr24; break;
2188
        case 0x83: conv= rgb15tobgr32; break;
2189
        case 0x84: conv= rgb16tobgr32; break;
2190
        case 0x86: conv= rgb24tobgr32; break;
2191
        case 0x88: conv= rgb32tobgr32; break;
2192
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2193
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2194
        }
2195
    }else{
2196
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2197
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2198
    }
2199

    
2200
    if(conv)
2201
    {
2202
        uint8_t *srcPtr= src[0];
2203
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2204
            srcPtr += ALT32_CORR;
2205

    
2206
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2207
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2208
        else
2209
        {
2210
            int i;
2211
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2212

    
2213
            for (i=0; i<srcSliceH; i++)
2214
            {
2215
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
2216
                srcPtr+= srcStride[0];
2217
                dstPtr+= dstStride[0];
2218
            }
2219
        }
2220
    }
2221
    return srcSliceH;
2222
}
2223

    
2224
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2225
                              int srcSliceH, uint8_t* dst[], int dstStride[]){
2226

    
2227
    rgb24toyv12(
2228
        src[0],
2229
        dst[0]+ srcSliceY    *dstStride[0],
2230
        dst[1]+(srcSliceY>>1)*dstStride[1],
2231
        dst[2]+(srcSliceY>>1)*dstStride[2],
2232
        c->srcW, srcSliceH,
2233
        dstStride[0], dstStride[1], srcStride[0]);
2234
    if (dst[3])
2235
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2236
    return srcSliceH;
2237
}
2238

    
2239
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2240
                             int srcSliceH, uint8_t* dst[], int dstStride[]){
2241
    int i;
2242

    
2243
    /* copy Y */
2244
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2245
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2246
    else{
2247
        uint8_t *srcPtr= src[0];
2248
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2249

    
2250
        for (i=0; i<srcSliceH; i++)
2251
        {
2252
            memcpy(dstPtr, srcPtr, c->srcW);
2253
            srcPtr+= srcStride[0];
2254
            dstPtr+= dstStride[0];
2255
        }
2256
    }
2257

    
2258
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){
2259
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2260
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
2261
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2262
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
2263
    }else{
2264
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2265
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
2266
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2267
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
2268
    }
2269
    if (dst[3])
2270
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2271
    return srcSliceH;
2272
}
2273

    
2274
/* unscaled copy like stuff (assumes nearly identical formats) */
2275
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2276
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2277
{
2278
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2279
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2280
    else
2281
    {
2282
        int i;
2283
        uint8_t *srcPtr= src[0];
2284
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2285
        int length=0;
2286

    
2287
        /* universal length finder */
2288
        while(length+c->srcW <= FFABS(dstStride[0])
2289
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2290
        assert(length!=0);
2291

    
2292
        for (i=0; i<srcSliceH; i++)
2293
        {
2294
            memcpy(dstPtr, srcPtr, length);
2295
            srcPtr+= srcStride[0];
2296
            dstPtr+= dstStride[0];
2297
        }
2298
    }
2299
    return srcSliceH;
2300
}
2301

    
2302
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2303
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2304
{
2305
    int plane, i, j;
2306
    for (plane=0; plane<4; plane++)
2307
    {
2308
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
2309
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2310
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
2311
        uint8_t *srcPtr= src[plane];
2312
        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2313

    
2314
        if (!dst[plane]) continue;
2315
        // ignore palette for GRAY8
2316
        if (plane == 1 && !dst[2]) continue;
2317
        if (!src[plane] || (plane == 1 && !src[2])){
2318
            if(is16BPS(c->dstFormat))
2319
                length*=2;
2320
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
2321
        }else
2322
        {
2323
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)){
2324
                if (!isBE(c->srcFormat)) srcPtr++;
2325
                for (i=0; i<height; i++){
2326
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2327
                    srcPtr+= srcStride[plane];
2328
                    dstPtr+= dstStride[plane];
2329
                }
2330
            }else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)){
2331
                for (i=0; i<height; i++){
2332
                    for (j=0; j<length; j++){
2333
                        dstPtr[ j<<1   ] = srcPtr[j];
2334
                        dstPtr[(j<<1)+1] = srcPtr[j];
2335
                    }
2336
                    srcPtr+= srcStride[plane];
2337
                    dstPtr+= dstStride[plane];
2338
                }
2339
            }else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2340
                  && isBE(c->srcFormat) != isBE(c->dstFormat)){
2341

    
2342
                for (i=0; i<height; i++){
2343
                    for (j=0; j<length; j++)
2344
                        ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2345
                    srcPtr+= srcStride[plane];
2346
                    dstPtr+= dstStride[plane];
2347
                }
2348
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2349
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2350
            else
2351
            {
2352
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2353
                    length*=2;
2354
                for (i=0; i<height; i++)
2355
                {
2356
                    memcpy(dstPtr, srcPtr, length);
2357
                    srcPtr+= srcStride[plane];
2358
                    dstPtr+= dstStride[plane];
2359
                }
2360
            }
2361
        }
2362
    }
2363
    return srcSliceH;
2364
}
2365

    
2366

    
2367
static void getSubSampleFactors(int *h, int *v, int format){
2368
    switch(format){
2369
    case PIX_FMT_UYVY422:
2370
    case PIX_FMT_YUYV422:
2371
        *h=1;
2372
        *v=0;
2373
        break;
2374
    case PIX_FMT_YUV420P:
2375
    case PIX_FMT_YUV420PLE:
2376
    case PIX_FMT_YUV420PBE:
2377
    case PIX_FMT_YUVA420P:
2378
    case PIX_FMT_GRAY16BE:
2379
    case PIX_FMT_GRAY16LE:
2380
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2381
    case PIX_FMT_NV12:
2382
    case PIX_FMT_NV21:
2383
        *h=1;
2384
        *v=1;
2385
        break;
2386
    case PIX_FMT_YUV440P:
2387
        *h=0;
2388
        *v=1;
2389
        break;
2390
    case PIX_FMT_YUV410P:
2391
        *h=2;
2392
        *v=2;
2393
        break;
2394
    case PIX_FMT_YUV444P:
2395
    case PIX_FMT_YUV444PLE:
2396
    case PIX_FMT_YUV444PBE:
2397
        *h=0;
2398
        *v=0;
2399
        break;
2400
    case PIX_FMT_YUV422P:
2401
    case PIX_FMT_YUV422PLE:
2402
    case PIX_FMT_YUV422PBE:
2403
        *h=1;
2404
        *v=0;
2405
        break;
2406
    case PIX_FMT_YUV411P:
2407
        *h=2;
2408
        *v=0;
2409
        break;
2410
    default:
2411
        *h=0;
2412
        *v=0;
2413
        break;
2414
    }
2415
}
2416

    
2417
static uint16_t roundToInt16(int64_t f){
2418
    int r= (f + (1<<15))>>16;
2419
         if (r<-0x7FFF) return 0x8000;
2420
    else if (r> 0x7FFF) return 0x7FFF;
2421
    else                return r;
2422
}
2423

    
2424
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2425
    int64_t crv =  inv_table[0];
2426
    int64_t cbu =  inv_table[1];
2427
    int64_t cgu = -inv_table[2];
2428
    int64_t cgv = -inv_table[3];
2429
    int64_t cy  = 1<<16;
2430
    int64_t oy  = 0;
2431

    
2432
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2433
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2434

    
2435
    c->brightness= brightness;
2436
    c->contrast  = contrast;
2437
    c->saturation= saturation;
2438
    c->srcRange  = srcRange;
2439
    c->dstRange  = dstRange;
2440
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2441

    
2442
    c->uOffset=   0x0400040004000400LL;
2443
    c->vOffset=   0x0400040004000400LL;
2444

    
2445
    if (!srcRange){
2446
        cy= (cy*255) / 219;
2447
        oy= 16<<16;
2448
    }else{
2449
        crv= (crv*224) / 255;
2450
        cbu= (cbu*224) / 255;
2451
        cgu= (cgu*224) / 255;
2452
        cgv= (cgv*224) / 255;
2453
    }
2454

    
2455
    cy = (cy *contrast             )>>16;
2456
    crv= (crv*contrast * saturation)>>32;
2457
    cbu= (cbu*contrast * saturation)>>32;
2458
    cgu= (cgu*contrast * saturation)>>32;
2459
    cgv= (cgv*contrast * saturation)>>32;
2460

    
2461
    oy -= 256*brightness;
2462

    
2463
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2464
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2465
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2466
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2467
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2468
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2469

    
2470
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2471
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2472
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2473
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2474
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2475
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2476

    
2477
    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2478
    //FIXME factorize
2479

    
2480
#ifdef COMPILE_ALTIVEC
2481
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2482
        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
2483
#endif
2484
    return 0;
2485
}
2486

    
2487
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2488
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2489

    
2490
    *inv_table = c->srcColorspaceTable;
2491
    *table     = c->dstColorspaceTable;
2492
    *srcRange  = c->srcRange;
2493
    *dstRange  = c->dstRange;
2494
    *brightness= c->brightness;
2495
    *contrast  = c->contrast;
2496
    *saturation= c->saturation;
2497

    
2498
    return 0;
2499
}
2500

    
2501
static int handle_jpeg(enum PixelFormat *format)
2502
{
2503
    switch (*format) {
2504
        case PIX_FMT_YUVJ420P:
2505
            *format = PIX_FMT_YUV420P;
2506
            return 1;
2507
        case PIX_FMT_YUVJ422P:
2508
            *format = PIX_FMT_YUV422P;
2509
            return 1;
2510
        case PIX_FMT_YUVJ444P:
2511
            *format = PIX_FMT_YUV444P;
2512
            return 1;
2513
        case PIX_FMT_YUVJ440P:
2514
            *format = PIX_FMT_YUV440P;
2515
            return 1;
2516
        default:
2517
            return 0;
2518
    }
2519
}
2520

    
2521
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2522
                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2523
{
2524

    
2525
    SwsContext *c;
2526
    int i;
2527
    int usesVFilter, usesHFilter;
2528
    int unscaled, needsDither;
2529
    int srcRange, dstRange;
2530
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2531
#if ARCH_X86
2532
    if (flags & SWS_CPU_CAPS_MMX)
2533
        __asm__ volatile("emms\n\t"::: "memory");
2534
#endif
2535

    
2536
#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
2537
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2538
#if   COMPILE_TEMPLATE_MMX2
2539
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2540
#elif COMPILE_TEMPLATE_AMD3DNOW
2541
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2542
#elif COMPILE_TEMPLATE_MMX
2543
    flags |= SWS_CPU_CAPS_MMX;
2544
#elif COMPILE_TEMPLATE_ALTIVEC
2545
    flags |= SWS_CPU_CAPS_ALTIVEC;
2546
#elif ARCH_BFIN
2547
    flags |= SWS_CPU_CAPS_BFIN;
2548
#endif
2549
#endif /* CONFIG_RUNTIME_CPUDETECT */
2550
    if (clip_table[512] != 255) globalInit();
2551
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2552

    
2553
    unscaled = (srcW == dstW && srcH == dstH);
2554
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2555
        && (fmt_depth(dstFormat))<24
2556
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2557

    
2558
    srcRange = handle_jpeg(&srcFormat);
2559
    dstRange = handle_jpeg(&dstFormat);
2560

    
2561
    if (!isSupportedIn(srcFormat))
2562
    {
2563
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2564
        return NULL;
2565
    }
2566
    if (!isSupportedOut(dstFormat))
2567
    {
2568
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2569
        return NULL;
2570
    }
2571

    
2572
    i= flags & ( SWS_POINT
2573
                |SWS_AREA
2574
                |SWS_BILINEAR
2575
                |SWS_FAST_BILINEAR
2576
                |SWS_BICUBIC
2577
                |SWS_X
2578
                |SWS_GAUSS
2579
                |SWS_LANCZOS
2580
                |SWS_SINC
2581
                |SWS_SPLINE
2582
                |SWS_BICUBLIN);
2583
    if(!i || (i & (i-1)))
2584
    {
2585
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2586
        return NULL;
2587
    }
2588

    
2589
    /* sanity check */
2590
    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
2591
    {
2592
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2593
               srcW, srcH, dstW, dstH);
2594
        return NULL;
2595
    }
2596
    if(srcW > VOFW || dstW > VOFW){
2597
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2598
        return NULL;
2599
    }
2600

    
2601
    if (!dstFilter) dstFilter= &dummyFilter;
2602
    if (!srcFilter) srcFilter= &dummyFilter;
2603

    
2604
    c= av_mallocz(sizeof(SwsContext));
2605

    
2606
    c->av_class = &sws_context_class;
2607
    c->srcW= srcW;
2608
    c->srcH= srcH;
2609
    c->dstW= dstW;
2610
    c->dstH= dstH;
2611
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2612
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2613
    c->flags= flags;
2614
    c->dstFormat= dstFormat;
2615
    c->srcFormat= srcFormat;
2616
    c->vRounder= 4* 0x0001000100010001ULL;
2617

    
2618
    usesHFilter= usesVFilter= 0;
2619
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2620
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2621
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2622
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2623
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2624
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2625
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2626
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2627

    
2628
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2629
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2630

    
2631
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2632
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2633

    
2634
    // drop some chroma lines if the user wants it
2635
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2636
    c->chrSrcVSubSample+= c->vChrDrop;
2637

    
2638
    // drop every other pixel for chroma calculation unless user wants full chroma
2639
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2640
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2641
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2642
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2643
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2644
        c->chrSrcHSubSample=1;
2645

    
2646
    if (param){
2647
        c->param[0] = param[0];
2648
        c->param[1] = param[1];
2649
    }else{
2650
        c->param[0] =
2651
        c->param[1] = SWS_PARAM_DEFAULT;
2652
    }
2653

    
2654
    // Note the -((-x)>>y) is so that we always round toward +inf.
2655
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2656
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2657
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2658
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2659

    
2660
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2661

    
2662
    /* unscaled special cases */
2663
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2664
    {
2665
        /* yv12_to_nv12 */
2666
        if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2667
        {
2668
            c->swScale= PlanarToNV12Wrapper;
2669
        }
2670
        /* yuv2bgr */
2671
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2672
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1))
2673
        {
2674
            c->swScale= ff_yuv2rgb_get_func_ptr(c);
2675
        }
2676

    
2677
        if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))
2678
        {
2679
            c->swScale= yvu9toyv12Wrapper;
2680
        }
2681

    
2682
        /* bgr24toYV12 */
2683
        if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2684
            c->swScale= bgr24toyv12Wrapper;
2685

    
2686
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2687
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2688
           && (isBGR(dstFormat) || isRGB(dstFormat))
2689
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2690
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2691
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2692
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2693
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2694
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2695
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2696
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2697
                                             && dstFormat != PIX_FMT_RGB32_1
2698
                                             && dstFormat != PIX_FMT_BGR32_1
2699
           && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
2700
           && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
2701
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2702
             c->swScale= rgb2rgbWrapper;
2703

    
2704
        if ((usePal(srcFormat) && (
2705
                 dstFormat == PIX_FMT_RGB32   ||
2706
                 dstFormat == PIX_FMT_RGB32_1 ||
2707
                 dstFormat == PIX_FMT_RGB24   ||
2708
                 dstFormat == PIX_FMT_BGR32   ||
2709
                 dstFormat == PIX_FMT_BGR32_1 ||
2710
                 dstFormat == PIX_FMT_BGR24)))
2711
             c->swScale= pal2rgbWrapper;
2712

    
2713
        if (srcFormat == PIX_FMT_YUV422P)
2714
        {
2715
            if (dstFormat == PIX_FMT_YUYV422)
2716
                c->swScale= YUV422PToYuy2Wrapper;
2717
            else if (dstFormat == PIX_FMT_UYVY422)
2718
                c->swScale= YUV422PToUyvyWrapper;
2719
        }
2720

    
2721
        /* LQ converters if -sws 0 or -sws 4*/
2722
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2723
            /* yv12_to_yuy2 */
2724
            if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)
2725
            {
2726
                if (dstFormat == PIX_FMT_YUYV422)
2727
                    c->swScale= PlanarToYuy2Wrapper;
2728
                else if (dstFormat == PIX_FMT_UYVY422)
2729
                    c->swScale= PlanarToUyvyWrapper;
2730
            }
2731
        }
2732
        if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2733
            c->swScale= YUYV2YUV420Wrapper;
2734
        if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2735
            c->swScale= UYVY2YUV420Wrapper;
2736
        if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2737
            c->swScale= YUYV2YUV422Wrapper;
2738
        if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2739
            c->swScale= UYVY2YUV422Wrapper;
2740

    
2741
#ifdef COMPILE_ALTIVEC
2742
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2743
            !(c->flags & SWS_BITEXACT) &&
2744
            srcFormat == PIX_FMT_YUV420P) {
2745
          // unscaled YV12 -> packed YUV, we want speed
2746
          if (dstFormat == PIX_FMT_YUYV422)
2747
              c->swScale= yv12toyuy2_unscaled_altivec;
2748
          else if (dstFormat == PIX_FMT_UYVY422)
2749
              c->swScale= yv12touyvy_unscaled_altivec;
2750
        }
2751
#endif
2752

    
2753
        /* simple copy */
2754
        if (  srcFormat == dstFormat
2755
            || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2756
            || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2757
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2758
            || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2759
            || (isGray(dstFormat) && isGray(srcFormat))
2760
            || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2761
                && c->chrDstHSubSample == c->chrSrcHSubSample
2762
                && c->chrDstVSubSample == c->chrSrcVSubSample))
2763
        {
2764
            if (isPacked(c->srcFormat))
2765
                c->swScale= packedCopy;
2766
            else /* Planar YUV or gray */
2767
                c->swScale= planarCopy;
2768
        }
2769
#if ARCH_BFIN
2770
        if (flags & SWS_CPU_CAPS_BFIN)
2771
            ff_bfin_get_unscaled_swscale (c);
2772
#endif
2773

    
2774
        if (c->swScale){
2775
            if (flags&SWS_PRINT_INFO)
2776
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2777
                                sws_format_name(srcFormat), sws_format_name(dstFormat));
2778
            return c;
2779
        }
2780
    }
2781

    
2782
    if (flags & SWS_CPU_CAPS_MMX2)
2783
    {
2784
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2785
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2786
        {
2787
            if (flags&SWS_PRINT_INFO)
2788
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2789
        }
2790
        if (usesHFilter) c->canMMX2BeUsed=0;
2791
    }
2792
    else
2793
        c->canMMX2BeUsed=0;
2794

    
2795
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2796
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2797

    
2798
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2799
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2800
    // n-2 is the last chrominance sample available
2801
    // this is not perfect, but no one should notice the difference, the more correct variant
2802
    // would be like the vertical one, but that would require some special code for the
2803
    // first and last pixel
2804
    if (flags&SWS_FAST_BILINEAR)
2805
    {
2806
        if (c->canMMX2BeUsed)
2807
        {
2808
            c->lumXInc+= 20;
2809
            c->chrXInc+= 20;
2810
        }
2811
        //we don't use the x86 asm scaler if MMX is available
2812
        else if (flags & SWS_CPU_CAPS_MMX)
2813
        {
2814
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2815
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2816
        }
2817
    }
2818

    
2819
    /* precalculate horizontal scaler filter coefficients */
2820
    {
2821
        const int filterAlign=
2822
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2823
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2824
            1;
2825

    
2826
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2827
                   srcW      ,       dstW, filterAlign, 1<<14,
2828
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2829
                   srcFilter->lumH, dstFilter->lumH, c->param);
2830
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2831
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2832
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2833
                   srcFilter->chrH, dstFilter->chrH, c->param);
2834

    
2835
#define MAX_FUNNY_CODE_SIZE 10000
2836
#if defined(COMPILE_MMX2)
2837
// can't downscale !!!
2838
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2839
        {
2840
#ifdef MAP_ANONYMOUS
2841
            c->funnyYCode  = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2842
            c->funnyUVCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2843
#elif HAVE_VIRTUALALLOC
2844
            c->funnyYCode  = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2845
            c->funnyUVCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2846
#else
2847
            c->funnyYCode  = av_malloc(MAX_FUNNY_CODE_SIZE);
2848
            c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2849
#endif
2850

    
2851
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2852
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2853
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2854
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2855

    
2856
            initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2857
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2858
        }
2859
#endif /* defined(COMPILE_MMX2) */
2860
    } // initialize horizontal stuff
2861

    
2862

    
2863

    
2864
    /* precalculate vertical scaler filter coefficients */
2865
    {
2866
        const int filterAlign=
2867
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2868
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2869
            1;
2870

    
2871
        initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2872
                   srcH      ,        dstH, filterAlign, (1<<12),
2873
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2874
                   srcFilter->lumV, dstFilter->lumV, c->param);
2875
        initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2876
                   c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2877
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2878
                   srcFilter->chrV, dstFilter->chrV, c->param);
2879

    
2880
#if HAVE_ALTIVEC
2881
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2882
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2883

    
2884
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2885
            int j;
2886
            short *p = (short *)&c->vYCoeffsBank[i];
2887
            for (j=0;j<8;j++)
2888
                p[j] = c->vLumFilter[i];
2889
        }
2890

    
2891
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2892
            int j;
2893
            short *p = (short *)&c->vCCoeffsBank[i];
2894
            for (j=0;j<8;j++)
2895
                p[j] = c->vChrFilter[i];
2896
        }
2897
#endif
2898
    }
2899

    
2900
    // calculate buffer sizes so that they won't run out while handling these damn slices
2901
    c->vLumBufSize= c->vLumFilterSize;
2902
    c->vChrBufSize= c->vChrFilterSize;
2903
    for (i=0; i<dstH; i++)
2904
    {
2905
        int chrI= i*c->chrDstH / dstH;
2906
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2907
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2908

    
2909
        nextSlice>>= c->chrSrcVSubSample;
2910
        nextSlice<<= c->chrSrcVSubSample;
2911
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2912
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2913
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2914
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2915
    }
2916

    
2917
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2918
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2919
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2920
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2921
        c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2922
    //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)
2923
    /* align at 16 bytes for AltiVec */
2924
    for (i=0; i<c->vLumBufSize; i++)
2925
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2926
    for (i=0; i<c->vChrBufSize; i++)
2927
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2928
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2929
        for (i=0; i<c->vLumBufSize; i++)
2930
            c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2931

    
2932
    //try to avoid drawing green stuff between the right end and the stride end
2933
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2934

    
2935
    assert(2*VOFW == VOF);
2936

    
2937
    assert(c->chrDstH <= dstH);
2938

    
2939
    if (flags&SWS_PRINT_INFO)
2940
    {
2941
#ifdef DITHER1XBPP
2942
        const char *dither= " dithered";
2943
#else
2944
        const char *dither= "";
2945
#endif
2946
        if (flags&SWS_FAST_BILINEAR)
2947
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2948
        else if (flags&SWS_BILINEAR)
2949
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2950
        else if (flags&SWS_BICUBIC)
2951
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2952
        else if (flags&SWS_X)
2953
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2954
        else if (flags&SWS_POINT)
2955
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2956
        else if (flags&SWS_AREA)
2957
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2958
        else if (flags&SWS_BICUBLIN)
2959
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2960
        else if (flags&SWS_GAUSS)
2961
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2962
        else if (flags&SWS_SINC)
2963
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2964
        else if (flags&SWS_LANCZOS)
2965
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2966
        else if (flags&SWS_SPLINE)
2967
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2968
        else
2969
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2970

    
2971
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2972
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2973
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2974
        else
2975
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2976
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2977

    
2978
        if (flags & SWS_CPU_CAPS_MMX2)
2979
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2980
        else if (flags & SWS_CPU_CAPS_3DNOW)
2981
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2982
        else if (flags & SWS_CPU_CAPS_MMX)
2983
            av_log(c, AV_LOG_INFO, "using MMX\n");
2984
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2985
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2986
        else
2987
            av_log(c, AV_LOG_INFO, "using C\n");
2988
    }
2989

    
2990
    if (flags & SWS_PRINT_INFO)
2991
    {
2992
        if (flags & SWS_CPU_CAPS_MMX)
2993
        {
2994
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2995
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2996
            else
2997
            {
2998
                if (c->hLumFilterSize==4)
2999
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
3000
                else if (c->hLumFilterSize==8)
3001
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
3002
                else
3003
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
3004

    
3005
                if (c->hChrFilterSize==4)
3006
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
3007
                else if (c->hChrFilterSize==8)
3008
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
3009
                else
3010
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
3011
            }
3012
        }
3013
        else
3014
        {
3015
#if ARCH_X86
3016
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
3017
#else
3018
            if (flags & SWS_FAST_BILINEAR)
3019
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
3020
            else
3021
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
3022
#endif
3023
        }
3024
        if (isPlanarYUV(dstFormat))
3025
        {
3026
            if (c->vLumFilterSize==1)
3027
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3028
            else
3029
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3030
        }
3031
        else
3032
        {
3033
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
3034
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
3035
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3036
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
3037
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3038
            else
3039
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3040
        }
3041

    
3042
        if (dstFormat==PIX_FMT_BGR24)
3043
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
3044
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
3045
        else if (dstFormat==PIX_FMT_RGB32)
3046
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3047
        else if (dstFormat==PIX_FMT_BGR565)
3048
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3049
        else if (dstFormat==PIX_FMT_BGR555)
3050
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3051

    
3052
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
3053
    }
3054
    if (flags & SWS_PRINT_INFO)
3055
    {
3056
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3057
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
3058
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3059
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
3060
    }
3061

    
3062
    c->swScale= getSwsFunc(c);
3063
    return c;
3064
}
3065

    
3066
static void reset_ptr(uint8_t* src[], int format){
3067
    if(!isALPHA(format))
3068
        src[3]=NULL;
3069
    if(!isPlanarYUV(format)){
3070
        src[3]=src[2]=NULL;
3071
        if(   format != PIX_FMT_PAL8
3072
           && format != PIX_FMT_RGB8
3073
           && format != PIX_FMT_BGR8
3074
           && format != PIX_FMT_RGB4_BYTE
3075
           && format != PIX_FMT_BGR4_BYTE
3076
          )
3077
            src[1]= NULL;
3078
    }
3079
}
3080

    
3081
/**
3082
 * swscale wrapper, so we don't need to export the SwsContext.
3083
 * Assumes planar YUV to be in YUV order instead of YVU.
3084
 */
3085
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3086
              int srcSliceH, uint8_t* dst[], int dstStride[]){
3087
    int i;
3088
    uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
3089
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
3090

    
3091
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
3092
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
3093
        return 0;
3094
    }
3095
    if (c->sliceDir == 0) {
3096
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
3097
    }
3098

    
3099
    if (usePal(c->srcFormat)){
3100
        for (i=0; i<256; i++){
3101
            int p, r, g, b,y,u,v;
3102
            if(c->srcFormat == PIX_FMT_PAL8){
3103
                p=((uint32_t*)(src[1]))[i];
3104
                r= (p>>16)&0xFF;
3105
                g= (p>> 8)&0xFF;
3106
                b=  p     &0xFF;
3107
            }else if(c->srcFormat == PIX_FMT_RGB8){
3108
                r= (i>>5    )*36;
3109
                g= ((i>>2)&7)*36;
3110
                b= (i&3     )*85;
3111
            }else if(c->srcFormat == PIX_FMT_BGR8){
3112
                b= (i>>6    )*85;
3113
                g= ((i>>3)&7)*36;
3114
                r= (i&7     )*36;
3115
            }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
3116
                r= (i>>3    )*255;
3117
                g= ((i>>1)&3)*85;
3118
                b= (i&1     )*255;
3119
            }else {
3120
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
3121
                b= (i>>3    )*255;
3122
                g= ((i>>1)&3)*85;
3123
                r= (i&1     )*255;
3124
            }
3125
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3126
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3127
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3128
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
3129

    
3130

    
3131
            switch(c->dstFormat) {
3132
            case PIX_FMT_BGR32:
3133
#if !HAVE_BIGENDIAN
3134
            case PIX_FMT_RGB24:
3135
#endif
3136
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
3137
                break;
3138
            case PIX_FMT_BGR32_1:
3139
#if HAVE_BIGENDIAN
3140
            case PIX_FMT_BGR24:
3141
#endif
3142
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
3143
                break;
3144
            case PIX_FMT_RGB32_1:
3145
#if HAVE_BIGENDIAN
3146
            case PIX_FMT_RGB24:
3147
#endif
3148
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3149
                break;
3150
            case PIX_FMT_RGB32:
3151
#if !HAVE_BIGENDIAN
3152
            case PIX_FMT_BGR24:
3153
#endif
3154
            default:
3155
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
3156
            }
3157
        }
3158
    }
3159

    
3160
    // copy strides, so they can safely be modified
3161
    if (c->sliceDir == 1) {
3162
        // slices go from top to bottom
3163
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3164
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3165

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

    
3169
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
3170
    } else {
3171
        // slices go from bottom to top => we flip the image internally
3172
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3173
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3174

    
3175
        src2[0] += (srcSliceH-1)*srcStride[0];
3176
        if (!usePal(c->srcFormat))
3177
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3178
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3179
        src2[3] += (srcSliceH-1)*srcStride[3];
3180
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
3181
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3182
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3183
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
3184

    
3185
        reset_ptr(src2, c->srcFormat);
3186
        reset_ptr(dst2, c->dstFormat);
3187

    
3188
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3189
    }
3190
}
3191

    
3192
#if LIBSWSCALE_VERSION_MAJOR < 1
3193
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3194
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
3195
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3196
}
3197
#endif
3198

    
3199
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3200
                                float lumaSharpen, float chromaSharpen,
3201
                                float chromaHShift, float chromaVShift,
3202
                                int verbose)
3203
{
3204
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3205

    
3206
    if (lumaGBlur!=0.0){
3207
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3208
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3209
    }else{
3210
        filter->lumH= sws_getIdentityVec();
3211
        filter->lumV= sws_getIdentityVec();
3212
    }
3213

    
3214
    if (chromaGBlur!=0.0){
3215
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3216
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3217
    }else{
3218
        filter->chrH= sws_getIdentityVec();
3219
        filter->chrV= sws_getIdentityVec();
3220
    }
3221

    
3222
    if (chromaSharpen!=0.0){
3223
        SwsVector *id= sws_getIdentityVec();
3224
        sws_scaleVec(filter->chrH, -chromaSharpen);
3225
        sws_scaleVec(filter->chrV, -chromaSharpen);
3226
        sws_addVec(filter->chrH, id);
3227
        sws_addVec(filter->chrV, id);
3228
        sws_freeVec(id);
3229
    }
3230

    
3231
    if (lumaSharpen!=0.0){
3232
        SwsVector *id= sws_getIdentityVec();
3233
        sws_scaleVec(filter->lumH, -lumaSharpen);
3234
        sws_scaleVec(filter->lumV, -lumaSharpen);
3235
        sws_addVec(filter->lumH, id);
3236
        sws_addVec(filter->lumV, id);
3237
        sws_freeVec(id);
3238
    }
3239

    
3240
    if (chromaHShift != 0.0)
3241
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3242

    
3243
    if (chromaVShift != 0.0)
3244
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3245

    
3246
    sws_normalizeVec(filter->chrH, 1.0);
3247
    sws_normalizeVec(filter->chrV, 1.0);
3248
    sws_normalizeVec(filter->lumH, 1.0);
3249
    sws_normalizeVec(filter->lumV, 1.0);
3250

    
3251
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3252
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3253

    
3254
    return filter;
3255
}
3256

    
3257
SwsVector *sws_getGaussianVec(double variance, double quality){
3258
    const int length= (int)(variance*quality + 0.5) | 1;
3259
    int i;
3260
    double *coeff= av_malloc(length*sizeof(double));
3261
    double middle= (length-1)*0.5;
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
    {
3269
        double dist= i-middle;
3270
        coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3271
    }
3272

    
3273
    sws_normalizeVec(vec, 1.0);
3274

    
3275
    return vec;
3276
}
3277

    
3278
SwsVector *sws_getConstVec(double c, int length){
3279
    int i;
3280
    double *coeff= av_malloc(length*sizeof(double));
3281
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3282

    
3283
    vec->coeff= coeff;
3284
    vec->length= length;
3285

    
3286
    for (i=0; i<length; i++)
3287
        coeff[i]= c;
3288

    
3289
    return vec;
3290
}
3291

    
3292

    
3293
SwsVector *sws_getIdentityVec(void){
3294
    return sws_getConstVec(1.0, 1);
3295
}
3296

    
3297
double sws_dcVec(SwsVector *a){
3298
    int i;
3299
    double sum=0;
3300

    
3301
    for (i=0; i<a->length; i++)
3302
        sum+= a->coeff[i];
3303

    
3304
    return sum;
3305
}
3306

    
3307
void sws_scaleVec(SwsVector *a, double scalar){
3308
    int i;
3309

    
3310
    for (i=0; i<a->length; i++)
3311
        a->coeff[i]*= scalar;
3312
}
3313

    
3314
void sws_normalizeVec(SwsVector *a, double height){
3315
    sws_scaleVec(a, height/sws_dcVec(a));
3316
}
3317

    
3318
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
3319
    int length= a->length + b->length - 1;
3320
    double *coeff= av_malloc(length*sizeof(double));
3321
    int i, j;
3322
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3323

    
3324
    vec->coeff= coeff;
3325
    vec->length= length;
3326

    
3327
    for (i=0; i<length; i++) coeff[i]= 0.0;
3328

    
3329
    for (i=0; i<a->length; i++)
3330
    {
3331
        for (j=0; j<b->length; j++)
3332
        {
3333
            coeff[i+j]+= a->coeff[i]*b->coeff[j];
3334
        }
3335
    }
3336

    
3337
    return vec;
3338
}
3339

    
3340
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
3341
    int length= FFMAX(a->length, b->length);
3342
    double *coeff= av_malloc(length*sizeof(double));
3343
    int i;
3344
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3345

    
3346
    vec->coeff= coeff;
3347
    vec->length= length;
3348

    
3349
    for (i=0; i<length; i++) coeff[i]= 0.0;
3350

    
3351
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3352
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3353

    
3354
    return vec;
3355
}
3356

    
3357
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
3358
    int length= FFMAX(a->length, b->length);
3359
    double *coeff= av_malloc(length*sizeof(double));
3360
    int i;
3361
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3362

    
3363
    vec->coeff= coeff;
3364
    vec->length= length;
3365

    
3366
    for (i=0; i<length; i++) coeff[i]= 0.0;
3367

    
3368
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3369
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3370

    
3371
    return vec;
3372
}
3373

    
3374
/* shift left / or right if "shift" is negative */
3375
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
3376
    int length= a->length + FFABS(shift)*2;
3377
    double *coeff= av_malloc(length*sizeof(double));
3378
    int i;
3379
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3380

    
3381
    vec->coeff= coeff;
3382
    vec->length= length;
3383

    
3384
    for (i=0; i<length; i++) coeff[i]= 0.0;
3385

    
3386
    for (i=0; i<a->length; i++)
3387
    {
3388
        coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3389
    }
3390

    
3391
    return vec;
3392
}
3393

    
3394
void sws_shiftVec(SwsVector *a, int shift){
3395
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3396
    av_free(a->coeff);
3397
    a->coeff= shifted->coeff;
3398
    a->length= shifted->length;
3399
    av_free(shifted);
3400
}
3401

    
3402
void sws_addVec(SwsVector *a, SwsVector *b){
3403
    SwsVector *sum= sws_sumVec(a, b);
3404
    av_free(a->coeff);
3405
    a->coeff= sum->coeff;
3406
    a->length= sum->length;
3407
    av_free(sum);
3408
}
3409

    
3410
void sws_subVec(SwsVector *a, SwsVector *b){
3411
    SwsVector *diff= sws_diffVec(a, b);
3412
    av_free(a->coeff);
3413
    a->coeff= diff->coeff;
3414
    a->length= diff->length;
3415
    av_free(diff);
3416
}
3417

    
3418
void sws_convVec(SwsVector *a, SwsVector *b){
3419
    SwsVector *conv= sws_getConvVec(a, b);
3420
    av_free(a->coeff);
3421
    a->coeff= conv->coeff;
3422
    a->length= conv->length;
3423
    av_free(conv);
3424
}
3425

    
3426
SwsVector *sws_cloneVec(SwsVector *a){
3427
    double *coeff= av_malloc(a->length*sizeof(double));
3428
    int i;
3429
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3430

    
3431
    vec->coeff= coeff;
3432
    vec->length= a->length;
3433

    
3434
    for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
3435

    
3436
    return vec;
3437
}
3438

    
3439
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){
3440
    int i;
3441
    double max=0;
3442
    double min=0;
3443
    double range;
3444

    
3445
    for (i=0; i<a->length; i++)
3446
        if (a->coeff[i]>max) max= a->coeff[i];
3447

    
3448
    for (i=0; i<a->length; i++)
3449
        if (a->coeff[i]<min) min= a->coeff[i];
3450

    
3451
    range= max - min;
3452

    
3453
    for (i=0; i<a->length; i++)
3454
    {
3455
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3456
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3457
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3458
        av_log(log_ctx, log_level, "|\n");
3459
    }
3460
}
3461

    
3462
#if LIBSWSCALE_VERSION_MAJOR < 1
3463
void sws_printVec(SwsVector *a){
3464
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3465
}
3466
#endif
3467

    
3468
void sws_freeVec(SwsVector *a){
3469
    if (!a) return;
3470
    av_freep(&a->coeff);
3471
    a->length=0;
3472
    av_free(a);
3473
}
3474

    
3475
void sws_freeFilter(SwsFilter *filter){
3476
    if (!filter) return;
3477

    
3478
    if (filter->lumH) sws_freeVec(filter->lumH);
3479
    if (filter->lumV) sws_freeVec(filter->lumV);
3480
    if (filter->chrH) sws_freeVec(filter->chrH);
3481
    if (filter->chrV) sws_freeVec(filter->chrV);
3482
    av_free(filter);
3483
}
3484

    
3485

    
3486
void sws_freeContext(SwsContext *c){
3487
    int i;
3488
    if (!c) return;
3489

    
3490
    if (c->lumPixBuf)
3491
    {
3492
        for (i=0; i<c->vLumBufSize; i++)
3493
            av_freep(&c->lumPixBuf[i]);
3494
        av_freep(&c->lumPixBuf);
3495
    }
3496

    
3497
    if (c->chrPixBuf)
3498
    {
3499
        for (i=0; i<c->vChrBufSize; i++)
3500
            av_freep(&c->chrPixBuf[i]);
3501
        av_freep(&c->chrPixBuf);
3502
    }
3503

    
3504
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
3505
        for (i=0; i<c->vLumBufSize; i++)
3506
            av_freep(&c->alpPixBuf[i]);
3507
        av_freep(&c->alpPixBuf);
3508
    }
3509

    
3510
    av_freep(&c->vLumFilter);
3511
    av_freep(&c->vChrFilter);
3512
    av_freep(&c->hLumFilter);
3513
    av_freep(&c->hChrFilter);
3514
#if HAVE_ALTIVEC
3515
    av_freep(&c->vYCoeffsBank);
3516
    av_freep(&c->vCCoeffsBank);
3517
#endif
3518

    
3519
    av_freep(&c->vLumFilterPos);
3520
    av_freep(&c->vChrFilterPos);
3521
    av_freep(&c->hLumFilterPos);
3522
    av_freep(&c->hChrFilterPos);
3523

    
3524
#if ARCH_X86 && CONFIG_GPL
3525
#ifdef MAP_ANONYMOUS
3526
    if (c->funnyYCode ) munmap(c->funnyYCode , MAX_FUNNY_CODE_SIZE);
3527
    if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
3528
#elif HAVE_VIRTUALALLOC
3529
    if (c->funnyYCode ) VirtualFree(c->funnyYCode , MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3530
    if (c->funnyUVCode) VirtualFree(c->funnyUVCode, MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3531
#else
3532
    av_free(c->funnyYCode );
3533
    av_free(c->funnyUVCode);
3534
#endif
3535
    c->funnyYCode=NULL;
3536
    c->funnyUVCode=NULL;
3537
#endif /* ARCH_X86 && CONFIG_GPL */
3538

    
3539
    av_freep(&c->lumMmx2Filter);
3540
    av_freep(&c->chrMmx2Filter);
3541
    av_freep(&c->lumMmx2FilterPos);
3542
    av_freep(&c->chrMmx2FilterPos);
3543
    av_freep(&c->yuvTable);
3544

    
3545
    av_free(c);
3546
}
3547

    
3548
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3549
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3550
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3551
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
3552
{
3553
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3554

    
3555
    if (!param)
3556
        param = default_param;
3557

    
3558
    if (context) {
3559
        if (context->srcW != srcW || context->srcH != srcH ||
3560
            context->srcFormat != srcFormat ||
3561
            context->dstW != dstW || context->dstH != dstH ||
3562
            context->dstFormat != dstFormat || context->flags != flags ||
3563
            context->param[0] != param[0] || context->param[1] != param[1])
3564
        {
3565
            sws_freeContext(context);
3566
            context = NULL;
3567
        }
3568
    }
3569
    if (!context) {
3570
        return sws_getContext(srcW, srcH, srcFormat,
3571
                              dstW, dstH, dstFormat, flags,
3572
                              srcFilter, dstFilter, param);
3573
    }
3574
    return context;
3575
}
3576