Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 0607b090

History | View | Annotate | Download (124 KB)

1
/*
2
 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3
 *
4
 * This file is part of FFmpeg.
5
 *
6
 * FFmpeg is free software; you can redistribute it and/or modify
7
 * it under the terms of the GNU General Public License as published by
8
 * the Free Software Foundation; either version 2 of the License, or
9
 * (at your option) any later version.
10
 *
11
 * FFmpeg is distributed in the hope that it will be useful,
12
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14
 * GNU General Public License for more details.
15
 *
16
 * You should have received a copy of the GNU General Public License
17
 * along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19
 *
20
 * the C code (not assembly, mmx, ...) of this file can be used
21
 * under the LGPL license too
22
 */
23

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

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

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

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

    
57
#define _SVID_SOURCE //needed for MAP_ANONYMOUS
58
#include <inttypes.h>
59
#include <string.h>
60
#include <math.h>
61
#include <stdio.h>
62
#include "config.h"
63
#include <assert.h>
64
#if HAVE_SYS_MMAN_H
65
#include <sys/mman.h>
66
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
67
#define MAP_ANONYMOUS MAP_ANON
68
#endif
69
#endif
70
#if HAVE_VIRTUALALLOC
71
#define WIN32_LEAN_AND_MEAN
72
#include <windows.h>
73
#endif
74
#include "swscale.h"
75
#include "swscale_internal.h"
76
#include "rgb2rgb.h"
77
#include "libavutil/intreadwrite.h"
78
#include "libavutil/x86_cpu.h"
79
#include "libavutil/bswap.h"
80

    
81
unsigned swscale_version(void)
82
{
83
    return LIBSWSCALE_VERSION_INT;
84
}
85

    
86
#undef MOVNTQ
87
#undef PAVGB
88

    
89
//#undef HAVE_MMX2
90
//#define HAVE_AMD3DNOW
91
//#undef HAVE_MMX
92
//#undef ARCH_X86
93
#define DITHER1XBPP
94

    
95
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
96

    
97
#define RET 0xC3 //near return opcode for x86
98

    
99
#ifdef M_PI
100
#define PI M_PI
101
#else
102
#define PI 3.14159265358979323846
103
#endif
104

    
105
#define isSupportedIn(x)    (       \
106
           (x)==PIX_FMT_YUV420P     \
107
        || (x)==PIX_FMT_YUVA420P    \
108
        || (x)==PIX_FMT_YUYV422     \
109
        || (x)==PIX_FMT_UYVY422     \
110
        || (x)==PIX_FMT_RGB48BE     \
111
        || (x)==PIX_FMT_RGB48LE     \
112
        || (x)==PIX_FMT_RGB32       \
113
        || (x)==PIX_FMT_RGB32_1     \
114
        || (x)==PIX_FMT_BGR24       \
115
        || (x)==PIX_FMT_BGR565      \
116
        || (x)==PIX_FMT_BGR555      \
117
        || (x)==PIX_FMT_BGR32       \
118
        || (x)==PIX_FMT_BGR32_1     \
119
        || (x)==PIX_FMT_RGB24       \
120
        || (x)==PIX_FMT_RGB565      \
121
        || (x)==PIX_FMT_RGB555      \
122
        || (x)==PIX_FMT_GRAY8       \
123
        || (x)==PIX_FMT_YUV410P     \
124
        || (x)==PIX_FMT_YUV440P     \
125
        || (x)==PIX_FMT_GRAY16BE    \
126
        || (x)==PIX_FMT_GRAY16LE    \
127
        || (x)==PIX_FMT_YUV444P     \
128
        || (x)==PIX_FMT_YUV422P     \
129
        || (x)==PIX_FMT_YUV411P     \
130
        || (x)==PIX_FMT_PAL8        \
131
        || (x)==PIX_FMT_BGR8        \
132
        || (x)==PIX_FMT_RGB8        \
133
        || (x)==PIX_FMT_BGR4_BYTE   \
134
        || (x)==PIX_FMT_RGB4_BYTE   \
135
        || (x)==PIX_FMT_YUV440P     \
136
        || (x)==PIX_FMT_MONOWHITE   \
137
        || (x)==PIX_FMT_MONOBLACK   \
138
        || (x)==PIX_FMT_YUV420PLE   \
139
        || (x)==PIX_FMT_YUV422PLE   \
140
        || (x)==PIX_FMT_YUV444PLE   \
141
        || (x)==PIX_FMT_YUV420PBE   \
142
        || (x)==PIX_FMT_YUV422PBE   \
143
        || (x)==PIX_FMT_YUV444PBE   \
144
    )
145
#define isSupportedOut(x)   (       \
146
           (x)==PIX_FMT_YUV420P     \
147
        || (x)==PIX_FMT_YUVA420P    \
148
        || (x)==PIX_FMT_YUYV422     \
149
        || (x)==PIX_FMT_UYVY422     \
150
        || (x)==PIX_FMT_YUV444P     \
151
        || (x)==PIX_FMT_YUV422P     \
152
        || (x)==PIX_FMT_YUV411P     \
153
        || isRGB(x)                 \
154
        || isBGR(x)                 \
155
        || (x)==PIX_FMT_NV12        \
156
        || (x)==PIX_FMT_NV21        \
157
        || (x)==PIX_FMT_GRAY16BE    \
158
        || (x)==PIX_FMT_GRAY16LE    \
159
        || (x)==PIX_FMT_GRAY8       \
160
        || (x)==PIX_FMT_YUV410P     \
161
        || (x)==PIX_FMT_YUV440P     \
162
        || (x)==PIX_FMT_YUV420PLE   \
163
        || (x)==PIX_FMT_YUV422PLE   \
164
        || (x)==PIX_FMT_YUV444PLE   \
165
        || (x)==PIX_FMT_YUV420PBE   \
166
        || (x)==PIX_FMT_YUV422PBE   \
167
        || (x)==PIX_FMT_YUV444PBE   \
168
    )
169
#define isPacked(x)         (       \
170
           (x)==PIX_FMT_PAL8        \
171
        || (x)==PIX_FMT_YUYV422     \
172
        || (x)==PIX_FMT_UYVY422     \
173
        || isRGB(x)                 \
174
        || isBGR(x)                 \
175
    )
176
#define usePal(x)           (       \
177
           (x)==PIX_FMT_PAL8        \
178
        || (x)==PIX_FMT_BGR4_BYTE   \
179
        || (x)==PIX_FMT_RGB4_BYTE   \
180
        || (x)==PIX_FMT_BGR8        \
181
        || (x)==PIX_FMT_RGB8        \
182
    )
183

    
184
#define RGB2YUV_SHIFT 15
185
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
186
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
187
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
188
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
189
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
190
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
191
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
192
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
193
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
194

    
195
extern const int32_t ff_yuv2rgb_coeffs[8][4];
196

    
197
static const double rgb2yuv_table[8][9]={
198
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
199
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
200
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
201
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
202
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
203
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
204
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
205
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
206
};
207

    
208
/*
209
NOTES
210
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
211

212
TODO
213
more intelligent misalignment avoidance for the horizontal scaler
214
write special vertical cubic upscale version
215
optimize C code (YV12 / minmax)
216
add support for packed pixel YUV input & output
217
add support for Y8 output
218
optimize BGR24 & BGR32
219
add BGR4 output support
220
write special BGR->BGR scaler
221
*/
222

    
223
#if ARCH_X86 && CONFIG_GPL
224
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
225
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
226
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
227
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
228
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
229
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
230
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
231
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
232

    
233
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
234
        0x0103010301030103LL,
235
        0x0200020002000200LL,};
236

    
237
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
238
        0x0602060206020602LL,
239
        0x0004000400040004LL,};
240

    
241
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
242
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
243
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
244
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
245
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
246
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
247

    
248
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
249
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
250
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
251

    
252
#ifdef FAST_BGR2YV12
253
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
254
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
255
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
256
#else
257
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
258
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
259
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
260
#endif /* FAST_BGR2YV12 */
261
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
262
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
263
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
264

    
265
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
266
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
267
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
268
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
269
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
270

    
271
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
272
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
273
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
274
};
275

    
276
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
277

    
278
#endif /* ARCH_X86 && CONFIG_GPL */
279

    
280
// clipping helper table for C implementations:
281
static unsigned char clip_table[768];
282

    
283
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
284

    
285
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4[2][8])={
286
{  1,   3,   1,   3,   1,   3,   1,   3, },
287
{  2,   0,   2,   0,   2,   0,   2,   0, },
288
};
289

    
290
DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8[2][8])={
291
{  6,   2,   6,   2,   6,   2,   6,   2, },
292
{  0,   4,   0,   4,   0,   4,   0,   4, },
293
};
294

    
295
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32[8][8])={
296
{ 17,   9,  23,  15,  16,   8,  22,  14, },
297
{  5,  29,   3,  27,   4,  28,   2,  26, },
298
{ 21,  13,  19,  11,  20,  12,  18,  10, },
299
{  0,  24,   6,  30,   1,  25,   7,  31, },
300
{ 16,   8,  22,  14,  17,   9,  23,  15, },
301
{  4,  28,   2,  26,   5,  29,   3,  27, },
302
{ 20,  12,  18,  10,  21,  13,  19,  11, },
303
{  1,  25,   7,  31,   0,  24,   6,  30, },
304
};
305

    
306
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73[8][8])={
307
{  0,  55,  14,  68,   3,  58,  17,  72, },
308
{ 37,  18,  50,  32,  40,  22,  54,  35, },
309
{  9,  64,   5,  59,  13,  67,   8,  63, },
310
{ 46,  27,  41,  23,  49,  31,  44,  26, },
311
{  2,  57,  16,  71,   1,  56,  15,  70, },
312
{ 39,  21,  52,  34,  38,  19,  51,  33, },
313
{ 11,  66,   7,  62,  10,  65,   6,  60, },
314
{ 48,  30,  43,  25,  47,  29,  42,  24, },
315
};
316

    
317
#if 1
318
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
319
{117,  62, 158, 103, 113,  58, 155, 100, },
320
{ 34, 199,  21, 186,  31, 196,  17, 182, },
321
{144,  89, 131,  76, 141,  86, 127,  72, },
322
{  0, 165,  41, 206,  10, 175,  52, 217, },
323
{110,  55, 151,  96, 120,  65, 162, 107, },
324
{ 28, 193,  14, 179,  38, 203,  24, 189, },
325
{138,  83, 124,  69, 148,  93, 134,  79, },
326
{  7, 172,  48, 213,   3, 168,  45, 210, },
327
};
328
#elif 1
329
// tries to correct a gamma of 1.5
330
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
331
{  0, 143,  18, 200,   2, 156,  25, 215, },
332
{ 78,  28, 125,  64,  89,  36, 138,  74, },
333
{ 10, 180,   3, 161,  16, 195,   8, 175, },
334
{109,  51,  93,  38, 121,  60, 105,  47, },
335
{  1, 152,  23, 210,   0, 147,  20, 205, },
336
{ 85,  33, 134,  71,  81,  30, 130,  67, },
337
{ 14, 190,   6, 171,  12, 185,   5, 166, },
338
{117,  57, 101,  44, 113,  54,  97,  41, },
339
};
340
#elif 1
341
// tries to correct a gamma of 2.0
342
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
343
{  0, 124,   8, 193,   0, 140,  12, 213, },
344
{ 55,  14, 104,  42,  66,  19, 119,  52, },
345
{  3, 168,   1, 145,   6, 187,   3, 162, },
346
{ 86,  31,  70,  21,  99,  39,  82,  28, },
347
{  0, 134,  11, 206,   0, 129,   9, 200, },
348
{ 62,  17, 114,  48,  58,  16, 109,  45, },
349
{  5, 181,   2, 157,   4, 175,   1, 151, },
350
{ 95,  36,  78,  26,  90,  34,  74,  24, },
351
};
352
#else
353
// tries to correct a gamma of 2.5
354
DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
355
{  0, 107,   3, 187,   0, 125,   6, 212, },
356
{ 39,   7,  86,  28,  49,  11, 102,  36, },
357
{  1, 158,   0, 131,   3, 180,   1, 151, },
358
{ 68,  19,  52,  12,  81,  25,  64,  17, },
359
{  0, 119,   5, 203,   0, 113,   4, 195, },
360
{ 45,   9,  96,  33,  42,   8,  91,  30, },
361
{  2, 172,   1, 144,   2, 165,   0, 137, },
362
{ 77,  23,  60,  15,  72,  21,  56,  14, },
363
};
364
#endif
365

    
366
const char *sws_format_name(enum PixelFormat format)
367
{
368
    switch (format) {
369
    case PIX_FMT_YUV420P:
370
        return "yuv420p";
371
    case PIX_FMT_YUVA420P:
372
        return "yuva420p";
373
    case PIX_FMT_YUYV422:
374
        return "yuyv422";
375
    case PIX_FMT_RGB24:
376
        return "rgb24";
377
    case PIX_FMT_BGR24:
378
        return "bgr24";
379
    case PIX_FMT_YUV422P:
380
        return "yuv422p";
381
    case PIX_FMT_YUV444P:
382
        return "yuv444p";
383
    case PIX_FMT_RGB32:
384
        return "rgb32";
385
    case PIX_FMT_YUV410P:
386
        return "yuv410p";
387
    case PIX_FMT_YUV411P:
388
        return "yuv411p";
389
    case PIX_FMT_RGB565:
390
        return "rgb565";
391
    case PIX_FMT_RGB555:
392
        return "rgb555";
393
    case PIX_FMT_GRAY16BE:
394
        return "gray16be";
395
    case PIX_FMT_GRAY16LE:
396
        return "gray16le";
397
    case PIX_FMT_GRAY8:
398
        return "gray8";
399
    case PIX_FMT_MONOWHITE:
400
        return "mono white";
401
    case PIX_FMT_MONOBLACK:
402
        return "mono black";
403
    case PIX_FMT_PAL8:
404
        return "Palette";
405
    case PIX_FMT_YUVJ420P:
406
        return "yuvj420p";
407
    case PIX_FMT_YUVJ422P:
408
        return "yuvj422p";
409
    case PIX_FMT_YUVJ444P:
410
        return "yuvj444p";
411
    case PIX_FMT_XVMC_MPEG2_MC:
412
        return "xvmc_mpeg2_mc";
413
    case PIX_FMT_XVMC_MPEG2_IDCT:
414
        return "xvmc_mpeg2_idct";
415
    case PIX_FMT_UYVY422:
416
        return "uyvy422";
417
    case PIX_FMT_UYYVYY411:
418
        return "uyyvyy411";
419
    case PIX_FMT_RGB32_1:
420
        return "rgb32x";
421
    case PIX_FMT_BGR32_1:
422
        return "bgr32x";
423
    case PIX_FMT_BGR32:
424
        return "bgr32";
425
    case PIX_FMT_BGR565:
426
        return "bgr565";
427
    case PIX_FMT_BGR555:
428
        return "bgr555";
429
    case PIX_FMT_BGR8:
430
        return "bgr8";
431
    case PIX_FMT_BGR4:
432
        return "bgr4";
433
    case PIX_FMT_BGR4_BYTE:
434
        return "bgr4 byte";
435
    case PIX_FMT_RGB8:
436
        return "rgb8";
437
    case PIX_FMT_RGB4:
438
        return "rgb4";
439
    case PIX_FMT_RGB4_BYTE:
440
        return "rgb4 byte";
441
    case PIX_FMT_RGB48BE:
442
        return "rgb48be";
443
    case PIX_FMT_RGB48LE:
444
        return "rgb48le";
445
    case PIX_FMT_NV12:
446
        return "nv12";
447
    case PIX_FMT_NV21:
448
        return "nv21";
449
    case PIX_FMT_YUV440P:
450
        return "yuv440p";
451
    case PIX_FMT_VDPAU_H264:
452
        return "vdpau_h264";
453
    case PIX_FMT_VDPAU_MPEG1:
454
        return "vdpau_mpeg1";
455
    case PIX_FMT_VDPAU_MPEG2:
456
        return "vdpau_mpeg2";
457
    case PIX_FMT_VDPAU_WMV3:
458
        return "vdpau_wmv3";
459
    case PIX_FMT_VDPAU_VC1:
460
        return "vdpau_vc1";
461
    case PIX_FMT_YUV420PLE:
462
        return "yuv420ple";
463
    case PIX_FMT_YUV422PLE:
464
        return "yuv422ple";
465
    case PIX_FMT_YUV444PLE:
466
        return "yuv444ple";
467
    case PIX_FMT_YUV420PBE:
468
        return "yuv420pbe";
469
    case PIX_FMT_YUV422PBE:
470
        return "yuv422pbe";
471
    case PIX_FMT_YUV444PBE:
472
        return "yuv444pbe";
473
    default:
474
        return "Unknown format";
475
    }
476
}
477

    
478
static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
479
                                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
480
                                                    const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
481
                                                    int dstW, int chrDstW, int big_endian)
482
{
483
    //FIXME Optimize (just quickly written not optimized..)
484
    int i;
485

    
486
    for (i = 0; i < dstW; i++) {
487
        int val = 1 << 10;
488
        int j;
489

    
490
        for (j = 0; j < lumFilterSize; j++)
491
            val += lumSrc[j][i] * lumFilter[j];
492

    
493
        if (big_endian) {
494
            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
495
        } else {
496
            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
497
        }
498
    }
499

    
500
    if (uDest) {
501
        for (i = 0; i < chrDstW; i++) {
502
            int u = 1 << 10;
503
            int v = 1 << 10;
504
            int j;
505

    
506
            for (j = 0; j < chrFilterSize; j++) {
507
                u += chrSrc[j][i       ] * chrFilter[j];
508
                v += chrSrc[j][i + VOFW] * chrFilter[j];
509
            }
510

    
511
            if (big_endian) {
512
                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
513
                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
514
            } else {
515
                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
516
                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
517
            }
518
        }
519
    }
520

    
521
    if (CONFIG_SWSCALE_ALPHA && aDest) {
522
        for (i = 0; i < dstW; i++) {
523
            int val = 1 << 10;
524
            int j;
525

    
526
            for (j = 0; j < lumFilterSize; j++)
527
                val += alpSrc[j][i] * lumFilter[j];
528

    
529
            if (big_endian) {
530
                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
531
            } else {
532
                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
533
            }
534
        }
535
    }
536
}
537

    
538
static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
539
                                 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
540
                                 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
541
                                 enum PixelFormat dstFormat)
542
{
543
    if (isBE(dstFormat)) {
544
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
545
                               chrFilter, chrSrc, chrFilterSize,
546
                               alpSrc,
547
                               dest, uDest, vDest, aDest,
548
                               dstW, chrDstW, 1);
549
    } else {
550
        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
551
                               chrFilter, chrSrc, chrFilterSize,
552
                               alpSrc,
553
                               dest, uDest, vDest, aDest,
554
                               dstW, chrDstW, 0);
555
    }
556
}
557

    
558
static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
559
                               const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
560
                               const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
561
{
562
    //FIXME Optimize (just quickly written not optimized..)
563
    int i;
564
    for (i=0; i<dstW; i++) {
565
        int val=1<<18;
566
        int j;
567
        for (j=0; j<lumFilterSize; j++)
568
            val += lumSrc[j][i] * lumFilter[j];
569

    
570
        dest[i]= av_clip_uint8(val>>19);
571
    }
572

    
573
    if (uDest)
574
        for (i=0; i<chrDstW; i++) {
575
            int u=1<<18;
576
            int v=1<<18;
577
            int j;
578
            for (j=0; j<chrFilterSize; j++) {
579
                u += chrSrc[j][i] * chrFilter[j];
580
                v += chrSrc[j][i + VOFW] * chrFilter[j];
581
            }
582

    
583
            uDest[i]= av_clip_uint8(u>>19);
584
            vDest[i]= av_clip_uint8(v>>19);
585
        }
586

    
587
    if (CONFIG_SWSCALE_ALPHA && aDest)
588
        for (i=0; i<dstW; i++) {
589
            int val=1<<18;
590
            int j;
591
            for (j=0; j<lumFilterSize; j++)
592
                val += alpSrc[j][i] * lumFilter[j];
593

    
594
            aDest[i]= av_clip_uint8(val>>19);
595
        }
596

    
597
}
598

    
599
static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
600
                                const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
601
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
602
{
603
    //FIXME Optimize (just quickly written not optimized..)
604
    int i;
605
    for (i=0; i<dstW; i++) {
606
        int val=1<<18;
607
        int j;
608
        for (j=0; j<lumFilterSize; j++)
609
            val += lumSrc[j][i] * lumFilter[j];
610

    
611
        dest[i]= av_clip_uint8(val>>19);
612
    }
613

    
614
    if (!uDest)
615
        return;
616

    
617
    if (dstFormat == PIX_FMT_NV12)
618
        for (i=0; i<chrDstW; i++) {
619
            int u=1<<18;
620
            int v=1<<18;
621
            int j;
622
            for (j=0; j<chrFilterSize; j++) {
623
                u += chrSrc[j][i] * chrFilter[j];
624
                v += chrSrc[j][i + VOFW] * chrFilter[j];
625
            }
626

    
627
            uDest[2*i]= av_clip_uint8(u>>19);
628
            uDest[2*i+1]= av_clip_uint8(v>>19);
629
        }
630
    else
631
        for (i=0; i<chrDstW; i++) {
632
            int u=1<<18;
633
            int v=1<<18;
634
            int j;
635
            for (j=0; j<chrFilterSize; j++) {
636
                u += chrSrc[j][i] * chrFilter[j];
637
                v += chrSrc[j][i + VOFW] * chrFilter[j];
638
            }
639

    
640
            uDest[2*i]= av_clip_uint8(v>>19);
641
            uDest[2*i+1]= av_clip_uint8(u>>19);
642
        }
643
}
644

    
645
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
646
    for (i=0; i<(dstW>>1); i++) {\
647
        int j;\
648
        int Y1 = 1<<18;\
649
        int Y2 = 1<<18;\
650
        int U  = 1<<18;\
651
        int V  = 1<<18;\
652
        int av_unused A1, A2;\
653
        type av_unused *r, *b, *g;\
654
        const int i2= 2*i;\
655
        \
656
        for (j=0; j<lumFilterSize; j++) {\
657
            Y1 += lumSrc[j][i2] * lumFilter[j];\
658
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
659
        }\
660
        for (j=0; j<chrFilterSize; j++) {\
661
            U += chrSrc[j][i] * chrFilter[j];\
662
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
663
        }\
664
        Y1>>=19;\
665
        Y2>>=19;\
666
        U >>=19;\
667
        V >>=19;\
668
        if (alpha) {\
669
            A1 = 1<<18;\
670
            A2 = 1<<18;\
671
            for (j=0; j<lumFilterSize; j++) {\
672
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
673
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
674
            }\
675
            A1>>=19;\
676
            A2>>=19;\
677
        }\
678

    
679
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
680
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
681
        if ((Y1|Y2|U|V)&256) {\
682
            if (Y1>255)   Y1=255; \
683
            else if (Y1<0)Y1=0;   \
684
            if (Y2>255)   Y2=255; \
685
            else if (Y2<0)Y2=0;   \
686
            if (U>255)    U=255;  \
687
            else if (U<0) U=0;    \
688
            if (V>255)    V=255;  \
689
            else if (V<0) V=0;    \
690
        }\
691
        if (alpha && ((A1|A2)&256)) {\
692
            A1=av_clip_uint8(A1);\
693
            A2=av_clip_uint8(A2);\
694
        }
695

    
696
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
697
    for (i=0; i<dstW; i++) {\
698
        int j;\
699
        int Y = 0;\
700
        int U = -128<<19;\
701
        int V = -128<<19;\
702
        int av_unused A;\
703
        int R,G,B;\
704
        \
705
        for (j=0; j<lumFilterSize; j++) {\
706
            Y += lumSrc[j][i     ] * lumFilter[j];\
707
        }\
708
        for (j=0; j<chrFilterSize; j++) {\
709
            U += chrSrc[j][i     ] * chrFilter[j];\
710
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
711
        }\
712
        Y >>=10;\
713
        U >>=10;\
714
        V >>=10;\
715
        if (alpha) {\
716
            A = rnd;\
717
            for (j=0; j<lumFilterSize; j++)\
718
                A += alpSrc[j][i     ] * lumFilter[j];\
719
            A >>=19;\
720
            if (A&256)\
721
                A = av_clip_uint8(A);\
722
        }\
723

    
724
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
725
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
726
        Y-= c->yuv2rgb_y_offset;\
727
        Y*= c->yuv2rgb_y_coeff;\
728
        Y+= rnd;\
729
        R= Y + V*c->yuv2rgb_v2r_coeff;\
730
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
731
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
732
        if ((R|G|B)&(0xC0000000)) {\
733
            if (R>=(256<<22))   R=(256<<22)-1; \
734
            else if (R<0)R=0;   \
735
            if (G>=(256<<22))   G=(256<<22)-1; \
736
            else if (G<0)G=0;   \
737
            if (B>=(256<<22))   B=(256<<22)-1; \
738
            else if (B<0)B=0;   \
739
        }\
740

    
741

    
742
#define YSCALE_YUV_2_GRAY16_C \
743
    for (i=0; i<(dstW>>1); i++) {\
744
        int j;\
745
        int Y1 = 1<<18;\
746
        int Y2 = 1<<18;\
747
        int U  = 1<<18;\
748
        int V  = 1<<18;\
749
        \
750
        const int i2= 2*i;\
751
        \
752
        for (j=0; j<lumFilterSize; j++) {\
753
            Y1 += lumSrc[j][i2] * lumFilter[j];\
754
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
755
        }\
756
        Y1>>=11;\
757
        Y2>>=11;\
758
        if ((Y1|Y2|U|V)&65536) {\
759
            if (Y1>65535)   Y1=65535; \
760
            else if (Y1<0)Y1=0;   \
761
            if (Y2>65535)   Y2=65535; \
762
            else if (Y2<0)Y2=0;   \
763
        }
764

    
765
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
766
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
767
    r = (type *)c->table_rV[V];   \
768
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
769
    b = (type *)c->table_bU[U];   \
770

    
771
#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
772
    for (i=0; i<(dstW>>1); i++) { \
773
        const int i2= 2*i;       \
774
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
775
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
776
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
777
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
778
        type av_unused *r, *b, *g;                                    \
779
        int av_unused A1, A2;                                         \
780
        if (alpha) {\
781
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
782
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
783
        }\
784

    
785
#define YSCALE_YUV_2_GRAY16_2_C   \
786
    for (i=0; i<(dstW>>1); i++) { \
787
        const int i2= 2*i;       \
788
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
789
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
790

    
791
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
792
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
793
    r = (type *)c->table_rV[V];\
794
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
795
    b = (type *)c->table_bU[U];\
796

    
797
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
798
    for (i=0; i<(dstW>>1); i++) {\
799
        const int i2= 2*i;\
800
        int Y1= buf0[i2  ]>>7;\
801
        int Y2= buf0[i2+1]>>7;\
802
        int U= (uvbuf1[i     ])>>7;\
803
        int V= (uvbuf1[i+VOFW])>>7;\
804
        type av_unused *r, *b, *g;\
805
        int av_unused A1, A2;\
806
        if (alpha) {\
807
            A1= abuf0[i2  ]>>7;\
808
            A2= abuf0[i2+1]>>7;\
809
        }\
810

    
811
#define YSCALE_YUV_2_GRAY16_1_C \
812
    for (i=0; i<(dstW>>1); i++) {\
813
        const int i2= 2*i;\
814
        int Y1= buf0[i2  ]<<1;\
815
        int Y2= buf0[i2+1]<<1;\
816

    
817
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
818
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
819
    r = (type *)c->table_rV[V];\
820
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
821
    b = (type *)c->table_bU[U];\
822

    
823
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
824
    for (i=0; i<(dstW>>1); i++) {\
825
        const int i2= 2*i;\
826
        int Y1= buf0[i2  ]>>7;\
827
        int Y2= buf0[i2+1]>>7;\
828
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
829
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
830
        type av_unused *r, *b, *g;\
831
        int av_unused A1, A2;\
832
        if (alpha) {\
833
            A1= abuf0[i2  ]>>7;\
834
            A2= abuf0[i2+1]>>7;\
835
        }\
836

    
837
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
838
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
839
    r = (type *)c->table_rV[V];\
840
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
841
    b = (type *)c->table_bU[U];\
842

    
843
#define YSCALE_YUV_2_MONO2_C \
844
    const uint8_t * const d128=dither_8x8_220[y&7];\
845
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
846
    for (i=0; i<dstW-7; i+=8) {\
847
        int acc;\
848
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
849
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
850
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
851
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
852
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
853
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
854
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
855
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
856
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
857
        dest++;\
858
    }\
859

    
860

    
861
#define YSCALE_YUV_2_MONOX_C \
862
    const uint8_t * const d128=dither_8x8_220[y&7];\
863
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
864
    int acc=0;\
865
    for (i=0; i<dstW-1; i+=2) {\
866
        int j;\
867
        int Y1=1<<18;\
868
        int Y2=1<<18;\
869
\
870
        for (j=0; j<lumFilterSize; j++) {\
871
            Y1 += lumSrc[j][i] * lumFilter[j];\
872
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
873
        }\
874
        Y1>>=19;\
875
        Y2>>=19;\
876
        if ((Y1|Y2)&256) {\
877
            if (Y1>255)   Y1=255;\
878
            else if (Y1<0)Y1=0;\
879
            if (Y2>255)   Y2=255;\
880
            else if (Y2<0)Y2=0;\
881
        }\
882
        acc+= acc + g[Y1+d128[(i+0)&7]];\
883
        acc+= acc + g[Y2+d128[(i+1)&7]];\
884
        if ((i&7)==6) {\
885
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
886
            dest++;\
887
        }\
888
    }
889

    
890

    
891
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
892
    switch(c->dstFormat) {\
893
    case PIX_FMT_RGB48BE:\
894
    case PIX_FMT_RGB48LE:\
895
        func(uint8_t,0)\
896
            ((uint8_t*)dest)[ 0]= r[Y1];\
897
            ((uint8_t*)dest)[ 1]= r[Y1];\
898
            ((uint8_t*)dest)[ 2]= g[Y1];\
899
            ((uint8_t*)dest)[ 3]= g[Y1];\
900
            ((uint8_t*)dest)[ 4]= b[Y1];\
901
            ((uint8_t*)dest)[ 5]= b[Y1];\
902
            ((uint8_t*)dest)[ 6]= r[Y2];\
903
            ((uint8_t*)dest)[ 7]= r[Y2];\
904
            ((uint8_t*)dest)[ 8]= g[Y2];\
905
            ((uint8_t*)dest)[ 9]= g[Y2];\
906
            ((uint8_t*)dest)[10]= b[Y2];\
907
            ((uint8_t*)dest)[11]= b[Y2];\
908
            dest+=12;\
909
        }\
910
        break;\
911
    case PIX_FMT_RGBA:\
912
    case PIX_FMT_BGRA:\
913
        if (CONFIG_SMALL) {\
914
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
915
            func(uint32_t,needAlpha)\
916
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
917
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
918
            }\
919
        } else {\
920
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
921
                func(uint32_t,1)\
922
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
923
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
924
                }\
925
            } else {\
926
                func(uint32_t,0)\
927
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
928
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
929
                }\
930
            }\
931
        }\
932
        break;\
933
    case PIX_FMT_ARGB:\
934
    case PIX_FMT_ABGR:\
935
        if (CONFIG_SMALL) {\
936
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
937
            func(uint32_t,needAlpha)\
938
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
939
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
940
            }\
941
        } else {\
942
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
943
                func(uint32_t,1)\
944
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
945
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
946
                }\
947
            } else {\
948
                func(uint32_t,0)\
949
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
950
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
951
                }\
952
            }\
953
        }                \
954
        break;\
955
    case PIX_FMT_RGB24:\
956
        func(uint8_t,0)\
957
            ((uint8_t*)dest)[0]= r[Y1];\
958
            ((uint8_t*)dest)[1]= g[Y1];\
959
            ((uint8_t*)dest)[2]= b[Y1];\
960
            ((uint8_t*)dest)[3]= r[Y2];\
961
            ((uint8_t*)dest)[4]= g[Y2];\
962
            ((uint8_t*)dest)[5]= b[Y2];\
963
            dest+=6;\
964
        }\
965
        break;\
966
    case PIX_FMT_BGR24:\
967
        func(uint8_t,0)\
968
            ((uint8_t*)dest)[0]= b[Y1];\
969
            ((uint8_t*)dest)[1]= g[Y1];\
970
            ((uint8_t*)dest)[2]= r[Y1];\
971
            ((uint8_t*)dest)[3]= b[Y2];\
972
            ((uint8_t*)dest)[4]= g[Y2];\
973
            ((uint8_t*)dest)[5]= r[Y2];\
974
            dest+=6;\
975
        }\
976
        break;\
977
    case PIX_FMT_RGB565:\
978
    case PIX_FMT_BGR565:\
979
        {\
980
            const int dr1= dither_2x2_8[y&1    ][0];\
981
            const int dg1= dither_2x2_4[y&1    ][0];\
982
            const int db1= dither_2x2_8[(y&1)^1][0];\
983
            const int dr2= dither_2x2_8[y&1    ][1];\
984
            const int dg2= dither_2x2_4[y&1    ][1];\
985
            const int db2= dither_2x2_8[(y&1)^1][1];\
986
            func(uint16_t,0)\
987
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
988
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
989
            }\
990
        }\
991
        break;\
992
    case PIX_FMT_RGB555:\
993
    case PIX_FMT_BGR555:\
994
        {\
995
            const int dr1= dither_2x2_8[y&1    ][0];\
996
            const int dg1= dither_2x2_8[y&1    ][1];\
997
            const int db1= dither_2x2_8[(y&1)^1][0];\
998
            const int dr2= dither_2x2_8[y&1    ][1];\
999
            const int dg2= dither_2x2_8[y&1    ][0];\
1000
            const int db2= dither_2x2_8[(y&1)^1][1];\
1001
            func(uint16_t,0)\
1002
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
1003
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
1004
            }\
1005
        }\
1006
        break;\
1007
    case PIX_FMT_RGB8:\
1008
    case PIX_FMT_BGR8:\
1009
        {\
1010
            const uint8_t * const d64= dither_8x8_73[y&7];\
1011
            const uint8_t * const d32= dither_8x8_32[y&7];\
1012
            func(uint8_t,0)\
1013
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
1014
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
1015
            }\
1016
        }\
1017
        break;\
1018
    case PIX_FMT_RGB4:\
1019
    case PIX_FMT_BGR4:\
1020
        {\
1021
            const uint8_t * const d64= dither_8x8_73 [y&7];\
1022
            const uint8_t * const d128=dither_8x8_220[y&7];\
1023
            func(uint8_t,0)\
1024
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
1025
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
1026
            }\
1027
        }\
1028
        break;\
1029
    case PIX_FMT_RGB4_BYTE:\
1030
    case PIX_FMT_BGR4_BYTE:\
1031
        {\
1032
            const uint8_t * const d64= dither_8x8_73 [y&7];\
1033
            const uint8_t * const d128=dither_8x8_220[y&7];\
1034
            func(uint8_t,0)\
1035
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
1036
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
1037
            }\
1038
        }\
1039
        break;\
1040
    case PIX_FMT_MONOBLACK:\
1041
    case PIX_FMT_MONOWHITE:\
1042
        {\
1043
            func_monoblack\
1044
        }\
1045
        break;\
1046
    case PIX_FMT_YUYV422:\
1047
        func2\
1048
            ((uint8_t*)dest)[2*i2+0]= Y1;\
1049
            ((uint8_t*)dest)[2*i2+1]= U;\
1050
            ((uint8_t*)dest)[2*i2+2]= Y2;\
1051
            ((uint8_t*)dest)[2*i2+3]= V;\
1052
        }                \
1053
        break;\
1054
    case PIX_FMT_UYVY422:\
1055
        func2\
1056
            ((uint8_t*)dest)[2*i2+0]= U;\
1057
            ((uint8_t*)dest)[2*i2+1]= Y1;\
1058
            ((uint8_t*)dest)[2*i2+2]= V;\
1059
            ((uint8_t*)dest)[2*i2+3]= Y2;\
1060
        }                \
1061
        break;\
1062
    case PIX_FMT_GRAY16BE:\
1063
        func_g16\
1064
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
1065
            ((uint8_t*)dest)[2*i2+1]= Y1;\
1066
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
1067
            ((uint8_t*)dest)[2*i2+3]= Y2;\
1068
        }                \
1069
        break;\
1070
    case PIX_FMT_GRAY16LE:\
1071
        func_g16\
1072
            ((uint8_t*)dest)[2*i2+0]= Y1;\
1073
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
1074
            ((uint8_t*)dest)[2*i2+2]= Y2;\
1075
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
1076
        }                \
1077
        break;\
1078
    }\
1079

    
1080

    
1081
static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1082
                                  const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1083
                                  const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1084
{
1085
    int i;
1086
    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)
1087
}
1088

    
1089
static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1090
                                    const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1091
                                    const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1092
{
1093
    int i;
1094
    int step= fmt_depth(c->dstFormat)/8;
1095
    int aidx= 3;
1096

    
1097
    switch(c->dstFormat) {
1098
    case PIX_FMT_ARGB:
1099
        dest++;
1100
        aidx= 0;
1101
    case PIX_FMT_RGB24:
1102
        aidx--;
1103
    case PIX_FMT_RGBA:
1104
        if (CONFIG_SMALL) {
1105
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1106
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1107
                dest[aidx]= needAlpha ? A : 255;
1108
                dest[0]= R>>22;
1109
                dest[1]= G>>22;
1110
                dest[2]= B>>22;
1111
                dest+= step;
1112
            }
1113
        } else {
1114
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1115
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1116
                    dest[aidx]= A;
1117
                    dest[0]= R>>22;
1118
                    dest[1]= G>>22;
1119
                    dest[2]= B>>22;
1120
                    dest+= step;
1121
                }
1122
            } else {
1123
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1124
                    dest[aidx]= 255;
1125
                    dest[0]= R>>22;
1126
                    dest[1]= G>>22;
1127
                    dest[2]= B>>22;
1128
                    dest+= step;
1129
                }
1130
            }
1131
        }
1132
        break;
1133
    case PIX_FMT_ABGR:
1134
        dest++;
1135
        aidx= 0;
1136
    case PIX_FMT_BGR24:
1137
        aidx--;
1138
    case PIX_FMT_BGRA:
1139
        if (CONFIG_SMALL) {
1140
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1141
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1142
                dest[aidx]= needAlpha ? A : 255;
1143
                dest[0]= B>>22;
1144
                dest[1]= G>>22;
1145
                dest[2]= R>>22;
1146
                dest+= step;
1147
            }
1148
        } else {
1149
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1150
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1151
                    dest[aidx]= A;
1152
                    dest[0]= B>>22;
1153
                    dest[1]= G>>22;
1154
                    dest[2]= R>>22;
1155
                    dest+= step;
1156
                }
1157
            } else {
1158
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1159
                    dest[aidx]= 255;
1160
                    dest[0]= B>>22;
1161
                    dest[1]= G>>22;
1162
                    dest[2]= R>>22;
1163
                    dest+= step;
1164
                }
1165
            }
1166
        }
1167
        break;
1168
    default:
1169
        assert(0);
1170
    }
1171
}
1172

    
1173
static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
1174
{
1175
    int i;
1176
    uint8_t *ptr = plane + stride*y;
1177
    for (i=0; i<height; i++) {
1178
        memset(ptr, val, width);
1179
        ptr += stride;
1180
    }
1181
}
1182

    
1183
static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)
1184
{
1185
    int i;
1186
    for (i = 0; i < width; i++) {
1187
        int r = src[i*6+0];
1188
        int g = src[i*6+2];
1189
        int b = src[i*6+4];
1190

    
1191
        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1192
    }
1193
}
1194

    
1195
static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1196
                             uint8_t *src1, uint8_t *src2, int width)
1197
{
1198
    int i;
1199
    assert(src1==src2);
1200
    for (i = 0; i < width; i++) {
1201
        int r = src1[6*i + 0];
1202
        int g = src1[6*i + 2];
1203
        int b = src1[6*i + 4];
1204

    
1205
        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1206
        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1207
    }
1208
}
1209

    
1210
static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1211
                                  uint8_t *src1, uint8_t *src2, int width)
1212
{
1213
    int i;
1214
    assert(src1==src2);
1215
    for (i = 0; i < width; i++) {
1216
        int r= src1[12*i + 0] + src1[12*i + 6];
1217
        int g= src1[12*i + 2] + src1[12*i + 8];
1218
        int b= src1[12*i + 4] + src1[12*i + 10];
1219

    
1220
        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1221
        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1222
    }
1223
}
1224

    
1225
#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1226
static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1227
{\
1228
    int i;\
1229
    for (i=0; i<width; i++) {\
1230
        int b= (((const type*)src)[i]>>shb)&maskb;\
1231
        int g= (((const type*)src)[i]>>shg)&maskg;\
1232
        int r= (((const type*)src)[i]>>shr)&maskr;\
1233
\
1234
        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1235
    }\
1236
}
1237

    
1238
BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1239
BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
1240
BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
1241
BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
1242
BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1243
BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1244

    
1245
static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1246
{
1247
    int i;
1248
    for (i=0; i<width; i++) {
1249
        dst[i]= src[4*i];
1250
    }
1251
}
1252

    
1253
#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1254
static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1255
{\
1256
    int i;\
1257
    for (i=0; i<width; i++) {\
1258
        int b= (((const type*)src)[i]&maskb)>>shb;\
1259
        int g= (((const type*)src)[i]&maskg)>>shg;\
1260
        int r= (((const type*)src)[i]&maskr)>>shr;\
1261
\
1262
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1263
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1264
    }\
1265
}\
1266
static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1267
{\
1268
    int i;\
1269
    for (i=0; i<width; i++) {\
1270
        int pix0= ((const type*)src)[2*i+0];\
1271
        int pix1= ((const type*)src)[2*i+1];\
1272
        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1273
        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1274
        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1275
        g&= maskg|(2*maskg);\
1276
\
1277
        g>>=shg;\
1278
\
1279
        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1280
        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1281
    }\
1282
}
1283

    
1284
BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1285
BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
1286
BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0,          0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
1287
BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0,          0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
1288
BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0,          0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1289
BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0,          0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1290

    
1291
static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1292
{
1293
    int i;
1294
    for (i=0; i<width; i++) {
1295
        int d= src[i];
1296

    
1297
        dst[i]= pal[d] & 0xFF;
1298
    }
1299
}
1300

    
1301
static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1302
                           const uint8_t *src1, const uint8_t *src2,
1303
                           long width, uint32_t *pal)
1304
{
1305
    int i;
1306
    assert(src1 == src2);
1307
    for (i=0; i<width; i++) {
1308
        int p= pal[src1[i]];
1309

    
1310
        dstU[i]= p>>8;
1311
        dstV[i]= p>>16;
1312
    }
1313
}
1314

    
1315
static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1316
{
1317
    int i, j;
1318
    for (i=0; i<width/8; i++) {
1319
        int d= ~src[i];
1320
        for(j=0; j<8; j++)
1321
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1322
    }
1323
}
1324

    
1325
static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1326
{
1327
    int i, j;
1328
    for (i=0; i<width/8; i++) {
1329
        int d= src[i];
1330
        for(j=0; j<8; j++)
1331
            dst[8*i+j]= ((d>>(7-j))&1)*255;
1332
    }
1333
}
1334

    
1335

    
1336
//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1337
//Plain C versions
1338
#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1339
#define COMPILE_C
1340
#endif
1341

    
1342
#if ARCH_PPC
1343
#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1344
#define COMPILE_ALTIVEC
1345
#endif
1346
#endif //ARCH_PPC
1347

    
1348
#if ARCH_X86
1349

    
1350
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1351
#define COMPILE_MMX
1352
#endif
1353

    
1354
#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1355
#define COMPILE_MMX2
1356
#endif
1357

    
1358
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1359
#define COMPILE_3DNOW
1360
#endif
1361
#endif //ARCH_X86
1362

    
1363
#define COMPILE_TEMPLATE_MMX 0
1364
#define COMPILE_TEMPLATE_MMX2 0
1365
#define COMPILE_TEMPLATE_AMD3DNOW 0
1366
#define COMPILE_TEMPLATE_ALTIVEC 0
1367

    
1368
#ifdef COMPILE_C
1369
#define RENAME(a) a ## _C
1370
#include "swscale_template.c"
1371
#endif
1372

    
1373
#ifdef COMPILE_ALTIVEC
1374
#undef RENAME
1375
#undef COMPILE_TEMPLATE_ALTIVEC
1376
#define COMPILE_TEMPLATE_ALTIVEC 1
1377
#define RENAME(a) a ## _altivec
1378
#include "swscale_template.c"
1379
#endif
1380

    
1381
#if ARCH_X86
1382

    
1383
//MMX versions
1384
#ifdef COMPILE_MMX
1385
#undef RENAME
1386
#undef COMPILE_TEMPLATE_MMX
1387
#undef COMPILE_TEMPLATE_MMX2
1388
#undef COMPILE_TEMPLATE_AMD3DNOW
1389
#define COMPILE_TEMPLATE_MMX 1
1390
#define COMPILE_TEMPLATE_MMX2 0
1391
#define COMPILE_TEMPLATE_AMD3DNOW 0
1392
#define RENAME(a) a ## _MMX
1393
#include "swscale_template.c"
1394
#endif
1395

    
1396
//MMX2 versions
1397
#ifdef COMPILE_MMX2
1398
#undef RENAME
1399
#undef COMPILE_TEMPLATE_MMX
1400
#undef COMPILE_TEMPLATE_MMX2
1401
#undef COMPILE_TEMPLATE_AMD3DNOW
1402
#define COMPILE_TEMPLATE_MMX 1
1403
#define COMPILE_TEMPLATE_MMX2 1
1404
#define COMPILE_TEMPLATE_AMD3DNOW 0
1405
#define RENAME(a) a ## _MMX2
1406
#include "swscale_template.c"
1407
#endif
1408

    
1409
//3DNOW versions
1410
#ifdef COMPILE_3DNOW
1411
#undef RENAME
1412
#undef COMPILE_TEMPLATE_MMX
1413
#undef COMPILE_TEMPLATE_MMX2
1414
#undef COMPILE_TEMPLATE_AMD3DNOW
1415
#define COMPILE_TEMPLATE_MMX 1
1416
#define COMPILE_TEMPLATE_MMX2 0
1417
#define COMPILE_TEMPLATE_AMD3DNOW 1
1418
#define RENAME(a) a ## _3DNow
1419
#include "swscale_template.c"
1420
#endif
1421

    
1422
#endif //ARCH_X86
1423

    
1424
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1425
{
1426
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1427
    if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
1428
    else           return getSplineCoeff(        0.0,
1429
                                          b+ 2.0*c + 3.0*d,
1430
                                                 c + 3.0*d,
1431
                                         -b- 3.0*c - 6.0*d,
1432
                                         dist-1.0);
1433
}
1434

    
1435
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1436
                             int srcW, int dstW, int filterAlign, int one, int flags,
1437
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1438
{
1439
    int i;
1440
    int filterSize;
1441
    int filter2Size;
1442
    int minFilterSize;
1443
    int64_t *filter=NULL;
1444
    int64_t *filter2=NULL;
1445
    const int64_t fone= 1LL<<54;
1446
    int ret= -1;
1447
#if ARCH_X86
1448
    if (flags & SWS_CPU_CAPS_MMX)
1449
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1450
#endif
1451

    
1452
    // NOTE: the +1 is for the MMX scaler which reads over the end
1453
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1454

    
1455
    if (FFABS(xInc - 0x10000) <10) { // unscaled
1456
        int i;
1457
        filterSize= 1;
1458
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1459

    
1460
        for (i=0; i<dstW; i++) {
1461
            filter[i*filterSize]= fone;
1462
            (*filterPos)[i]=i;
1463
        }
1464

    
1465
    } else if (flags&SWS_POINT) { // lame looking point sampling mode
1466
        int i;
1467
        int xDstInSrc;
1468
        filterSize= 1;
1469
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1470

    
1471
        xDstInSrc= xInc/2 - 0x8000;
1472
        for (i=0; i<dstW; i++) {
1473
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1474

    
1475
            (*filterPos)[i]= xx;
1476
            filter[i]= fone;
1477
            xDstInSrc+= xInc;
1478
        }
1479
    } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
1480
        int i;
1481
        int xDstInSrc;
1482
        filterSize= 2;
1483
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1484

    
1485
        xDstInSrc= xInc/2 - 0x8000;
1486
        for (i=0; i<dstW; i++) {
1487
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1488
            int j;
1489

    
1490
            (*filterPos)[i]= xx;
1491
            //bilinear upscale / linear interpolate / area averaging
1492
            for (j=0; j<filterSize; j++) {
1493
                int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1494
                if (coeff<0) coeff=0;
1495
                filter[i*filterSize + j]= coeff;
1496
                xx++;
1497
            }
1498
            xDstInSrc+= xInc;
1499
        }
1500
    } else {
1501
        int xDstInSrc;
1502
        int sizeFactor;
1503

    
1504
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1505
        else if (flags&SWS_X)            sizeFactor=  8;
1506
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1507
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1508
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1509
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1510
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1511
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1512
        else {
1513
            sizeFactor= 0; //GCC warning killer
1514
            assert(0);
1515
        }
1516

    
1517
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1518
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1519

    
1520
        if (filterSize > srcW-2) filterSize=srcW-2;
1521

    
1522
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1523

    
1524
        xDstInSrc= xInc - 0x10000;
1525
        for (i=0; i<dstW; i++) {
1526
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1527
            int j;
1528
            (*filterPos)[i]= xx;
1529
            for (j=0; j<filterSize; j++) {
1530
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1531
                double floatd;
1532
                int64_t coeff;
1533

    
1534
                if (xInc > 1<<16)
1535
                    d= d*dstW/srcW;
1536
                floatd= d * (1.0/(1<<30));
1537

    
1538
                if (flags & SWS_BICUBIC) {
1539
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1540
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1541
                    int64_t dd = ( d*d)>>30;
1542
                    int64_t ddd= (dd*d)>>30;
1543

    
1544
                    if      (d < 1LL<<30)
1545
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1546
                    else if (d < 1LL<<31)
1547
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1548
                    else
1549
                        coeff=0.0;
1550
                    coeff *= fone>>(30+24);
1551
                }
1552
/*                else if (flags & SWS_X) {
1553
                    double p= param ? param*0.01 : 0.3;
1554
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1555
                    coeff*= pow(2.0, - p*d*d);
1556
                }*/
1557
                else if (flags & SWS_X) {
1558
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1559
                    double c;
1560

    
1561
                    if (floatd<1.0)
1562
                        c = cos(floatd*PI);
1563
                    else
1564
                        c=-1.0;
1565
                    if (c<0.0)      c= -pow(-c, A);
1566
                    else            c=  pow( c, A);
1567
                    coeff= (c*0.5 + 0.5)*fone;
1568
                } else if (flags & SWS_AREA) {
1569
                    int64_t d2= d - (1<<29);
1570
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1571
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1572
                    else coeff=0.0;
1573
                    coeff *= fone>>(30+16);
1574
                } else if (flags & SWS_GAUSS) {
1575
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1576
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1577
                } else if (flags & SWS_SINC) {
1578
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1579
                } else if (flags & SWS_LANCZOS) {
1580
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1581
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1582
                    if (floatd>p) coeff=0;
1583
                } else if (flags & SWS_BILINEAR) {
1584
                    coeff= (1<<30) - d;
1585
                    if (coeff<0) coeff=0;
1586
                    coeff *= fone >> 30;
1587
                } else if (flags & SWS_SPLINE) {
1588
                    double p=-2.196152422706632;
1589
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1590
                } else {
1591
                    coeff= 0.0; //GCC warning killer
1592
                    assert(0);
1593
                }
1594

    
1595
                filter[i*filterSize + j]= coeff;
1596
                xx++;
1597
            }
1598
            xDstInSrc+= 2*xInc;
1599
        }
1600
    }
1601

    
1602
    /* apply src & dst Filter to filter -> filter2
1603
       av_free(filter);
1604
    */
1605
    assert(filterSize>0);
1606
    filter2Size= filterSize;
1607
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1608
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1609
    assert(filter2Size>0);
1610
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1611

    
1612
    for (i=0; i<dstW; i++) {
1613
        int j, k;
1614

    
1615
        if(srcFilter) {
1616
            for (k=0; k<srcFilter->length; k++) {
1617
                for (j=0; j<filterSize; j++)
1618
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1619
            }
1620
        } else {
1621
            for (j=0; j<filterSize; j++)
1622
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1623
        }
1624
        //FIXME dstFilter
1625

    
1626
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1627
    }
1628
    av_freep(&filter);
1629

    
1630
    /* try to reduce the filter-size (step1 find size and shift left) */
1631
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1632
    minFilterSize= 0;
1633
    for (i=dstW-1; i>=0; i--) {
1634
        int min= filter2Size;
1635
        int j;
1636
        int64_t cutOff=0.0;
1637

    
1638
        /* get rid off near zero elements on the left by shifting left */
1639
        for (j=0; j<filter2Size; j++) {
1640
            int k;
1641
            cutOff += FFABS(filter2[i*filter2Size]);
1642

    
1643
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1644

    
1645
            /* preserve monotonicity because the core can't handle the filter otherwise */
1646
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1647

    
1648
            // move filter coefficients left
1649
            for (k=1; k<filter2Size; k++)
1650
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1651
            filter2[i*filter2Size + k - 1]= 0;
1652
            (*filterPos)[i]++;
1653
        }
1654

    
1655
        cutOff=0;
1656
        /* count near zeros on the right */
1657
        for (j=filter2Size-1; j>0; j--) {
1658
            cutOff += FFABS(filter2[i*filter2Size + j]);
1659

    
1660
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1661
            min--;
1662
        }
1663

    
1664
        if (min>minFilterSize) minFilterSize= min;
1665
    }
1666

    
1667
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1668
        // we can handle the special case 4,
1669
        // so we don't want to go to the full 8
1670
        if (minFilterSize < 5)
1671
            filterAlign = 4;
1672

    
1673
        // We really don't want to waste our time
1674
        // doing useless computation, so fall back on
1675
        // the scalar C code for very small filters.
1676
        // Vectorizing is worth it only if you have a
1677
        // decent-sized vector.
1678
        if (minFilterSize < 3)
1679
            filterAlign = 1;
1680
    }
1681

    
1682
    if (flags & SWS_CPU_CAPS_MMX) {
1683
        // special case for unscaled vertical filtering
1684
        if (minFilterSize == 1 && filterAlign == 2)
1685
            filterAlign= 1;
1686
    }
1687

    
1688
    assert(minFilterSize > 0);
1689
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1690
    assert(filterSize > 0);
1691
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1692
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1693
        goto error;
1694
    *outFilterSize= filterSize;
1695

    
1696
    if (flags&SWS_PRINT_INFO)
1697
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1698
    /* try to reduce the filter-size (step2 reduce it) */
1699
    for (i=0; i<dstW; i++) {
1700
        int j;
1701

    
1702
        for (j=0; j<filterSize; j++) {
1703
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1704
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1705
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1706
                filter[i*filterSize + j]= 0;
1707
        }
1708
    }
1709

    
1710

    
1711
    //FIXME try to align filterPos if possible
1712

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

    
1726
        if ((*filterPos)[i] + filterSize > srcW) {
1727
            int shift= (*filterPos)[i] + filterSize - srcW;
1728
            // move filter coefficients right to compensate for filterPos
1729
            for (j=filterSize-2; j>=0; j--) {
1730
                int right= FFMIN(j + shift, filterSize-1);
1731
                filter[i*filterSize +right] += filter[i*filterSize +j];
1732
                filter[i*filterSize +j]=0;
1733
            }
1734
            (*filterPos)[i]= srcW - filterSize;
1735
        }
1736
    }
1737

    
1738
    // Note the +1 is for the MMX scaler which reads over the end
1739
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1740
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1741

    
1742
    /* normalize & store in outFilter */
1743
    for (i=0; i<dstW; i++) {
1744
        int j;
1745
        int64_t error=0;
1746
        int64_t sum=0;
1747

    
1748
        for (j=0; j<filterSize; j++) {
1749
            sum+= filter[i*filterSize + j];
1750
        }
1751
        sum= (sum + one/2)/ one;
1752
        for (j=0; j<*outFilterSize; j++) {
1753
            int64_t v= filter[i*filterSize + j] + error;
1754
            int intV= ROUNDED_DIV(v, sum);
1755
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1756
            error= v - intV*sum;
1757
        }
1758
    }
1759

    
1760
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1761
    for (i=0; i<*outFilterSize; i++) {
1762
        int j= dstW*(*outFilterSize);
1763
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1764
    }
1765

    
1766
    ret=0;
1767
error:
1768
    av_free(filter);
1769
    av_free(filter2);
1770
    return ret;
1771
}
1772

    
1773
#ifdef COMPILE_MMX2
1774
static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
1775
{
1776
    uint8_t *fragmentA;
1777
    x86_reg imm8OfPShufW1A;
1778
    x86_reg imm8OfPShufW2A;
1779
    x86_reg fragmentLengthA;
1780
    uint8_t *fragmentB;
1781
    x86_reg imm8OfPShufW1B;
1782
    x86_reg imm8OfPShufW2B;
1783
    x86_reg fragmentLengthB;
1784
    int fragmentPos;
1785

    
1786
    int xpos, i;
1787

    
1788
    // create an optimized horizontal scaling routine
1789

    
1790
    //code fragment
1791

    
1792
    __asm__ volatile(
1793
        "jmp                         9f                 \n\t"
1794
    // Begin
1795
        "0:                                             \n\t"
1796
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1797
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1798
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1799
        "punpcklbw                %%mm7, %%mm1          \n\t"
1800
        "punpcklbw                %%mm7, %%mm0          \n\t"
1801
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1802
        "1:                                             \n\t"
1803
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1804
        "2:                                             \n\t"
1805
        "psubw                    %%mm1, %%mm0          \n\t"
1806
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1807
        "pmullw                   %%mm3, %%mm0          \n\t"
1808
        "psllw                       $7, %%mm1          \n\t"
1809
        "paddw                    %%mm1, %%mm0          \n\t"
1810

    
1811
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1812

    
1813
        "add                         $8, %%"REG_a"      \n\t"
1814
    // End
1815
        "9:                                             \n\t"
1816
//        "int $3                                         \n\t"
1817
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1818
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1819
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1820
        "dec                         %1                 \n\t"
1821
        "dec                         %2                 \n\t"
1822
        "sub                         %0, %1             \n\t"
1823
        "sub                         %0, %2             \n\t"
1824
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1825
        "sub                         %0, %3             \n\t"
1826

    
1827

    
1828
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1829
        "=r" (fragmentLengthA)
1830
    );
1831

    
1832
    __asm__ volatile(
1833
        "jmp                         9f                 \n\t"
1834
    // Begin
1835
        "0:                                             \n\t"
1836
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1837
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1838
        "punpcklbw                %%mm7, %%mm0          \n\t"
1839
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1840
        "1:                                             \n\t"
1841
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1842
        "2:                                             \n\t"
1843
        "psubw                    %%mm1, %%mm0          \n\t"
1844
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1845
        "pmullw                   %%mm3, %%mm0          \n\t"
1846
        "psllw                       $7, %%mm1          \n\t"
1847
        "paddw                    %%mm1, %%mm0          \n\t"
1848

    
1849
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1850

    
1851
        "add                         $8, %%"REG_a"      \n\t"
1852
    // End
1853
        "9:                                             \n\t"
1854
//        "int                       $3                   \n\t"
1855
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1856
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1857
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1858
        "dec                         %1                 \n\t"
1859
        "dec                         %2                 \n\t"
1860
        "sub                         %0, %1             \n\t"
1861
        "sub                         %0, %2             \n\t"
1862
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1863
        "sub                         %0, %3             \n\t"
1864

    
1865

    
1866
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1867
        "=r" (fragmentLengthB)
1868
    );
1869

    
1870
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1871
    fragmentPos=0;
1872

    
1873
    for (i=0; i<dstW/numSplits; i++) {
1874
        int xx=xpos>>16;
1875

    
1876
        if ((i&3) == 0) {
1877
            int a=0;
1878
            int b=((xpos+xInc)>>16) - xx;
1879
            int c=((xpos+xInc*2)>>16) - xx;
1880
            int d=((xpos+xInc*3)>>16) - xx;
1881
            int inc                = (d+1<4);
1882
            uint8_t *fragment      = (d+1<4) ? fragmentB       : fragmentA;
1883
            x86_reg imm8OfPShufW1  = (d+1<4) ? imm8OfPShufW1B  : imm8OfPShufW1A;
1884
            x86_reg imm8OfPShufW2  = (d+1<4) ? imm8OfPShufW2B  : imm8OfPShufW2A;
1885
            x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
1886
            int maxShift= 3-(d+inc);
1887
            int shift=0;
1888

    
1889
            if (filterCode) {
1890
                filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1891
                filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1892
                filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1893
                filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1894
                filterPos[i/2]= xx;
1895

    
1896
                memcpy(filterCode + fragmentPos, fragment, fragmentLength);
1897

    
1898
                filterCode[fragmentPos + imm8OfPShufW1]=
1899
                    (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
1900
                filterCode[fragmentPos + imm8OfPShufW2]=
1901
                    a | (b<<2) | (c<<4) | (d<<6);
1902

    
1903
                if (i+4-inc>=dstW) shift=maxShift; //avoid overread
1904
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1905

    
1906
                if (shift && i>=shift) {
1907
                    filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
1908
                    filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
1909
                    filterPos[i/2]-=shift;
1910
                }
1911
            }
1912

    
1913
            fragmentPos+= fragmentLength;
1914

    
1915
            if (filterCode)
1916
                filterCode[fragmentPos]= RET;
1917
        }
1918
        xpos+=xInc;
1919
    }
1920
    if (filterCode)
1921
        filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1922

    
1923
    return fragmentPos + 1;
1924
}
1925
#endif /* COMPILE_MMX2 */
1926

    
1927
static void globalInit(void)
1928
{
1929
    // generating tables:
1930
    int i;
1931
    for (i=0; i<768; i++) {
1932
        int c= av_clip_uint8(i-256);
1933
        clip_table[i]=c;
1934
    }
1935
}
1936

    
1937
static SwsFunc getSwsFunc(SwsContext *c)
1938
{
1939
#if CONFIG_RUNTIME_CPUDETECT
1940
    int flags = c->flags;
1941

    
1942
#if ARCH_X86 && CONFIG_GPL
1943
    // ordered per speed fastest first
1944
    if (flags & SWS_CPU_CAPS_MMX2) {
1945
        sws_init_swScale_MMX2(c);
1946
        return swScale_MMX2;
1947
    } else if (flags & SWS_CPU_CAPS_3DNOW) {
1948
        sws_init_swScale_3DNow(c);
1949
        return swScale_3DNow;
1950
    } else if (flags & SWS_CPU_CAPS_MMX) {
1951
        sws_init_swScale_MMX(c);
1952
        return swScale_MMX;
1953
    } else {
1954
        sws_init_swScale_C(c);
1955
        return swScale_C;
1956
    }
1957

    
1958
#else
1959
#if ARCH_PPC
1960
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1961
        sws_init_swScale_altivec(c);
1962
        return swScale_altivec;
1963
    } else {
1964
        sws_init_swScale_C(c);
1965
        return swScale_C;
1966
    }
1967
#endif
1968
    sws_init_swScale_C(c);
1969
    return swScale_C;
1970
#endif /* ARCH_X86 && CONFIG_GPL */
1971
#else //CONFIG_RUNTIME_CPUDETECT
1972
#if   COMPILE_TEMPLATE_MMX2
1973
    sws_init_swScale_MMX2(c);
1974
    return swScale_MMX2;
1975
#elif COMPILE_TEMPLATE_AMD3DNOW
1976
    sws_init_swScale_3DNow(c);
1977
    return swScale_3DNow;
1978
#elif COMPILE_TEMPLATE_MMX
1979
    sws_init_swScale_MMX(c);
1980
    return swScale_MMX;
1981
#elif COMPILE_TEMPLATE_ALTIVEC
1982
    sws_init_swScale_altivec(c);
1983
    return swScale_altivec;
1984
#else
1985
    sws_init_swScale_C(c);
1986
    return swScale_C;
1987
#endif
1988
#endif //!CONFIG_RUNTIME_CPUDETECT
1989
}
1990

    
1991
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1992
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
1993
{
1994
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1995
    /* Copy Y plane */
1996
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1997
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1998
    else {
1999
        int i;
2000
        const uint8_t *srcPtr= src[0];
2001
        uint8_t *dstPtr= dst;
2002
        for (i=0; i<srcSliceH; i++) {
2003
            memcpy(dstPtr, srcPtr, c->srcW);
2004
            srcPtr+= srcStride[0];
2005
            dstPtr+= dstStride[0];
2006
        }
2007
    }
2008
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
2009
    if (c->dstFormat == PIX_FMT_NV12)
2010
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
2011
    else
2012
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
2013

    
2014
    return srcSliceH;
2015
}
2016

    
2017
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2018
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2019
{
2020
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2021

    
2022
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2023

    
2024
    return srcSliceH;
2025
}
2026

    
2027
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2028
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2029
{
2030
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2031

    
2032
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
2033

    
2034
    return srcSliceH;
2035
}
2036

    
2037
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2038
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
2039
{
2040
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2041

    
2042
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2043

    
2044
    return srcSliceH;
2045
}
2046

    
2047
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2048
                                int srcSliceH, uint8_t* dstParam[], int dstStride[])
2049
{
2050
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2051

    
2052
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2053

    
2054
    return srcSliceH;
2055
}
2056

    
2057
static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2058
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2059
{
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
{
2075
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2076
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2077
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2078

    
2079
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2080

    
2081
    return srcSliceH;
2082
}
2083

    
2084
static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2085
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2086
{
2087
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2088
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2089
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2090

    
2091
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2092

    
2093
    if (dstParam[3])
2094
        fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2095

    
2096
    return srcSliceH;
2097
}
2098

    
2099
static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2100
                               int srcSliceH, uint8_t* dstParam[], int dstStride[])
2101
{
2102
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2103
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2104
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
2105

    
2106
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2107

    
2108
    return srcSliceH;
2109
}
2110

    
2111
static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2112
                          int srcSliceH, uint8_t* dst[], int dstStride[])
2113
{
2114
    const enum PixelFormat srcFormat= c->srcFormat;
2115
    const enum PixelFormat dstFormat= c->dstFormat;
2116
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
2117
                 const uint8_t *palette)=NULL;
2118
    int i;
2119
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2120
    uint8_t *srcPtr= src[0];
2121

    
2122
    if (!usePal(srcFormat))
2123
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2124
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2125

    
2126
    switch(dstFormat) {
2127
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
2128
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
2129
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
2130
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
2131
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
2132
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
2133
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2134
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2135
    }
2136

    
2137

    
2138
    for (i=0; i<srcSliceH; i++) {
2139
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
2140
        srcPtr+= srcStride[0];
2141
        dstPtr+= dstStride[0];
2142
    }
2143

    
2144
    return srcSliceH;
2145
}
2146

    
2147
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
2148
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2149
                          int srcSliceH, uint8_t* dst[], int dstStride[])
2150
{
2151
    const enum PixelFormat srcFormat= c->srcFormat;
2152
    const enum PixelFormat dstFormat= c->dstFormat;
2153
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2154
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2155
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2156
    const int dstId= fmt_depth(dstFormat) >> 2;
2157
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2158

    
2159
    /* BGR -> BGR */
2160
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
2161
       || (isRGB(srcFormat) && isRGB(dstFormat))) {
2162
        switch(srcId | (dstId<<4)) {
2163
        case 0x34: conv= rgb16to15; break;
2164
        case 0x36: conv= rgb24to15; break;
2165
        case 0x38: conv= rgb32to15; break;
2166
        case 0x43: conv= rgb15to16; break;
2167
        case 0x46: conv= rgb24to16; break;
2168
        case 0x48: conv= rgb32to16; break;
2169
        case 0x63: conv= rgb15to24; break;
2170
        case 0x64: conv= rgb16to24; break;
2171
        case 0x68: conv= rgb32to24; break;
2172
        case 0x83: conv= rgb15to32; break;
2173
        case 0x84: conv= rgb16to32; break;
2174
        case 0x86: conv= rgb24to32; break;
2175
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2176
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2177
        }
2178
    } else if (  (isBGR(srcFormat) && isRGB(dstFormat))
2179
             || (isRGB(srcFormat) && isBGR(dstFormat))) {
2180
        switch(srcId | (dstId<<4)) {
2181
        case 0x33: conv= rgb15tobgr15; break;
2182
        case 0x34: conv= rgb16tobgr15; break;
2183
        case 0x36: conv= rgb24tobgr15; break;
2184
        case 0x38: conv= rgb32tobgr15; break;
2185
        case 0x43: conv= rgb15tobgr16; break;
2186
        case 0x44: conv= rgb16tobgr16; break;
2187
        case 0x46: conv= rgb24tobgr16; break;
2188
        case 0x48: conv= rgb32tobgr16; break;
2189
        case 0x63: conv= rgb15tobgr24; break;
2190
        case 0x64: conv= rgb16tobgr24; break;
2191
        case 0x66: conv= rgb24tobgr24; break;
2192
        case 0x68: conv= rgb32tobgr24; break;
2193
        case 0x83: conv= rgb15tobgr32; break;
2194
        case 0x84: conv= rgb16tobgr32; break;
2195
        case 0x86: conv= rgb24tobgr32; break;
2196
        case 0x88: conv= rgb32tobgr32; break;
2197
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2198
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2199
        }
2200
    } else {
2201
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2202
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2203
    }
2204

    
2205
    if(conv) {
2206
        uint8_t *srcPtr= src[0];
2207
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2208
            srcPtr += ALT32_CORR;
2209

    
2210
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2211
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2212
        else {
2213
            int i;
2214
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2215

    
2216
            for (i=0; i<srcSliceH; i++) {
2217
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
2218
                srcPtr+= srcStride[0];
2219
                dstPtr+= dstStride[0];
2220
            }
2221
        }
2222
    }
2223
    return srcSliceH;
2224
}
2225

    
2226
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2227
                              int srcSliceH, uint8_t* dst[], int dstStride[])
2228
{
2229

    
2230
    rgb24toyv12(
2231
        src[0],
2232
        dst[0]+ srcSliceY    *dstStride[0],
2233
        dst[1]+(srcSliceY>>1)*dstStride[1],
2234
        dst[2]+(srcSliceY>>1)*dstStride[2],
2235
        c->srcW, srcSliceH,
2236
        dstStride[0], dstStride[1], srcStride[0]);
2237
    if (dst[3])
2238
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2239
    return srcSliceH;
2240
}
2241

    
2242
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2243
                             int srcSliceH, uint8_t* dst[], int dstStride[])
2244
{
2245
    int i;
2246

    
2247
    /* copy Y */
2248
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2249
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2250
    else {
2251
        uint8_t *srcPtr= src[0];
2252
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2253

    
2254
        for (i=0; i<srcSliceH; i++) {
2255
            memcpy(dstPtr, srcPtr, c->srcW);
2256
            srcPtr+= srcStride[0];
2257
            dstPtr+= dstStride[0];
2258
        }
2259
    }
2260

    
2261
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
2262
        planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2263
                 srcSliceH >> 2, srcStride[1], dstStride[1]);
2264
        planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2265
                 srcSliceH >> 2, srcStride[2], dstStride[2]);
2266
    } else {
2267
        planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2268
                 srcSliceH >> 2, srcStride[1], dstStride[2]);
2269
        planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2270
                 srcSliceH >> 2, srcStride[2], dstStride[1]);
2271
    }
2272
    if (dst[3])
2273
        fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2274
    return srcSliceH;
2275
}
2276

    
2277
/* unscaled copy like stuff (assumes nearly identical formats) */
2278
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2279
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2280
{
2281
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2282
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2283
    else {
2284
        int i;
2285
        uint8_t *srcPtr= src[0];
2286
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2287
        int length=0;
2288

    
2289
        /* universal length finder */
2290
        while(length+c->srcW <= FFABS(dstStride[0])
2291
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2292
        assert(length!=0);
2293

    
2294
        for (i=0; i<srcSliceH; i++) {
2295
            memcpy(dstPtr, srcPtr, length);
2296
            srcPtr+= srcStride[0];
2297
            dstPtr+= dstStride[0];
2298
        }
2299
    }
2300
    return srcSliceH;
2301
}
2302

    
2303
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2304
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2305
{
2306
    int plane, i, j;
2307
    for (plane=0; plane<4; plane++) {
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
            if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
2323
                if (!isBE(c->srcFormat)) srcPtr++;
2324
                for (i=0; i<height; i++) {
2325
                    for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2326
                    srcPtr+= srcStride[plane];
2327
                    dstPtr+= dstStride[plane];
2328
                }
2329
            } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
2330
                for (i=0; i<height; i++) {
2331
                    for (j=0; j<length; j++) {
2332
                        dstPtr[ j<<1   ] = srcPtr[j];
2333
                        dstPtr[(j<<1)+1] = srcPtr[j];
2334
                    }
2335
                    srcPtr+= srcStride[plane];
2336
                    dstPtr+= dstStride[plane];
2337
                }
2338
            } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2339
                  && isBE(c->srcFormat) != isBE(c->dstFormat)) {
2340

    
2341
                for (i=0; i<height; i++) {
2342
                    for (j=0; j<length; j++)
2343
                        ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2344
                    srcPtr+= srcStride[plane];
2345
                    dstPtr+= dstStride[plane];
2346
                }
2347
            } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2348
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2349
            else {
2350
                if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2351
                    length*=2;
2352
                for (i=0; i<height; i++) {
2353
                    memcpy(dstPtr, srcPtr, length);
2354
                    srcPtr+= srcStride[plane];
2355
                    dstPtr+= dstStride[plane];
2356
                }
2357
            }
2358
        }
2359
    }
2360
    return srcSliceH;
2361
}
2362

    
2363

    
2364
static void getSubSampleFactors(int *h, int *v, int format)
2365
{
2366
    switch(format) {
2367
    case PIX_FMT_UYVY422:
2368
    case PIX_FMT_YUYV422:
2369
        *h=1;
2370
        *v=0;
2371
        break;
2372
    case PIX_FMT_YUV420P:
2373
    case PIX_FMT_YUV420PLE:
2374
    case PIX_FMT_YUV420PBE:
2375
    case PIX_FMT_YUVA420P:
2376
    case PIX_FMT_GRAY16BE:
2377
    case PIX_FMT_GRAY16LE:
2378
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2379
    case PIX_FMT_NV12:
2380
    case PIX_FMT_NV21:
2381
        *h=1;
2382
        *v=1;
2383
        break;
2384
    case PIX_FMT_YUV440P:
2385
        *h=0;
2386
        *v=1;
2387
        break;
2388
    case PIX_FMT_YUV410P:
2389
        *h=2;
2390
        *v=2;
2391
        break;
2392
    case PIX_FMT_YUV444P:
2393
    case PIX_FMT_YUV444PLE:
2394
    case PIX_FMT_YUV444PBE:
2395
        *h=0;
2396
        *v=0;
2397
        break;
2398
    case PIX_FMT_YUV422P:
2399
    case PIX_FMT_YUV422PLE:
2400
    case PIX_FMT_YUV422PBE:
2401
        *h=1;
2402
        *v=0;
2403
        break;
2404
    case PIX_FMT_YUV411P:
2405
        *h=2;
2406
        *v=0;
2407
        break;
2408
    default:
2409
        *h=0;
2410
        *v=0;
2411
        break;
2412
    }
2413
}
2414

    
2415
static uint16_t roundToInt16(int64_t f)
2416
{
2417
    int r= (f + (1<<15))>>16;
2418
         if (r<-0x7FFF) return 0x8000;
2419
    else if (r> 0x7FFF) return 0x7FFF;
2420
    else                return r;
2421
}
2422

    
2423
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)
2424
{
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
{
2489
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2490

    
2491
    *inv_table = c->srcColorspaceTable;
2492
    *table     = c->dstColorspaceTable;
2493
    *srcRange  = c->srcRange;
2494
    *dstRange  = c->dstRange;
2495
    *brightness= c->brightness;
2496
    *contrast  = c->contrast;
2497
    *saturation= c->saturation;
2498

    
2499
    return 0;
2500
}
2501

    
2502
static int handle_jpeg(enum PixelFormat *format)
2503
{
2504
    switch (*format) {
2505
    case PIX_FMT_YUVJ420P:
2506
        *format = PIX_FMT_YUV420P;
2507
        return 1;
2508
    case PIX_FMT_YUVJ422P:
2509
        *format = PIX_FMT_YUV422P;
2510
        return 1;
2511
    case PIX_FMT_YUVJ444P:
2512
        *format = PIX_FMT_YUV444P;
2513
        return 1;
2514
    case PIX_FMT_YUVJ440P:
2515
        *format = PIX_FMT_YUV440P;
2516
        return 1;
2517
    default:
2518
        return 0;
2519
    }
2520
}
2521

    
2522
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2523
                           SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2524
{
2525

    
2526
    SwsContext *c;
2527
    int i;
2528
    int usesVFilter, usesHFilter;
2529
    int unscaled, needsDither;
2530
    int srcRange, dstRange;
2531
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2532
#if ARCH_X86
2533
    if (flags & SWS_CPU_CAPS_MMX)
2534
        __asm__ volatile("emms\n\t"::: "memory");
2535
#endif
2536

    
2537
#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
2538
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2539
#if   COMPILE_TEMPLATE_MMX2
2540
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2541
#elif COMPILE_TEMPLATE_AMD3DNOW
2542
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2543
#elif COMPILE_TEMPLATE_MMX
2544
    flags |= SWS_CPU_CAPS_MMX;
2545
#elif COMPILE_TEMPLATE_ALTIVEC
2546
    flags |= SWS_CPU_CAPS_ALTIVEC;
2547
#elif ARCH_BFIN
2548
    flags |= SWS_CPU_CAPS_BFIN;
2549
#endif
2550
#endif /* CONFIG_RUNTIME_CPUDETECT */
2551
    if (clip_table[512] != 255) globalInit();
2552
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2553

    
2554
    unscaled = (srcW == dstW && srcH == dstH);
2555
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2556
        && (fmt_depth(dstFormat))<24
2557
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2558

    
2559
    srcRange = handle_jpeg(&srcFormat);
2560
    dstRange = handle_jpeg(&dstFormat);
2561

    
2562
    if (!isSupportedIn(srcFormat)) {
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
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2568
        return NULL;
2569
    }
2570

    
2571
    i= flags & ( SWS_POINT
2572
                |SWS_AREA
2573
                |SWS_BILINEAR
2574
                |SWS_FAST_BILINEAR
2575
                |SWS_BICUBIC
2576
                |SWS_X
2577
                |SWS_GAUSS
2578
                |SWS_LANCZOS
2579
                |SWS_SINC
2580
                |SWS_SPLINE
2581
                |SWS_BICUBLIN);
2582
    if(!i || (i & (i-1))) {
2583
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2584
        return NULL;
2585
    }
2586

    
2587
    /* sanity check */
2588
    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
2589
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2590
               srcW, srcH, dstW, dstH);
2591
        return NULL;
2592
    }
2593
    if(srcW > VOFW || dstW > VOFW) {
2594
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2595
        return NULL;
2596
    }
2597

    
2598
    if (!dstFilter) dstFilter= &dummyFilter;
2599
    if (!srcFilter) srcFilter= &dummyFilter;
2600

    
2601
    c= av_mallocz(sizeof(SwsContext));
2602

    
2603
    c->av_class = &sws_context_class;
2604
    c->srcW= srcW;
2605
    c->srcH= srcH;
2606
    c->dstW= dstW;
2607
    c->dstH= dstH;
2608
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2609
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2610
    c->flags= flags;
2611
    c->dstFormat= dstFormat;
2612
    c->srcFormat= srcFormat;
2613
    c->vRounder= 4* 0x0001000100010001ULL;
2614

    
2615
    usesHFilter= usesVFilter= 0;
2616
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2617
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2618
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2619
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2620
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2621
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2622
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2623
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2624

    
2625
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2626
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2627

    
2628
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2629
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2630

    
2631
    // drop some chroma lines if the user wants it
2632
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2633
    c->chrSrcVSubSample+= c->vChrDrop;
2634

    
2635
    // drop every other pixel for chroma calculation unless user wants full chroma
2636
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2637
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2638
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2639
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2640
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2641
        c->chrSrcHSubSample=1;
2642

    
2643
    if (param) {
2644
        c->param[0] = param[0];
2645
        c->param[1] = param[1];
2646
    } else {
2647
        c->param[0] =
2648
        c->param[1] = SWS_PARAM_DEFAULT;
2649
    }
2650

    
2651
    // Note the -((-x)>>y) is so that we always round toward +inf.
2652
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2653
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2654
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2655
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2656

    
2657
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2658

    
2659
    /* unscaled special cases */
2660
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {
2661
        /* yv12_to_nv12 */
2662
        if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
2663
            c->swScale= PlanarToNV12Wrapper;
2664
        }
2665
        /* yuv2bgr */
2666
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2667
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
2668
            c->swScale= ff_yuv2rgb_get_func_ptr(c);
2669
        }
2670

    
2671
        if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
2672
            c->swScale= yvu9toyv12Wrapper;
2673
        }
2674

    
2675
        /* bgr24toYV12 */
2676
        if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2677
            c->swScale= bgr24toyv12Wrapper;
2678

    
2679
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2680
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2681
           && (isBGR(dstFormat) || isRGB(dstFormat))
2682
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2683
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2684
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2685
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2686
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2687
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2688
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2689
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2690
                                             && dstFormat != PIX_FMT_RGB32_1
2691
                                             && dstFormat != PIX_FMT_BGR32_1
2692
           && srcFormat != PIX_FMT_RGB48LE   && dstFormat != PIX_FMT_RGB48LE
2693
           && srcFormat != PIX_FMT_RGB48BE   && dstFormat != PIX_FMT_RGB48BE
2694
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2695
             c->swScale= rgb2rgbWrapper;
2696

    
2697
        if ((usePal(srcFormat) && (
2698
                 dstFormat == PIX_FMT_RGB32   ||
2699
                 dstFormat == PIX_FMT_RGB32_1 ||
2700
                 dstFormat == PIX_FMT_RGB24   ||
2701
                 dstFormat == PIX_FMT_BGR32   ||
2702
                 dstFormat == PIX_FMT_BGR32_1 ||
2703
                 dstFormat == PIX_FMT_BGR24)))
2704
             c->swScale= pal2rgbWrapper;
2705

    
2706
        if (srcFormat == PIX_FMT_YUV422P) {
2707
            if (dstFormat == PIX_FMT_YUYV422)
2708
                c->swScale= YUV422PToYuy2Wrapper;
2709
            else if (dstFormat == PIX_FMT_UYVY422)
2710
                c->swScale= YUV422PToUyvyWrapper;
2711
        }
2712

    
2713
        /* LQ converters if -sws 0 or -sws 4*/
2714
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
2715
            /* yv12_to_yuy2 */
2716
            if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
2717
                if (dstFormat == PIX_FMT_YUYV422)
2718
                    c->swScale= PlanarToYuy2Wrapper;
2719
                else if (dstFormat == PIX_FMT_UYVY422)
2720
                    c->swScale= PlanarToUyvyWrapper;
2721
            }
2722
        }
2723
        if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2724
            c->swScale= YUYV2YUV420Wrapper;
2725
        if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2726
            c->swScale= UYVY2YUV420Wrapper;
2727
        if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2728
            c->swScale= YUYV2YUV422Wrapper;
2729
        if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2730
            c->swScale= UYVY2YUV422Wrapper;
2731

    
2732
#ifdef COMPILE_ALTIVEC
2733
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2734
            !(c->flags & SWS_BITEXACT) &&
2735
            srcFormat == PIX_FMT_YUV420P) {
2736
          // unscaled YV12 -> packed YUV, we want speed
2737
          if (dstFormat == PIX_FMT_YUYV422)
2738
              c->swScale= yv12toyuy2_unscaled_altivec;
2739
          else if (dstFormat == PIX_FMT_UYVY422)
2740
              c->swScale= yv12touyvy_unscaled_altivec;
2741
        }
2742
#endif
2743

    
2744
        /* simple copy */
2745
        if (  srcFormat == dstFormat
2746
            || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2747
            || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2748
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2749
            || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2750
            || (isGray(dstFormat) && isGray(srcFormat))
2751
            || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2752
                && c->chrDstHSubSample == c->chrSrcHSubSample
2753
                && c->chrDstVSubSample == c->chrSrcVSubSample
2754
                && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
2755
                && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
2756
        {
2757
            if (isPacked(c->srcFormat))
2758
                c->swScale= packedCopy;
2759
            else /* Planar YUV or gray */
2760
                c->swScale= planarCopy;
2761
        }
2762
#if ARCH_BFIN
2763
        if (flags & SWS_CPU_CAPS_BFIN)
2764
            ff_bfin_get_unscaled_swscale (c);
2765
#endif
2766

    
2767
        if (c->swScale) {
2768
            if (flags&SWS_PRINT_INFO)
2769
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2770
                       sws_format_name(srcFormat), sws_format_name(dstFormat));
2771
            return c;
2772
        }
2773
    }
2774

    
2775
    if (flags & SWS_CPU_CAPS_MMX2) {
2776
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2777
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
2778
            if (flags&SWS_PRINT_INFO)
2779
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2780
        }
2781
        if (usesHFilter) c->canMMX2BeUsed=0;
2782
    }
2783
    else
2784
        c->canMMX2BeUsed=0;
2785

    
2786
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2787
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2788

    
2789
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2790
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2791
    // n-2 is the last chrominance sample available
2792
    // this is not perfect, but no one should notice the difference, the more correct variant
2793
    // would be like the vertical one, but that would require some special code for the
2794
    // first and last pixel
2795
    if (flags&SWS_FAST_BILINEAR) {
2796
        if (c->canMMX2BeUsed) {
2797
            c->lumXInc+= 20;
2798
            c->chrXInc+= 20;
2799
        }
2800
        //we don't use the x86 asm scaler if MMX is available
2801
        else if (flags & SWS_CPU_CAPS_MMX) {
2802
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2803
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2804
        }
2805
    }
2806

    
2807
    /* precalculate horizontal scaler filter coefficients */
2808
    {
2809
        const int filterAlign=
2810
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2811
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2812
            1;
2813

    
2814
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2815
                   srcW      ,       dstW, filterAlign, 1<<14,
2816
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2817
                   srcFilter->lumH, dstFilter->lumH, c->param);
2818
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2819
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2820
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2821
                   srcFilter->chrH, dstFilter->chrH, c->param);
2822

    
2823
#if defined(COMPILE_MMX2)
2824
// can't downscale !!!
2825
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
2826
            c->lumMmx2FilterCodeSize = initMMX2HScaler(      dstW, c->lumXInc, NULL, NULL, NULL, 8);
2827
            c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
2828

    
2829
#ifdef MAP_ANONYMOUS
2830
            c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2831
            c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2832
#elif HAVE_VIRTUALALLOC
2833
            c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2834
            c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2835
#else
2836
            c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
2837
            c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
2838
#endif
2839

    
2840
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2841
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2842
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2843
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2844

    
2845
            initMMX2HScaler(      dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2846
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2847

    
2848
#ifdef MAP_ANONYMOUS
2849
            mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
2850
            mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
2851
#endif
2852
        }
2853
#endif /* defined(COMPILE_MMX2) */
2854
    } // initialize horizontal stuff
2855

    
2856

    
2857

    
2858
    /* precalculate vertical scaler filter coefficients */
2859
    {
2860
        const int filterAlign=
2861
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2862
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2863
            1;
2864

    
2865
        initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2866
                   srcH      ,        dstH, filterAlign, (1<<12),
2867
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2868
                   srcFilter->lumV, dstFilter->lumV, c->param);
2869
        initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2870
                   c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2871
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2872
                   srcFilter->chrV, dstFilter->chrV, c->param);
2873

    
2874
#ifdef COMPILE_ALTIVEC
2875
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2876
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2877

    
2878
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2879
            int j;
2880
            short *p = (short *)&c->vYCoeffsBank[i];
2881
            for (j=0;j<8;j++)
2882
                p[j] = c->vLumFilter[i];
2883
        }
2884

    
2885
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2886
            int j;
2887
            short *p = (short *)&c->vCCoeffsBank[i];
2888
            for (j=0;j<8;j++)
2889
                p[j] = c->vChrFilter[i];
2890
        }
2891
#endif
2892
    }
2893

    
2894
    // calculate buffer sizes so that they won't run out while handling these damn slices
2895
    c->vLumBufSize= c->vLumFilterSize;
2896
    c->vChrBufSize= c->vChrFilterSize;
2897
    for (i=0; i<dstH; i++) {
2898
        int chrI= i*c->chrDstH / dstH;
2899
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2900
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2901

    
2902
        nextSlice>>= c->chrSrcVSubSample;
2903
        nextSlice<<= c->chrSrcVSubSample;
2904
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2905
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2906
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2907
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2908
    }
2909

    
2910
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2911
    // allocate several megabytes to handle all possible cases)
2912
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2913
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2914
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2915
        c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2916
    //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)
2917
    /* align at 16 bytes for AltiVec */
2918
    for (i=0; i<c->vLumBufSize; i++)
2919
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2920
    for (i=0; i<c->vChrBufSize; i++)
2921
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2922
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2923
        for (i=0; i<c->vLumBufSize; i++)
2924
            c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2925

    
2926
    //try to avoid drawing green stuff between the right end and the stride end
2927
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2928

    
2929
    assert(2*VOFW == VOF);
2930

    
2931
    assert(c->chrDstH <= dstH);
2932

    
2933
    if (flags&SWS_PRINT_INFO) {
2934
#ifdef DITHER1XBPP
2935
        const char *dither= " dithered";
2936
#else
2937
        const char *dither= "";
2938
#endif
2939
        if (flags&SWS_FAST_BILINEAR)
2940
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2941
        else if (flags&SWS_BILINEAR)
2942
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2943
        else if (flags&SWS_BICUBIC)
2944
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2945
        else if (flags&SWS_X)
2946
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2947
        else if (flags&SWS_POINT)
2948
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2949
        else if (flags&SWS_AREA)
2950
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2951
        else if (flags&SWS_BICUBLIN)
2952
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2953
        else if (flags&SWS_GAUSS)
2954
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2955
        else if (flags&SWS_SINC)
2956
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2957
        else if (flags&SWS_LANCZOS)
2958
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2959
        else if (flags&SWS_SPLINE)
2960
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2961
        else
2962
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2963

    
2964
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2965
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2966
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2967
        else
2968
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2969
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2970

    
2971
        if (flags & SWS_CPU_CAPS_MMX2)
2972
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2973
        else if (flags & SWS_CPU_CAPS_3DNOW)
2974
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2975
        else if (flags & SWS_CPU_CAPS_MMX)
2976
            av_log(c, AV_LOG_INFO, "using MMX\n");
2977
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2978
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2979
        else
2980
            av_log(c, AV_LOG_INFO, "using C\n");
2981
    }
2982

    
2983
    if (flags & SWS_PRINT_INFO) {
2984
        if (flags & SWS_CPU_CAPS_MMX) {
2985
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2986
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2987
            else {
2988
                if (c->hLumFilterSize==4)
2989
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2990
                else if (c->hLumFilterSize==8)
2991
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2992
                else
2993
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2994

    
2995
                if (c->hChrFilterSize==4)
2996
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2997
                else if (c->hChrFilterSize==8)
2998
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2999
                else
3000
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
3001
            }
3002
        } else {
3003
#if ARCH_X86
3004
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
3005
#else
3006
            if (flags & SWS_FAST_BILINEAR)
3007
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
3008
            else
3009
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
3010
#endif
3011
        }
3012
        if (isPlanarYUV(dstFormat)) {
3013
            if (c->vLumFilterSize==1)
3014
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3015
            else
3016
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3017
        } else {
3018
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
3019
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
3020
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3021
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
3022
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3023
            else
3024
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3025
        }
3026

    
3027
        if (dstFormat==PIX_FMT_BGR24)
3028
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
3029
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
3030
        else if (dstFormat==PIX_FMT_RGB32)
3031
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3032
        else if (dstFormat==PIX_FMT_BGR565)
3033
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3034
        else if (dstFormat==PIX_FMT_BGR555)
3035
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
3036

    
3037
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
3038
    }
3039
    if (flags & SWS_PRINT_INFO) {
3040
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3041
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
3042
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
3043
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
3044
    }
3045

    
3046
    c->swScale= getSwsFunc(c);
3047
    return c;
3048
}
3049

    
3050
static void reset_ptr(uint8_t* src[], int format)
3051
{
3052
    if(!isALPHA(format))
3053
        src[3]=NULL;
3054
    if(!isPlanarYUV(format)) {
3055
        src[3]=src[2]=NULL;
3056
        if(   format != PIX_FMT_PAL8
3057
           && format != PIX_FMT_RGB8
3058
           && format != PIX_FMT_BGR8
3059
           && format != PIX_FMT_RGB4_BYTE
3060
           && format != PIX_FMT_BGR4_BYTE
3061
          )
3062
            src[1]= NULL;
3063
    }
3064
}
3065

    
3066
/**
3067
 * swscale wrapper, so we don't need to export the SwsContext.
3068
 * Assumes planar YUV to be in YUV order instead of YVU.
3069
 */
3070
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3071
              int srcSliceH, uint8_t* dst[], int dstStride[])
3072
{
3073
    int i;
3074
    uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
3075
    uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
3076

    
3077
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
3078
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
3079
        return 0;
3080
    }
3081
    if (c->sliceDir == 0) {
3082
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
3083
    }
3084

    
3085
    if (usePal(c->srcFormat)) {
3086
        for (i=0; i<256; i++) {
3087
            int p, r, g, b,y,u,v;
3088
            if(c->srcFormat == PIX_FMT_PAL8) {
3089
                p=((uint32_t*)(src[1]))[i];
3090
                r= (p>>16)&0xFF;
3091
                g= (p>> 8)&0xFF;
3092
                b=  p     &0xFF;
3093
            } else if(c->srcFormat == PIX_FMT_RGB8) {
3094
                r= (i>>5    )*36;
3095
                g= ((i>>2)&7)*36;
3096
                b= (i&3     )*85;
3097
            } else if(c->srcFormat == PIX_FMT_BGR8) {
3098
                b= (i>>6    )*85;
3099
                g= ((i>>3)&7)*36;
3100
                r= (i&7     )*36;
3101
            } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
3102
                r= (i>>3    )*255;
3103
                g= ((i>>1)&3)*85;
3104
                b= (i&1     )*255;
3105
            } else {
3106
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
3107
                b= (i>>3    )*255;
3108
                g= ((i>>1)&3)*85;
3109
                r= (i&1     )*255;
3110
            }
3111
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3112
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3113
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3114
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
3115

    
3116

    
3117
            switch(c->dstFormat) {
3118
            case PIX_FMT_BGR32:
3119
#if !HAVE_BIGENDIAN
3120
            case PIX_FMT_RGB24:
3121
#endif
3122
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
3123
                break;
3124
            case PIX_FMT_BGR32_1:
3125
#if HAVE_BIGENDIAN
3126
            case PIX_FMT_BGR24:
3127
#endif
3128
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
3129
                break;
3130
            case PIX_FMT_RGB32_1:
3131
#if HAVE_BIGENDIAN
3132
            case PIX_FMT_RGB24:
3133
#endif
3134
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3135
                break;
3136
            case PIX_FMT_RGB32:
3137
#if !HAVE_BIGENDIAN
3138
            case PIX_FMT_BGR24:
3139
#endif
3140
            default:
3141
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
3142
            }
3143
        }
3144
    }
3145

    
3146
    // copy strides, so they can safely be modified
3147
    if (c->sliceDir == 1) {
3148
        // slices go from top to bottom
3149
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3150
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3151

    
3152
        reset_ptr(src2, c->srcFormat);
3153
        reset_ptr(dst2, c->dstFormat);
3154

    
3155
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
3156
    } else {
3157
        // slices go from bottom to top => we flip the image internally
3158
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3159
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3160

    
3161
        src2[0] += (srcSliceH-1)*srcStride[0];
3162
        if (!usePal(c->srcFormat))
3163
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3164
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3165
        src2[3] += (srcSliceH-1)*srcStride[3];
3166
        dst2[0] += ( c->dstH                      -1)*dstStride[0];
3167
        dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3168
        dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3169
        dst2[3] += ( c->dstH                      -1)*dstStride[3];
3170

    
3171
        reset_ptr(src2, c->srcFormat);
3172
        reset_ptr(dst2, c->dstFormat);
3173

    
3174
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3175
    }
3176
}
3177

    
3178
#if LIBSWSCALE_VERSION_MAJOR < 1
3179
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3180
                      int srcSliceH, uint8_t* dst[], int dstStride[])
3181
{
3182
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3183
}
3184
#endif
3185

    
3186
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3187
                                float lumaSharpen, float chromaSharpen,
3188
                                float chromaHShift, float chromaVShift,
3189
                                int verbose)
3190
{
3191
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3192

    
3193
    if (lumaGBlur!=0.0) {
3194
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3195
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3196
    } else {
3197
        filter->lumH= sws_getIdentityVec();
3198
        filter->lumV= sws_getIdentityVec();
3199
    }
3200

    
3201
    if (chromaGBlur!=0.0) {
3202
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3203
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3204
    } else {
3205
        filter->chrH= sws_getIdentityVec();
3206
        filter->chrV= sws_getIdentityVec();
3207
    }
3208

    
3209
    if (chromaSharpen!=0.0) {
3210
        SwsVector *id= sws_getIdentityVec();
3211
        sws_scaleVec(filter->chrH, -chromaSharpen);
3212
        sws_scaleVec(filter->chrV, -chromaSharpen);
3213
        sws_addVec(filter->chrH, id);
3214
        sws_addVec(filter->chrV, id);
3215
        sws_freeVec(id);
3216
    }
3217

    
3218
    if (lumaSharpen!=0.0) {
3219
        SwsVector *id= sws_getIdentityVec();
3220
        sws_scaleVec(filter->lumH, -lumaSharpen);
3221
        sws_scaleVec(filter->lumV, -lumaSharpen);
3222
        sws_addVec(filter->lumH, id);
3223
        sws_addVec(filter->lumV, id);
3224
        sws_freeVec(id);
3225
    }
3226

    
3227
    if (chromaHShift != 0.0)
3228
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3229

    
3230
    if (chromaVShift != 0.0)
3231
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3232

    
3233
    sws_normalizeVec(filter->chrH, 1.0);
3234
    sws_normalizeVec(filter->chrV, 1.0);
3235
    sws_normalizeVec(filter->lumH, 1.0);
3236
    sws_normalizeVec(filter->lumV, 1.0);
3237

    
3238
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3239
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3240

    
3241
    return filter;
3242
}
3243

    
3244
SwsVector *sws_allocVec(int length)
3245
{
3246
    SwsVector *vec = av_malloc(sizeof(SwsVector));
3247
    if (!vec)
3248
        return NULL;
3249
    vec->length = length;
3250
    vec->coeff  = av_malloc(sizeof(double) * length);
3251
    if (!vec->coeff)
3252
        av_freep(&vec);
3253
    return vec;
3254
}
3255

    
3256
SwsVector *sws_getGaussianVec(double variance, double quality)
3257
{
3258
    const int length= (int)(variance*quality + 0.5) | 1;
3259
    int i;
3260
    double middle= (length-1)*0.5;
3261
    SwsVector *vec= sws_allocVec(length);
3262

    
3263
    if (!vec)
3264
        return NULL;
3265

    
3266
    for (i=0; i<length; i++) {
3267
        double dist= i-middle;
3268
        vec->coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3269
    }
3270

    
3271
    sws_normalizeVec(vec, 1.0);
3272

    
3273
    return vec;
3274
}
3275

    
3276
SwsVector *sws_getConstVec(double c, int length)
3277
{
3278
    int i;
3279
    SwsVector *vec= sws_allocVec(length);
3280

    
3281
    if (!vec)
3282
        return NULL;
3283

    
3284
    for (i=0; i<length; i++)
3285
        vec->coeff[i]= c;
3286

    
3287
    return vec;
3288
}
3289

    
3290

    
3291
SwsVector *sws_getIdentityVec(void)
3292
{
3293
    return sws_getConstVec(1.0, 1);
3294
}
3295

    
3296
double sws_dcVec(SwsVector *a)
3297
{
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
{
3309
    int i;
3310

    
3311
    for (i=0; i<a->length; i++)
3312
        a->coeff[i]*= scalar;
3313
}
3314

    
3315
void sws_normalizeVec(SwsVector *a, double height)
3316
{
3317
    sws_scaleVec(a, height/sws_dcVec(a));
3318
}
3319

    
3320
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
3321
{
3322
    int length= a->length + b->length - 1;
3323
    int i, j;
3324
    SwsVector *vec= sws_getConstVec(0.0, length);
3325

    
3326
    if (!vec)
3327
        return NULL;
3328

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

    
3335
    return vec;
3336
}
3337

    
3338
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
3339
{
3340
    int length= FFMAX(a->length, b->length);
3341
    int i;
3342
    SwsVector *vec= sws_getConstVec(0.0, length);
3343

    
3344
    if (!vec)
3345
        return NULL;
3346

    
3347
    for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3348
    for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3349

    
3350
    return vec;
3351
}
3352

    
3353
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
3354
{
3355
    int length= FFMAX(a->length, b->length);
3356
    int i;
3357
    SwsVector *vec= sws_getConstVec(0.0, length);
3358

    
3359
    if (!vec)
3360
        return NULL;
3361

    
3362
    for (i=0; i<a->length; i++) vec->coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3363
    for (i=0; i<b->length; i++) vec->coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3364

    
3365
    return vec;
3366
}
3367

    
3368
/* shift left / or right if "shift" is negative */
3369
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
3370
{
3371
    int length= a->length + FFABS(shift)*2;
3372
    int i;
3373
    SwsVector *vec= sws_getConstVec(0.0, length);
3374

    
3375
    if (!vec)
3376
        return NULL;
3377

    
3378
    for (i=0; i<a->length; i++) {
3379
        vec->coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3380
    }
3381

    
3382
    return vec;
3383
}
3384

    
3385
void sws_shiftVec(SwsVector *a, int shift)
3386
{
3387
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3388
    av_free(a->coeff);
3389
    a->coeff= shifted->coeff;
3390
    a->length= shifted->length;
3391
    av_free(shifted);
3392
}
3393

    
3394
void sws_addVec(SwsVector *a, SwsVector *b)
3395
{
3396
    SwsVector *sum= sws_sumVec(a, b);
3397
    av_free(a->coeff);
3398
    a->coeff= sum->coeff;
3399
    a->length= sum->length;
3400
    av_free(sum);
3401
}
3402

    
3403
void sws_subVec(SwsVector *a, SwsVector *b)
3404
{
3405
    SwsVector *diff= sws_diffVec(a, b);
3406
    av_free(a->coeff);
3407
    a->coeff= diff->coeff;
3408
    a->length= diff->length;
3409
    av_free(diff);
3410
}
3411

    
3412
void sws_convVec(SwsVector *a, SwsVector *b)
3413
{
3414
    SwsVector *conv= sws_getConvVec(a, b);
3415
    av_free(a->coeff);
3416
    a->coeff= conv->coeff;
3417
    a->length= conv->length;
3418
    av_free(conv);
3419
}
3420

    
3421
SwsVector *sws_cloneVec(SwsVector *a)
3422
{
3423
    int i;
3424
    SwsVector *vec= sws_allocVec(a->length);
3425

    
3426
    if (!vec)
3427
        return NULL;
3428

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

    
3431
    return vec;
3432
}
3433

    
3434
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
3435
{
3436
    int i;
3437
    double max=0;
3438
    double min=0;
3439
    double range;
3440

    
3441
    for (i=0; i<a->length; i++)
3442
        if (a->coeff[i]>max) max= a->coeff[i];
3443

    
3444
    for (i=0; i<a->length; i++)
3445
        if (a->coeff[i]<min) min= a->coeff[i];
3446

    
3447
    range= max - min;
3448

    
3449
    for (i=0; i<a->length; i++) {
3450
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3451
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3452
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3453
        av_log(log_ctx, log_level, "|\n");
3454
    }
3455
}
3456

    
3457
#if LIBSWSCALE_VERSION_MAJOR < 1
3458
void sws_printVec(SwsVector *a)
3459
{
3460
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3461
}
3462
#endif
3463

    
3464
void sws_freeVec(SwsVector *a)
3465
{
3466
    if (!a) return;
3467
    av_freep(&a->coeff);
3468
    a->length=0;
3469
    av_free(a);
3470
}
3471

    
3472
void sws_freeFilter(SwsFilter *filter)
3473
{
3474
    if (!filter) return;
3475

    
3476
    if (filter->lumH) sws_freeVec(filter->lumH);
3477
    if (filter->lumV) sws_freeVec(filter->lumV);
3478
    if (filter->chrH) sws_freeVec(filter->chrH);
3479
    if (filter->chrV) sws_freeVec(filter->chrV);
3480
    av_free(filter);
3481
}
3482

    
3483

    
3484
void sws_freeContext(SwsContext *c)
3485
{
3486
    int i;
3487
    if (!c) return;
3488

    
3489
    if (c->lumPixBuf) {
3490
        for (i=0; i<c->vLumBufSize; i++)
3491
            av_freep(&c->lumPixBuf[i]);
3492
        av_freep(&c->lumPixBuf);
3493
    }
3494

    
3495
    if (c->chrPixBuf) {
3496
        for (i=0; i<c->vChrBufSize; i++)
3497
            av_freep(&c->chrPixBuf[i]);
3498
        av_freep(&c->chrPixBuf);
3499
    }
3500

    
3501
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
3502
        for (i=0; i<c->vLumBufSize; i++)
3503
            av_freep(&c->alpPixBuf[i]);
3504
        av_freep(&c->alpPixBuf);
3505
    }
3506

    
3507
    av_freep(&c->vLumFilter);
3508
    av_freep(&c->vChrFilter);
3509
    av_freep(&c->hLumFilter);
3510
    av_freep(&c->hChrFilter);
3511
#ifdef COMPILE_ALTIVEC
3512
    av_freep(&c->vYCoeffsBank);
3513
    av_freep(&c->vCCoeffsBank);
3514
#endif
3515

    
3516
    av_freep(&c->vLumFilterPos);
3517
    av_freep(&c->vChrFilterPos);
3518
    av_freep(&c->hLumFilterPos);
3519
    av_freep(&c->hChrFilterPos);
3520

    
3521
#if ARCH_X86 && CONFIG_GPL
3522
#ifdef MAP_ANONYMOUS
3523
    if (c->lumMmx2FilterCode) munmap(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize);
3524
    if (c->chrMmx2FilterCode) munmap(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize);
3525
#elif HAVE_VIRTUALALLOC
3526
    if (c->lumMmx2FilterCode) VirtualFree(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, MEM_RELEASE);
3527
    if (c->chrMmx2FilterCode) VirtualFree(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, MEM_RELEASE);
3528
#else
3529
    av_free(c->lumMmx2FilterCode);
3530
    av_free(c->chrMmx2FilterCode);
3531
#endif
3532
    c->lumMmx2FilterCode=NULL;
3533
    c->chrMmx2FilterCode=NULL;
3534
#endif /* ARCH_X86 && CONFIG_GPL */
3535

    
3536
    av_freep(&c->lumMmx2Filter);
3537
    av_freep(&c->chrMmx2Filter);
3538
    av_freep(&c->lumMmx2FilterPos);
3539
    av_freep(&c->chrMmx2FilterPos);
3540
    av_freep(&c->yuvTable);
3541

    
3542
    av_free(c);
3543
}
3544

    
3545
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3546
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3547
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3548
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
3549
{
3550
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3551

    
3552
    if (!param)
3553
        param = default_param;
3554

    
3555
    if (context) {
3556
        if (context->srcW != srcW || context->srcH != srcH ||
3557
            context->srcFormat != srcFormat ||
3558
            context->dstW != dstW || context->dstH != dstH ||
3559
            context->dstFormat != dstFormat || context->flags != flags ||
3560
            context->param[0] != param[0] || context->param[1] != param[1])
3561
        {
3562
            sws_freeContext(context);
3563
            context = NULL;
3564
        }
3565
    }
3566
    if (!context) {
3567
        return sws_getContext(srcW, srcH, srcFormat,
3568
                              dstW, dstH, dstFormat, flags,
3569
                              srcFilter, dstFilter, param);
3570
    }
3571
    return context;
3572
}
3573