Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 13394e8c

History | View | Annotate | Download (104 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 <unistd.h>
63
#include "config.h"
64
#include <assert.h>
65
#ifdef HAVE_SYS_MMAN_H
66
#include <sys/mman.h>
67
#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
68
#define MAP_ANONYMOUS MAP_ANON
69
#endif
70
#endif
71
#include "swscale.h"
72
#include "swscale_internal.h"
73
#include "rgb2rgb.h"
74
#include "libavutil/x86_cpu.h"
75
#include "libavutil/bswap.h"
76

    
77
unsigned swscale_version(void)
78
{
79
    return LIBSWSCALE_VERSION_INT;
80
}
81

    
82
#undef MOVNTQ
83
#undef PAVGB
84

    
85
//#undef HAVE_MMX2
86
//#define HAVE_3DNOW
87
//#undef HAVE_MMX
88
//#undef ARCH_X86
89
//#define WORDS_BIGENDIAN
90
#define DITHER1XBPP
91

    
92
#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
93

    
94
#define RET 0xC3 //near return opcode for X86
95

    
96
#ifdef M_PI
97
#define PI M_PI
98
#else
99
#define PI 3.14159265358979323846
100
#endif
101

    
102
#define isSupportedIn(x)    (       \
103
           (x)==PIX_FMT_YUV420P     \
104
        || (x)==PIX_FMT_YUVA420P    \
105
        || (x)==PIX_FMT_YUYV422     \
106
        || (x)==PIX_FMT_UYVY422     \
107
        || (x)==PIX_FMT_RGB32       \
108
        || (x)==PIX_FMT_RGB32_1     \
109
        || (x)==PIX_FMT_BGR24       \
110
        || (x)==PIX_FMT_BGR565      \
111
        || (x)==PIX_FMT_BGR555      \
112
        || (x)==PIX_FMT_BGR32       \
113
        || (x)==PIX_FMT_BGR32_1     \
114
        || (x)==PIX_FMT_RGB24       \
115
        || (x)==PIX_FMT_RGB565      \
116
        || (x)==PIX_FMT_RGB555      \
117
        || (x)==PIX_FMT_GRAY8       \
118
        || (x)==PIX_FMT_YUV410P     \
119
        || (x)==PIX_FMT_YUV440P     \
120
        || (x)==PIX_FMT_GRAY16BE    \
121
        || (x)==PIX_FMT_GRAY16LE    \
122
        || (x)==PIX_FMT_YUV444P     \
123
        || (x)==PIX_FMT_YUV422P     \
124
        || (x)==PIX_FMT_YUV411P     \
125
        || (x)==PIX_FMT_PAL8        \
126
        || (x)==PIX_FMT_BGR8        \
127
        || (x)==PIX_FMT_RGB8        \
128
        || (x)==PIX_FMT_BGR4_BYTE   \
129
        || (x)==PIX_FMT_RGB4_BYTE   \
130
        || (x)==PIX_FMT_YUV440P     \
131
        || (x)==PIX_FMT_MONOWHITE   \
132
        || (x)==PIX_FMT_MONOBLACK   \
133
    )
134
#define isSupportedOut(x)   (       \
135
           (x)==PIX_FMT_YUV420P     \
136
        || (x)==PIX_FMT_YUYV422     \
137
        || (x)==PIX_FMT_UYVY422     \
138
        || (x)==PIX_FMT_YUV444P     \
139
        || (x)==PIX_FMT_YUV422P     \
140
        || (x)==PIX_FMT_YUV411P     \
141
        || isRGB(x)                 \
142
        || isBGR(x)                 \
143
        || (x)==PIX_FMT_NV12        \
144
        || (x)==PIX_FMT_NV21        \
145
        || (x)==PIX_FMT_GRAY16BE    \
146
        || (x)==PIX_FMT_GRAY16LE    \
147
        || (x)==PIX_FMT_GRAY8       \
148
        || (x)==PIX_FMT_YUV410P     \
149
        || (x)==PIX_FMT_YUV440P     \
150
    )
151
#define isPacked(x)         (       \
152
           (x)==PIX_FMT_PAL8        \
153
        || (x)==PIX_FMT_YUYV422     \
154
        || (x)==PIX_FMT_UYVY422     \
155
        || isRGB(x)                 \
156
        || isBGR(x)                 \
157
    )
158

    
159
#define RGB2YUV_SHIFT 15
160
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
161
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
162
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
163
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
164
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
165
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
166
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
167
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
168
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
169

    
170
extern const int32_t Inverse_Table_6_9[8][4];
171

    
172
static const double rgb2yuv_table[8][9]={
173
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
174
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
175
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
176
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
177
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
178
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
179
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
180
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
181
};
182

    
183
/*
184
NOTES
185
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
186

187
TODO
188
more intelligent misalignment avoidance for the horizontal scaler
189
write special vertical cubic upscale version
190
Optimize C code (yv12 / minmax)
191
add support for packed pixel yuv input & output
192
add support for Y8 output
193
optimize bgr24 & bgr32
194
add BGR4 output support
195
write special BGR->BGR scaler
196
*/
197

    
198
#if defined(ARCH_X86) && defined (CONFIG_GPL)
199
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
200
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
201
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
202
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
203
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
204
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
205
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
206
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
207

    
208
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
209
        0x0103010301030103LL,
210
        0x0200020002000200LL,};
211

    
212
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
213
        0x0602060206020602LL,
214
        0x0004000400040004LL,};
215

    
216
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
217
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
218
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
219
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
220
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
221
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
222

    
223
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
224
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
225
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
226

    
227
#ifdef FAST_BGR2YV12
228
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
229
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
230
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
231
#else
232
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
233
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
234
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
235
#endif /* FAST_BGR2YV12 */
236
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
237
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
238
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
239

    
240
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
241
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
242
DECLARE_ALIGNED(8, const uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
243
DECLARE_ALIGNED(8, const uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
244
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
245

    
246
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toUV[2][4]) = {
247
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
248
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
249
};
250

    
251
DECLARE_ALIGNED(8, const uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
252

    
253
#endif /* defined(ARCH_X86) */
254

    
255
// clipping helper table for C implementations:
256
static unsigned char clip_table[768];
257

    
258
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
259

    
260
const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={
261
{  1,   3,   1,   3,   1,   3,   1,   3, },
262
{  2,   0,   2,   0,   2,   0,   2,   0, },
263
};
264

    
265
const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={
266
{  6,   2,   6,   2,   6,   2,   6,   2, },
267
{  0,   4,   0,   4,   0,   4,   0,   4, },
268
};
269

    
270
const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={
271
{ 17,   9,  23,  15,  16,   8,  22,  14, },
272
{  5,  29,   3,  27,   4,  28,   2,  26, },
273
{ 21,  13,  19,  11,  20,  12,  18,  10, },
274
{  0,  24,   6,  30,   1,  25,   7,  31, },
275
{ 16,   8,  22,  14,  17,   9,  23,  15, },
276
{  4,  28,   2,  26,   5,  29,   3,  27, },
277
{ 20,  12,  18,  10,  21,  13,  19,  11, },
278
{  1,  25,   7,  31,   0,  24,   6,  30, },
279
};
280

    
281
#if 0
282
const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={
283
{  0,  48,  12,  60,   3,  51,  15,  63, },
284
{ 32,  16,  44,  28,  35,  19,  47,  31, },
285
{  8,  56,   4,  52,  11,  59,   7,  55, },
286
{ 40,  24,  36,  20,  43,  27,  39,  23, },
287
{  2,  50,  14,  62,   1,  49,  13,  61, },
288
{ 34,  18,  46,  30,  33,  17,  45,  29, },
289
{ 10,  58,   6,  54,   9,  57,   5,  53, },
290
{ 42,  26,  38,  22,  41,  25,  37,  21, },
291
};
292
#endif
293

    
294
const uint8_t  __attribute__((aligned(8))) dither_8x8_73[8][8]={
295
{  0,  55,  14,  68,   3,  58,  17,  72, },
296
{ 37,  18,  50,  32,  40,  22,  54,  35, },
297
{  9,  64,   5,  59,  13,  67,   8,  63, },
298
{ 46,  27,  41,  23,  49,  31,  44,  26, },
299
{  2,  57,  16,  71,   1,  56,  15,  70, },
300
{ 39,  21,  52,  34,  38,  19,  51,  33, },
301
{ 11,  66,   7,  62,  10,  65,   6,  60, },
302
{ 48,  30,  43,  25,  47,  29,  42,  24, },
303
};
304

    
305
#if 0
306
const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={
307
{ 68,  36,  92,  60,  66,  34,  90,  58, },
308
{ 20, 116,  12, 108,  18, 114,  10, 106, },
309
{ 84,  52,  76,  44,  82,  50,  74,  42, },
310
{  0,  96,  24, 120,   6, 102,  30, 126, },
311
{ 64,  32,  88,  56,  70,  38,  94,  62, },
312
{ 16, 112,   8, 104,  22, 118,  14, 110, },
313
{ 80,  48,  72,  40,  86,  54,  78,  46, },
314
{  4, 100,  28, 124,   2,  98,  26, 122, },
315
};
316
#endif
317

    
318
#if 1
319
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
320
{117,  62, 158, 103, 113,  58, 155, 100, },
321
{ 34, 199,  21, 186,  31, 196,  17, 182, },
322
{144,  89, 131,  76, 141,  86, 127,  72, },
323
{  0, 165,  41, 206,  10, 175,  52, 217, },
324
{110,  55, 151,  96, 120,  65, 162, 107, },
325
{ 28, 193,  14, 179,  38, 203,  24, 189, },
326
{138,  83, 124,  69, 148,  93, 134,  79, },
327
{  7, 172,  48, 213,   3, 168,  45, 210, },
328
};
329
#elif 1
330
// tries to correct a gamma of 1.5
331
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
332
{  0, 143,  18, 200,   2, 156,  25, 215, },
333
{ 78,  28, 125,  64,  89,  36, 138,  74, },
334
{ 10, 180,   3, 161,  16, 195,   8, 175, },
335
{109,  51,  93,  38, 121,  60, 105,  47, },
336
{  1, 152,  23, 210,   0, 147,  20, 205, },
337
{ 85,  33, 134,  71,  81,  30, 130,  67, },
338
{ 14, 190,   6, 171,  12, 185,   5, 166, },
339
{117,  57, 101,  44, 113,  54,  97,  41, },
340
};
341
#elif 1
342
// tries to correct a gamma of 2.0
343
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
344
{  0, 124,   8, 193,   0, 140,  12, 213, },
345
{ 55,  14, 104,  42,  66,  19, 119,  52, },
346
{  3, 168,   1, 145,   6, 187,   3, 162, },
347
{ 86,  31,  70,  21,  99,  39,  82,  28, },
348
{  0, 134,  11, 206,   0, 129,   9, 200, },
349
{ 62,  17, 114,  48,  58,  16, 109,  45, },
350
{  5, 181,   2, 157,   4, 175,   1, 151, },
351
{ 95,  36,  78,  26,  90,  34,  74,  24, },
352
};
353
#else
354
// tries to correct a gamma of 2.5
355
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
356
{  0, 107,   3, 187,   0, 125,   6, 212, },
357
{ 39,   7,  86,  28,  49,  11, 102,  36, },
358
{  1, 158,   0, 131,   3, 180,   1, 151, },
359
{ 68,  19,  52,  12,  81,  25,  64,  17, },
360
{  0, 119,   5, 203,   0, 113,   4, 195, },
361
{ 45,   9,  96,  33,  42,   8,  91,  30, },
362
{  2, 172,   1, 144,   2, 165,   0, 137, },
363
{ 77,  23,  60,  15,  72,  21,  56,  14, },
364
};
365
#endif
366

    
367
const char *sws_format_name(enum PixelFormat format)
368
{
369
    switch (format) {
370
        case PIX_FMT_YUV420P:
371
            return "yuv420p";
372
        case PIX_FMT_YUVA420P:
373
            return "yuva420p";
374
        case PIX_FMT_YUYV422:
375
            return "yuyv422";
376
        case PIX_FMT_RGB24:
377
            return "rgb24";
378
        case PIX_FMT_BGR24:
379
            return "bgr24";
380
        case PIX_FMT_YUV422P:
381
            return "yuv422p";
382
        case PIX_FMT_YUV444P:
383
            return "yuv444p";
384
        case PIX_FMT_RGB32:
385
            return "rgb32";
386
        case PIX_FMT_YUV410P:
387
            return "yuv410p";
388
        case PIX_FMT_YUV411P:
389
            return "yuv411p";
390
        case PIX_FMT_RGB565:
391
            return "rgb565";
392
        case PIX_FMT_RGB555:
393
            return "rgb555";
394
        case PIX_FMT_GRAY16BE:
395
            return "gray16be";
396
        case PIX_FMT_GRAY16LE:
397
            return "gray16le";
398
        case PIX_FMT_GRAY8:
399
            return "gray8";
400
        case PIX_FMT_MONOWHITE:
401
            return "mono white";
402
        case PIX_FMT_MONOBLACK:
403
            return "mono black";
404
        case PIX_FMT_PAL8:
405
            return "Palette";
406
        case PIX_FMT_YUVJ420P:
407
            return "yuvj420p";
408
        case PIX_FMT_YUVJ422P:
409
            return "yuvj422p";
410
        case PIX_FMT_YUVJ444P:
411
            return "yuvj444p";
412
        case PIX_FMT_XVMC_MPEG2_MC:
413
            return "xvmc_mpeg2_mc";
414
        case PIX_FMT_XVMC_MPEG2_IDCT:
415
            return "xvmc_mpeg2_idct";
416
        case PIX_FMT_UYVY422:
417
            return "uyvy422";
418
        case PIX_FMT_UYYVYY411:
419
            return "uyyvyy411";
420
        case PIX_FMT_RGB32_1:
421
            return "rgb32x";
422
        case PIX_FMT_BGR32_1:
423
            return "bgr32x";
424
        case PIX_FMT_BGR32:
425
            return "bgr32";
426
        case PIX_FMT_BGR565:
427
            return "bgr565";
428
        case PIX_FMT_BGR555:
429
            return "bgr555";
430
        case PIX_FMT_BGR8:
431
            return "bgr8";
432
        case PIX_FMT_BGR4:
433
            return "bgr4";
434
        case PIX_FMT_BGR4_BYTE:
435
            return "bgr4 byte";
436
        case PIX_FMT_RGB8:
437
            return "rgb8";
438
        case PIX_FMT_RGB4:
439
            return "rgb4";
440
        case PIX_FMT_RGB4_BYTE:
441
            return "rgb4 byte";
442
        case PIX_FMT_NV12:
443
            return "nv12";
444
        case PIX_FMT_NV21:
445
            return "nv21";
446
        case PIX_FMT_YUV440P:
447
            return "yuv440p";
448
        default:
449
            return "Unknown format";
450
    }
451
}
452

    
453
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
454
                               int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
455
                               uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
456
{
457
    //FIXME Optimize (just quickly writen not opti..)
458
    int i;
459
    for (i=0; i<dstW; i++)
460
    {
461
        int val=1<<18;
462
        int j;
463
        for (j=0; j<lumFilterSize; j++)
464
            val += lumSrc[j][i] * lumFilter[j];
465

    
466
        dest[i]= av_clip_uint8(val>>19);
467
    }
468

    
469
    if (uDest)
470
        for (i=0; i<chrDstW; i++)
471
        {
472
            int u=1<<18;
473
            int v=1<<18;
474
            int j;
475
            for (j=0; j<chrFilterSize; j++)
476
            {
477
                u += chrSrc[j][i] * chrFilter[j];
478
                v += chrSrc[j][i + VOFW] * chrFilter[j];
479
            }
480

    
481
            uDest[i]= av_clip_uint8(u>>19);
482
            vDest[i]= av_clip_uint8(v>>19);
483
        }
484
}
485

    
486
static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
487
                                int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
488
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
489
{
490
    //FIXME Optimize (just quickly writen not opti..)
491
    int i;
492
    for (i=0; i<dstW; i++)
493
    {
494
        int val=1<<18;
495
        int j;
496
        for (j=0; j<lumFilterSize; j++)
497
            val += lumSrc[j][i] * lumFilter[j];
498

    
499
        dest[i]= av_clip_uint8(val>>19);
500
    }
501

    
502
    if (!uDest)
503
        return;
504

    
505
    if (dstFormat == PIX_FMT_NV12)
506
        for (i=0; i<chrDstW; i++)
507
        {
508
            int u=1<<18;
509
            int v=1<<18;
510
            int j;
511
            for (j=0; j<chrFilterSize; j++)
512
            {
513
                u += chrSrc[j][i] * chrFilter[j];
514
                v += chrSrc[j][i + VOFW] * chrFilter[j];
515
            }
516

    
517
            uDest[2*i]= av_clip_uint8(u>>19);
518
            uDest[2*i+1]= av_clip_uint8(v>>19);
519
        }
520
    else
521
        for (i=0; i<chrDstW; i++)
522
        {
523
            int u=1<<18;
524
            int v=1<<18;
525
            int j;
526
            for (j=0; j<chrFilterSize; j++)
527
            {
528
                u += chrSrc[j][i] * chrFilter[j];
529
                v += chrSrc[j][i + VOFW] * chrFilter[j];
530
            }
531

    
532
            uDest[2*i]= av_clip_uint8(v>>19);
533
            uDest[2*i+1]= av_clip_uint8(u>>19);
534
        }
535
}
536

    
537
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type) \
538
    for (i=0; i<(dstW>>1); i++){\
539
        int j;\
540
        int Y1 = 1<<18;\
541
        int Y2 = 1<<18;\
542
        int U  = 1<<18;\
543
        int V  = 1<<18;\
544
        type av_unused *r, *b, *g;\
545
        const int i2= 2*i;\
546
        \
547
        for (j=0; j<lumFilterSize; j++)\
548
        {\
549
            Y1 += lumSrc[j][i2] * lumFilter[j];\
550
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
551
        }\
552
        for (j=0; j<chrFilterSize; j++)\
553
        {\
554
            U += chrSrc[j][i] * chrFilter[j];\
555
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
556
        }\
557
        Y1>>=19;\
558
        Y2>>=19;\
559
        U >>=19;\
560
        V >>=19;\
561

    
562
#define YSCALE_YUV_2_PACKEDX_C(type) \
563
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type)\
564
        if ((Y1|Y2|U|V)&256)\
565
        {\
566
            if (Y1>255)   Y1=255; \
567
            else if (Y1<0)Y1=0;   \
568
            if (Y2>255)   Y2=255; \
569
            else if (Y2<0)Y2=0;   \
570
            if (U>255)    U=255;  \
571
            else if (U<0) U=0;    \
572
            if (V>255)    V=255;  \
573
            else if (V<0) V=0;    \
574
        }
575

    
576
#define YSCALE_YUV_2_PACKEDX_FULL_C \
577
    for (i=0; i<dstW; i++){\
578
        int j;\
579
        int Y = 0;\
580
        int U = -128<<19;\
581
        int V = -128<<19;\
582
        int R,G,B;\
583
        \
584
        for (j=0; j<lumFilterSize; j++){\
585
            Y += lumSrc[j][i     ] * lumFilter[j];\
586
        }\
587
        for (j=0; j<chrFilterSize; j++){\
588
            U += chrSrc[j][i     ] * chrFilter[j];\
589
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
590
        }\
591
        Y >>=10;\
592
        U >>=10;\
593
        V >>=10;\
594

    
595
#define YSCALE_YUV_2_RGBX_FULL_C(rnd) \
596
    YSCALE_YUV_2_PACKEDX_FULL_C\
597
        Y-= c->yuv2rgb_y_offset;\
598
        Y*= c->yuv2rgb_y_coeff;\
599
        Y+= rnd;\
600
        R= Y + V*c->yuv2rgb_v2r_coeff;\
601
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
602
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
603
        if ((R|G|B)&(0xC0000000)){\
604
            if (R>=(256<<22))   R=(256<<22)-1; \
605
            else if (R<0)R=0;   \
606
            if (G>=(256<<22))   G=(256<<22)-1; \
607
            else if (G<0)G=0;   \
608
            if (B>=(256<<22))   B=(256<<22)-1; \
609
            else if (B<0)B=0;   \
610
        }\
611

    
612

    
613
#define YSCALE_YUV_2_GRAY16_C \
614
    for (i=0; i<(dstW>>1); i++){\
615
        int j;\
616
        int Y1 = 1<<18;\
617
        int Y2 = 1<<18;\
618
        int U  = 1<<18;\
619
        int V  = 1<<18;\
620
        \
621
        const int i2= 2*i;\
622
        \
623
        for (j=0; j<lumFilterSize; j++)\
624
        {\
625
            Y1 += lumSrc[j][i2] * lumFilter[j];\
626
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
627
        }\
628
        Y1>>=11;\
629
        Y2>>=11;\
630
        if ((Y1|Y2|U|V)&65536)\
631
        {\
632
            if (Y1>65535)   Y1=65535; \
633
            else if (Y1<0)Y1=0;   \
634
            if (Y2>65535)   Y2=65535; \
635
            else if (Y2<0)Y2=0;   \
636
        }
637

    
638
#define YSCALE_YUV_2_RGBX_C(type) \
639
    YSCALE_YUV_2_PACKEDX_C(type)  /* FIXME fix tables so that cliping is not needed and then use _NOCLIP*/\
640
    r = (type *)c->table_rV[V];   \
641
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
642
    b = (type *)c->table_bU[U];   \
643

    
644
#define YSCALE_YUV_2_PACKED2_C   \
645
    for (i=0; i<(dstW>>1); i++){ \
646
        const int i2= 2*i;       \
647
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
648
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
649
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
650
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
651

    
652
#define YSCALE_YUV_2_GRAY16_2_C   \
653
    for (i=0; i<(dstW>>1); i++){ \
654
        const int i2= 2*i;       \
655
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
656
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
657

    
658
#define YSCALE_YUV_2_RGB2_C(type) \
659
    YSCALE_YUV_2_PACKED2_C\
660
    type *r, *b, *g;\
661
    r = (type *)c->table_rV[V];\
662
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
663
    b = (type *)c->table_bU[U];\
664

    
665
#define YSCALE_YUV_2_PACKED1_C \
666
    for (i=0; i<(dstW>>1); i++){\
667
        const int i2= 2*i;\
668
        int Y1= buf0[i2  ]>>7;\
669
        int Y2= buf0[i2+1]>>7;\
670
        int U= (uvbuf1[i     ])>>7;\
671
        int V= (uvbuf1[i+VOFW])>>7;\
672

    
673
#define YSCALE_YUV_2_GRAY16_1_C \
674
    for (i=0; i<(dstW>>1); i++){\
675
        const int i2= 2*i;\
676
        int Y1= buf0[i2  ]<<1;\
677
        int Y2= buf0[i2+1]<<1;\
678

    
679
#define YSCALE_YUV_2_RGB1_C(type) \
680
    YSCALE_YUV_2_PACKED1_C\
681
    type *r, *b, *g;\
682
    r = (type *)c->table_rV[V];\
683
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
684
    b = (type *)c->table_bU[U];\
685

    
686
#define YSCALE_YUV_2_PACKED1B_C \
687
    for (i=0; i<(dstW>>1); i++){\
688
        const int i2= 2*i;\
689
        int Y1= buf0[i2  ]>>7;\
690
        int Y2= buf0[i2+1]>>7;\
691
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
692
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
693

    
694
#define YSCALE_YUV_2_RGB1B_C(type) \
695
    YSCALE_YUV_2_PACKED1B_C\
696
    type *r, *b, *g;\
697
    r = (type *)c->table_rV[V];\
698
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
699
    b = (type *)c->table_bU[U];\
700

    
701
#define YSCALE_YUV_2_MONO2_C \
702
    const uint8_t * const d128=dither_8x8_220[y&7];\
703
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
704
    for (i=0; i<dstW-7; i+=8){\
705
        int acc;\
706
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
707
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
708
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
709
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
710
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
711
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
712
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
713
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
714
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
715
        dest++;\
716
    }\
717

    
718

    
719
#define YSCALE_YUV_2_MONOX_C \
720
    const uint8_t * const d128=dither_8x8_220[y&7];\
721
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
722
    int acc=0;\
723
    for (i=0; i<dstW-1; i+=2){\
724
        int j;\
725
        int Y1=1<<18;\
726
        int Y2=1<<18;\
727
\
728
        for (j=0; j<lumFilterSize; j++)\
729
        {\
730
            Y1 += lumSrc[j][i] * lumFilter[j];\
731
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
732
        }\
733
        Y1>>=19;\
734
        Y2>>=19;\
735
        if ((Y1|Y2)&256)\
736
        {\
737
            if (Y1>255)   Y1=255;\
738
            else if (Y1<0)Y1=0;\
739
            if (Y2>255)   Y2=255;\
740
            else if (Y2<0)Y2=0;\
741
        }\
742
        acc+= acc + g[Y1+d128[(i+0)&7]];\
743
        acc+= acc + g[Y2+d128[(i+1)&7]];\
744
        if ((i&7)==6){\
745
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
746
            dest++;\
747
        }\
748
    }
749

    
750

    
751
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
752
    switch(c->dstFormat)\
753
    {\
754
    case PIX_FMT_RGB32:\
755
    case PIX_FMT_BGR32:\
756
    case PIX_FMT_RGB32_1:\
757
    case PIX_FMT_BGR32_1:\
758
        func(uint32_t)\
759
            ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
760
            ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
761
        }                \
762
        break;\
763
    case PIX_FMT_RGB24:\
764
        func(uint8_t)\
765
            ((uint8_t*)dest)[0]= r[Y1];\
766
            ((uint8_t*)dest)[1]= g[Y1];\
767
            ((uint8_t*)dest)[2]= b[Y1];\
768
            ((uint8_t*)dest)[3]= r[Y2];\
769
            ((uint8_t*)dest)[4]= g[Y2];\
770
            ((uint8_t*)dest)[5]= b[Y2];\
771
            dest+=6;\
772
        }\
773
        break;\
774
    case PIX_FMT_BGR24:\
775
        func(uint8_t)\
776
            ((uint8_t*)dest)[0]= b[Y1];\
777
            ((uint8_t*)dest)[1]= g[Y1];\
778
            ((uint8_t*)dest)[2]= r[Y1];\
779
            ((uint8_t*)dest)[3]= b[Y2];\
780
            ((uint8_t*)dest)[4]= g[Y2];\
781
            ((uint8_t*)dest)[5]= r[Y2];\
782
            dest+=6;\
783
        }\
784
        break;\
785
    case PIX_FMT_RGB565:\
786
    case PIX_FMT_BGR565:\
787
        {\
788
            const int dr1= dither_2x2_8[y&1    ][0];\
789
            const int dg1= dither_2x2_4[y&1    ][0];\
790
            const int db1= dither_2x2_8[(y&1)^1][0];\
791
            const int dr2= dither_2x2_8[y&1    ][1];\
792
            const int dg2= dither_2x2_4[y&1    ][1];\
793
            const int db2= dither_2x2_8[(y&1)^1][1];\
794
            func(uint16_t)\
795
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
796
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
797
            }\
798
        }\
799
        break;\
800
    case PIX_FMT_RGB555:\
801
    case PIX_FMT_BGR555:\
802
        {\
803
            const int dr1= dither_2x2_8[y&1    ][0];\
804
            const int dg1= dither_2x2_8[y&1    ][1];\
805
            const int db1= dither_2x2_8[(y&1)^1][0];\
806
            const int dr2= dither_2x2_8[y&1    ][1];\
807
            const int dg2= dither_2x2_8[y&1    ][0];\
808
            const int db2= dither_2x2_8[(y&1)^1][1];\
809
            func(uint16_t)\
810
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
811
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
812
            }\
813
        }\
814
        break;\
815
    case PIX_FMT_RGB8:\
816
    case PIX_FMT_BGR8:\
817
        {\
818
            const uint8_t * const d64= dither_8x8_73[y&7];\
819
            const uint8_t * const d32= dither_8x8_32[y&7];\
820
            func(uint8_t)\
821
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
822
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
823
            }\
824
        }\
825
        break;\
826
    case PIX_FMT_RGB4:\
827
    case PIX_FMT_BGR4:\
828
        {\
829
            const uint8_t * const d64= dither_8x8_73 [y&7];\
830
            const uint8_t * const d128=dither_8x8_220[y&7];\
831
            func(uint8_t)\
832
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
833
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
834
            }\
835
        }\
836
        break;\
837
    case PIX_FMT_RGB4_BYTE:\
838
    case PIX_FMT_BGR4_BYTE:\
839
        {\
840
            const uint8_t * const d64= dither_8x8_73 [y&7];\
841
            const uint8_t * const d128=dither_8x8_220[y&7];\
842
            func(uint8_t)\
843
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
844
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
845
            }\
846
        }\
847
        break;\
848
    case PIX_FMT_MONOBLACK:\
849
    case PIX_FMT_MONOWHITE:\
850
        {\
851
            func_monoblack\
852
        }\
853
        break;\
854
    case PIX_FMT_YUYV422:\
855
        func2\
856
            ((uint8_t*)dest)[2*i2+0]= Y1;\
857
            ((uint8_t*)dest)[2*i2+1]= U;\
858
            ((uint8_t*)dest)[2*i2+2]= Y2;\
859
            ((uint8_t*)dest)[2*i2+3]= V;\
860
        }                \
861
        break;\
862
    case PIX_FMT_UYVY422:\
863
        func2\
864
            ((uint8_t*)dest)[2*i2+0]= U;\
865
            ((uint8_t*)dest)[2*i2+1]= Y1;\
866
            ((uint8_t*)dest)[2*i2+2]= V;\
867
            ((uint8_t*)dest)[2*i2+3]= Y2;\
868
        }                \
869
        break;\
870
    case PIX_FMT_GRAY16BE:\
871
        func_g16\
872
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
873
            ((uint8_t*)dest)[2*i2+1]= Y1;\
874
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
875
            ((uint8_t*)dest)[2*i2+3]= Y2;\
876
        }                \
877
        break;\
878
    case PIX_FMT_GRAY16LE:\
879
        func_g16\
880
            ((uint8_t*)dest)[2*i2+0]= Y1;\
881
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
882
            ((uint8_t*)dest)[2*i2+2]= Y2;\
883
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
884
        }                \
885
        break;\
886
    }\
887

    
888

    
889
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
890
                                  int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
891
                                  uint8_t *dest, int dstW, int y)
892
{
893
    int i;
894
    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
895
}
896

    
897
static inline void yuv2rgbXinC_full(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
898
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
899
                                    uint8_t *dest, int dstW, int y)
900
{
901
    int i;
902
    int step= fmt_depth(c->dstFormat)/8;
903
    int aidx= 3;
904

    
905
    switch(c->dstFormat){
906
    case PIX_FMT_ARGB:
907
        dest++;
908
        aidx= 0;
909
    case PIX_FMT_RGB24:
910
        aidx--;
911
    case PIX_FMT_RGBA:
912
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
913
            dest[aidx]= 0;
914
            dest[0]= R>>22;
915
            dest[1]= G>>22;
916
            dest[2]= B>>22;
917
            dest+= step;
918
        }
919
        break;
920
    case PIX_FMT_ABGR:
921
        dest++;
922
        aidx= 0;
923
    case PIX_FMT_BGR24:
924
        aidx--;
925
    case PIX_FMT_BGRA:
926
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
927
            dest[aidx]= 0;
928
            dest[0]= B>>22;
929
            dest[1]= G>>22;
930
            dest[2]= R>>22;
931
            dest+= step;
932
        }
933
        break;
934
    default:
935
        assert(0);
936
    }
937
}
938

    
939
//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
940
//Plain C versions
941
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)
942
#define COMPILE_C
943
#endif
944

    
945
#ifdef ARCH_POWERPC
946
#if (defined (HAVE_ALTIVEC) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
947
#define COMPILE_ALTIVEC
948
#endif //HAVE_ALTIVEC
949
#endif //ARCH_POWERPC
950

    
951
#if defined(ARCH_X86)
952

    
953
#if ((defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
954
#define COMPILE_MMX
955
#endif
956

    
957
#if (defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
958
#define COMPILE_MMX2
959
#endif
960

    
961
#if ((defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
962
#define COMPILE_3DNOW
963
#endif
964
#endif //ARCH_X86 || ARCH_X86_64
965

    
966
#undef HAVE_MMX
967
#undef HAVE_MMX2
968
#undef HAVE_3DNOW
969

    
970
#ifdef COMPILE_C
971
#undef HAVE_MMX
972
#undef HAVE_MMX2
973
#undef HAVE_3DNOW
974
#undef HAVE_ALTIVEC
975
#define RENAME(a) a ## _C
976
#include "swscale_template.c"
977
#endif
978

    
979
#ifdef COMPILE_ALTIVEC
980
#undef RENAME
981
#define HAVE_ALTIVEC
982
#define RENAME(a) a ## _altivec
983
#include "swscale_template.c"
984
#endif
985

    
986
#if defined(ARCH_X86)
987

    
988
//X86 versions
989
/*
990
#undef RENAME
991
#undef HAVE_MMX
992
#undef HAVE_MMX2
993
#undef HAVE_3DNOW
994
#define ARCH_X86
995
#define RENAME(a) a ## _X86
996
#include "swscale_template.c"
997
*/
998
//MMX versions
999
#ifdef COMPILE_MMX
1000
#undef RENAME
1001
#define HAVE_MMX
1002
#undef HAVE_MMX2
1003
#undef HAVE_3DNOW
1004
#define RENAME(a) a ## _MMX
1005
#include "swscale_template.c"
1006
#endif
1007

    
1008
//MMX2 versions
1009
#ifdef COMPILE_MMX2
1010
#undef RENAME
1011
#define HAVE_MMX
1012
#define HAVE_MMX2
1013
#undef HAVE_3DNOW
1014
#define RENAME(a) a ## _MMX2
1015
#include "swscale_template.c"
1016
#endif
1017

    
1018
//3DNOW versions
1019
#ifdef COMPILE_3DNOW
1020
#undef RENAME
1021
#define HAVE_MMX
1022
#undef HAVE_MMX2
1023
#define HAVE_3DNOW
1024
#define RENAME(a) a ## _3DNow
1025
#include "swscale_template.c"
1026
#endif
1027

    
1028
#endif //ARCH_X86 || ARCH_X86_64
1029

    
1030
// minor note: the HAVE_xyz is messed up after that line so don't use it
1031

    
1032
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1033
{
1034
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1035
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
1036
    else                return getSplineCoeff(        0.0,
1037
                                             b+ 2.0*c + 3.0*d,
1038
                                                    c + 3.0*d,
1039
                                            -b- 3.0*c - 6.0*d,
1040
                                            dist-1.0);
1041
}
1042

    
1043
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1044
                             int srcW, int dstW, int filterAlign, int one, int flags,
1045
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1046
{
1047
    int i;
1048
    int filterSize;
1049
    int filter2Size;
1050
    int minFilterSize;
1051
    int64_t *filter=NULL;
1052
    int64_t *filter2=NULL;
1053
    const int64_t fone= 1LL<<54;
1054
    int ret= -1;
1055
#if defined(ARCH_X86)
1056
    if (flags & SWS_CPU_CAPS_MMX)
1057
        asm volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1058
#endif
1059

    
1060
    // Note the +1 is for the MMXscaler which reads over the end
1061
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1062

    
1063
    if (FFABS(xInc - 0x10000) <10) // unscaled
1064
    {
1065
        int i;
1066
        filterSize= 1;
1067
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1068

    
1069
        for (i=0; i<dstW; i++)
1070
        {
1071
            filter[i*filterSize]= fone;
1072
            (*filterPos)[i]=i;
1073
        }
1074

    
1075
    }
1076
    else if (flags&SWS_POINT) // lame looking point sampling mode
1077
    {
1078
        int i;
1079
        int xDstInSrc;
1080
        filterSize= 1;
1081
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1082

    
1083
        xDstInSrc= xInc/2 - 0x8000;
1084
        for (i=0; i<dstW; i++)
1085
        {
1086
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1087

    
1088
            (*filterPos)[i]= xx;
1089
            filter[i]= fone;
1090
            xDstInSrc+= xInc;
1091
        }
1092
    }
1093
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1094
    {
1095
        int i;
1096
        int xDstInSrc;
1097
        if      (flags&SWS_BICUBIC) filterSize= 4;
1098
        else if (flags&SWS_X      ) filterSize= 4;
1099
        else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA
1100
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1101

    
1102
        xDstInSrc= xInc/2 - 0x8000;
1103
        for (i=0; i<dstW; i++)
1104
        {
1105
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1106
            int j;
1107

    
1108
            (*filterPos)[i]= xx;
1109
                //Bilinear upscale / linear interpolate / Area averaging
1110
                for (j=0; j<filterSize; j++)
1111
                {
1112
                    int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1113
                    if (coeff<0) coeff=0;
1114
                    filter[i*filterSize + j]= coeff;
1115
                    xx++;
1116
                }
1117
            xDstInSrc+= xInc;
1118
        }
1119
    }
1120
    else
1121
    {
1122
        int xDstInSrc;
1123
        int sizeFactor;
1124

    
1125
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1126
        else if (flags&SWS_X)            sizeFactor=  8;
1127
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1128
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1129
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1130
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1131
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1132
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1133
        else {
1134
            sizeFactor= 0; //GCC warning killer
1135
            assert(0);
1136
        }
1137

    
1138
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1139
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1140

    
1141
        if (filterSize > srcW-2) filterSize=srcW-2;
1142

    
1143
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1144

    
1145
        xDstInSrc= xInc - 0x10000;
1146
        for (i=0; i<dstW; i++)
1147
        {
1148
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1149
            int j;
1150
            (*filterPos)[i]= xx;
1151
            for (j=0; j<filterSize; j++)
1152
            {
1153
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1154
                double floatd;
1155
                int64_t coeff;
1156

    
1157
                if (xInc > 1<<16)
1158
                    d= d*dstW/srcW;
1159
                floatd= d * (1.0/(1<<30));
1160

    
1161
                if (flags & SWS_BICUBIC)
1162
                {
1163
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1164
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1165
                    int64_t dd = ( d*d)>>30;
1166
                    int64_t ddd= (dd*d)>>30;
1167

    
1168
                    if      (d < 1LL<<30)
1169
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1170
                    else if (d < 1LL<<31)
1171
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1172
                    else
1173
                        coeff=0.0;
1174
                    coeff *= fone>>(30+24);
1175
                }
1176
/*                else if (flags & SWS_X)
1177
                {
1178
                    double p= param ? param*0.01 : 0.3;
1179
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1180
                    coeff*= pow(2.0, - p*d*d);
1181
                }*/
1182
                else if (flags & SWS_X)
1183
                {
1184
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1185
                    double c;
1186

    
1187
                    if (floatd<1.0)
1188
                        c = cos(floatd*PI);
1189
                    else
1190
                        c=-1.0;
1191
                    if (c<0.0)      c= -pow(-c, A);
1192
                    else            c=  pow( c, A);
1193
                    coeff= (c*0.5 + 0.5)*fone;
1194
                }
1195
                else if (flags & SWS_AREA)
1196
                {
1197
                    int64_t d2= d - (1<<29);
1198
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1199
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1200
                    else coeff=0.0;
1201
                    coeff *= fone>>(30+16);
1202
                }
1203
                else if (flags & SWS_GAUSS)
1204
                {
1205
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1206
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1207
                }
1208
                else if (flags & SWS_SINC)
1209
                {
1210
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1211
                }
1212
                else if (flags & SWS_LANCZOS)
1213
                {
1214
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1215
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1216
                    if (floatd>p) coeff=0;
1217
                }
1218
                else if (flags & SWS_BILINEAR)
1219
                {
1220
                    coeff= (1<<30) - d;
1221
                    if (coeff<0) coeff=0;
1222
                    coeff *= fone >> 30;
1223
                }
1224
                else if (flags & SWS_SPLINE)
1225
                {
1226
                    double p=-2.196152422706632;
1227
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1228
                }
1229
                else {
1230
                    coeff= 0.0; //GCC warning killer
1231
                    assert(0);
1232
                }
1233

    
1234
                filter[i*filterSize + j]= coeff;
1235
                xx++;
1236
            }
1237
            xDstInSrc+= 2*xInc;
1238
        }
1239
    }
1240

    
1241
    /* apply src & dst Filter to filter -> filter2
1242
       av_free(filter);
1243
    */
1244
    assert(filterSize>0);
1245
    filter2Size= filterSize;
1246
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1247
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1248
    assert(filter2Size>0);
1249
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1250

    
1251
    for (i=0; i<dstW; i++)
1252
    {
1253
        int j, k;
1254

    
1255
        if(srcFilter){
1256
            for (k=0; k<srcFilter->length; k++){
1257
                for (j=0; j<filterSize; j++)
1258
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1259
            }
1260
        }else{
1261
            for (j=0; j<filterSize; j++)
1262
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1263
        }
1264
        //FIXME dstFilter
1265

    
1266
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1267
    }
1268
    av_freep(&filter);
1269

    
1270
    /* try to reduce the filter-size (step1 find size and shift left) */
1271
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1272
    minFilterSize= 0;
1273
    for (i=dstW-1; i>=0; i--)
1274
    {
1275
        int min= filter2Size;
1276
        int j;
1277
        int64_t cutOff=0.0;
1278

    
1279
        /* get rid off near zero elements on the left by shifting left */
1280
        for (j=0; j<filter2Size; j++)
1281
        {
1282
            int k;
1283
            cutOff += FFABS(filter2[i*filter2Size]);
1284

    
1285
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1286

    
1287
            /* preserve monotonicity because the core can't handle the filter otherwise */
1288
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1289

    
1290
            // Move filter coeffs left
1291
            for (k=1; k<filter2Size; k++)
1292
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1293
            filter2[i*filter2Size + k - 1]= 0;
1294
            (*filterPos)[i]++;
1295
        }
1296

    
1297
        cutOff=0;
1298
        /* count near zeros on the right */
1299
        for (j=filter2Size-1; j>0; j--)
1300
        {
1301
            cutOff += FFABS(filter2[i*filter2Size + j]);
1302

    
1303
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1304
            min--;
1305
        }
1306

    
1307
        if (min>minFilterSize) minFilterSize= min;
1308
    }
1309

    
1310
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1311
        // we can handle the special case 4,
1312
        // so we don't want to go to the full 8
1313
        if (minFilterSize < 5)
1314
            filterAlign = 4;
1315

    
1316
        // we really don't want to waste our time
1317
        // doing useless computation, so fall-back on
1318
        // the scalar C code for very small filter.
1319
        // vectorizing is worth it only if you have
1320
        // decent-sized vector.
1321
        if (minFilterSize < 3)
1322
            filterAlign = 1;
1323
    }
1324

    
1325
    if (flags & SWS_CPU_CAPS_MMX) {
1326
        // special case for unscaled vertical filtering
1327
        if (minFilterSize == 1 && filterAlign == 2)
1328
            filterAlign= 1;
1329
    }
1330

    
1331
    assert(minFilterSize > 0);
1332
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1333
    assert(filterSize > 0);
1334
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1335
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1336
        goto error;
1337
    *outFilterSize= filterSize;
1338

    
1339
    if (flags&SWS_PRINT_INFO)
1340
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1341
    /* try to reduce the filter-size (step2 reduce it) */
1342
    for (i=0; i<dstW; i++)
1343
    {
1344
        int j;
1345

    
1346
        for (j=0; j<filterSize; j++)
1347
        {
1348
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1349
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1350
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1351
                filter[i*filterSize + j]= 0;
1352
        }
1353
    }
1354

    
1355

    
1356
    //FIXME try to align filterpos if possible
1357

    
1358
    //fix borders
1359
    for (i=0; i<dstW; i++)
1360
    {
1361
        int j;
1362
        if ((*filterPos)[i] < 0)
1363
        {
1364
            // Move filter coeffs left to compensate for filterPos
1365
            for (j=1; j<filterSize; j++)
1366
            {
1367
                int left= FFMAX(j + (*filterPos)[i], 0);
1368
                filter[i*filterSize + left] += filter[i*filterSize + j];
1369
                filter[i*filterSize + j]=0;
1370
            }
1371
            (*filterPos)[i]= 0;
1372
        }
1373

    
1374
        if ((*filterPos)[i] + filterSize > srcW)
1375
        {
1376
            int shift= (*filterPos)[i] + filterSize - srcW;
1377
            // Move filter coeffs right to compensate for filterPos
1378
            for (j=filterSize-2; j>=0; j--)
1379
            {
1380
                int right= FFMIN(j + shift, filterSize-1);
1381
                filter[i*filterSize +right] += filter[i*filterSize +j];
1382
                filter[i*filterSize +j]=0;
1383
            }
1384
            (*filterPos)[i]= srcW - filterSize;
1385
        }
1386
    }
1387

    
1388
    // Note the +1 is for the MMXscaler which reads over the end
1389
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1390
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1391

    
1392
    /* Normalize & Store in outFilter */
1393
    for (i=0; i<dstW; i++)
1394
    {
1395
        int j;
1396
        int64_t error=0;
1397
        int64_t sum=0;
1398

    
1399
        for (j=0; j<filterSize; j++)
1400
        {
1401
            sum+= filter[i*filterSize + j];
1402
        }
1403
        sum= (sum + one/2)/ one;
1404
        for (j=0; j<*outFilterSize; j++)
1405
        {
1406
            int64_t v= filter[i*filterSize + j] + error;
1407
            int intV= ROUNDED_DIV(v, sum);
1408
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1409
            error= v - intV*sum;
1410
        }
1411
    }
1412

    
1413
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1414
    for (i=0; i<*outFilterSize; i++)
1415
    {
1416
        int j= dstW*(*outFilterSize);
1417
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1418
    }
1419

    
1420
    ret=0;
1421
error:
1422
    av_free(filter);
1423
    av_free(filter2);
1424
    return ret;
1425
}
1426

    
1427
#ifdef COMPILE_MMX2
1428
static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1429
{
1430
    uint8_t *fragmentA;
1431
    long imm8OfPShufW1A;
1432
    long imm8OfPShufW2A;
1433
    long fragmentLengthA;
1434
    uint8_t *fragmentB;
1435
    long imm8OfPShufW1B;
1436
    long imm8OfPShufW2B;
1437
    long fragmentLengthB;
1438
    int fragmentPos;
1439

    
1440
    int xpos, i;
1441

    
1442
    // create an optimized horizontal scaling routine
1443

    
1444
    //code fragment
1445

    
1446
    asm volatile(
1447
        "jmp                         9f                 \n\t"
1448
    // Begin
1449
        "0:                                             \n\t"
1450
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1451
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1452
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1453
        "punpcklbw                %%mm7, %%mm1          \n\t"
1454
        "punpcklbw                %%mm7, %%mm0          \n\t"
1455
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1456
        "1:                                             \n\t"
1457
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1458
        "2:                                             \n\t"
1459
        "psubw                    %%mm1, %%mm0          \n\t"
1460
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1461
        "pmullw                   %%mm3, %%mm0          \n\t"
1462
        "psllw                       $7, %%mm1          \n\t"
1463
        "paddw                    %%mm1, %%mm0          \n\t"
1464

    
1465
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1466

    
1467
        "add                         $8, %%"REG_a"      \n\t"
1468
    // End
1469
        "9:                                             \n\t"
1470
//        "int $3                                         \n\t"
1471
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1472
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1473
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1474
        "dec                         %1                 \n\t"
1475
        "dec                         %2                 \n\t"
1476
        "sub                         %0, %1             \n\t"
1477
        "sub                         %0, %2             \n\t"
1478
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1479
        "sub                         %0, %3             \n\t"
1480

    
1481

    
1482
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1483
        "=r" (fragmentLengthA)
1484
    );
1485

    
1486
    asm volatile(
1487
        "jmp                         9f                 \n\t"
1488
    // Begin
1489
        "0:                                             \n\t"
1490
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1491
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1492
        "punpcklbw                %%mm7, %%mm0          \n\t"
1493
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1494
        "1:                                             \n\t"
1495
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1496
        "2:                                             \n\t"
1497
        "psubw                    %%mm1, %%mm0          \n\t"
1498
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1499
        "pmullw                   %%mm3, %%mm0          \n\t"
1500
        "psllw                       $7, %%mm1          \n\t"
1501
        "paddw                    %%mm1, %%mm0          \n\t"
1502

    
1503
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1504

    
1505
        "add                         $8, %%"REG_a"      \n\t"
1506
    // End
1507
        "9:                                             \n\t"
1508
//        "int                       $3                   \n\t"
1509
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1510
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1511
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1512
        "dec                         %1                 \n\t"
1513
        "dec                         %2                 \n\t"
1514
        "sub                         %0, %1             \n\t"
1515
        "sub                         %0, %2             \n\t"
1516
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1517
        "sub                         %0, %3             \n\t"
1518

    
1519

    
1520
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1521
        "=r" (fragmentLengthB)
1522
    );
1523

    
1524
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1525
    fragmentPos=0;
1526

    
1527
    for (i=0; i<dstW/numSplits; i++)
1528
    {
1529
        int xx=xpos>>16;
1530

    
1531
        if ((i&3) == 0)
1532
        {
1533
            int a=0;
1534
            int b=((xpos+xInc)>>16) - xx;
1535
            int c=((xpos+xInc*2)>>16) - xx;
1536
            int d=((xpos+xInc*3)>>16) - xx;
1537

    
1538
            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1539
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1540
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1541
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1542
            filterPos[i/2]= xx;
1543

    
1544
            if (d+1<4)
1545
            {
1546
                int maxShift= 3-(d+1);
1547
                int shift=0;
1548

    
1549
                memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1550

    
1551
                funnyCode[fragmentPos + imm8OfPShufW1B]=
1552
                    (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1553
                funnyCode[fragmentPos + imm8OfPShufW2B]=
1554
                    a | (b<<2) | (c<<4) | (d<<6);
1555

    
1556
                if (i+3>=dstW) shift=maxShift; //avoid overread
1557
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1558

    
1559
                if (shift && i>=shift)
1560
                {
1561
                    funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1562
                    funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1563
                    filterPos[i/2]-=shift;
1564
                }
1565

    
1566
                fragmentPos+= fragmentLengthB;
1567
            }
1568
            else
1569
            {
1570
                int maxShift= 3-d;
1571
                int shift=0;
1572

    
1573
                memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
1574

    
1575
                funnyCode[fragmentPos + imm8OfPShufW1A]=
1576
                funnyCode[fragmentPos + imm8OfPShufW2A]=
1577
                    a | (b<<2) | (c<<4) | (d<<6);
1578

    
1579
                if (i+4>=dstW) shift=maxShift; //avoid overread
1580
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
1581

    
1582
                if (shift && i>=shift)
1583
                {
1584
                    funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1585
                    funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1586
                    filterPos[i/2]-=shift;
1587
                }
1588

    
1589
                fragmentPos+= fragmentLengthA;
1590
            }
1591

    
1592
            funnyCode[fragmentPos]= RET;
1593
        }
1594
        xpos+=xInc;
1595
    }
1596
    filterPos[i/2]= xpos>>16; // needed to jump to the next part
1597
}
1598
#endif /* COMPILE_MMX2 */
1599

    
1600
static void globalInit(void){
1601
    // generating tables:
1602
    int i;
1603
    for (i=0; i<768; i++){
1604
        int c= av_clip_uint8(i-256);
1605
        clip_table[i]=c;
1606
    }
1607
}
1608

    
1609
static SwsFunc getSwsFunc(int flags){
1610

    
1611
#if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)
1612
#if defined(ARCH_X86)
1613
    // ordered per speed fastest first
1614
    if (flags & SWS_CPU_CAPS_MMX2)
1615
        return swScale_MMX2;
1616
    else if (flags & SWS_CPU_CAPS_3DNOW)
1617
        return swScale_3DNow;
1618
    else if (flags & SWS_CPU_CAPS_MMX)
1619
        return swScale_MMX;
1620
    else
1621
        return swScale_C;
1622

    
1623
#else
1624
#ifdef ARCH_POWERPC
1625
    if (flags & SWS_CPU_CAPS_ALTIVEC)
1626
        return swScale_altivec;
1627
    else
1628
        return swScale_C;
1629
#endif
1630
    return swScale_C;
1631
#endif /* defined(ARCH_X86) */
1632
#else //RUNTIME_CPUDETECT
1633
#ifdef HAVE_MMX2
1634
    return swScale_MMX2;
1635
#elif defined (HAVE_3DNOW)
1636
    return swScale_3DNow;
1637
#elif defined (HAVE_MMX)
1638
    return swScale_MMX;
1639
#elif defined (HAVE_ALTIVEC)
1640
    return swScale_altivec;
1641
#else
1642
    return swScale_C;
1643
#endif
1644
#endif //!RUNTIME_CPUDETECT
1645
}
1646

    
1647
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1648
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1649
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1650
    /* Copy Y plane */
1651
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1652
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1653
    else
1654
    {
1655
        int i;
1656
        uint8_t *srcPtr= src[0];
1657
        uint8_t *dstPtr= dst;
1658
        for (i=0; i<srcSliceH; i++)
1659
        {
1660
            memcpy(dstPtr, srcPtr, c->srcW);
1661
            srcPtr+= srcStride[0];
1662
            dstPtr+= dstStride[0];
1663
        }
1664
    }
1665
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1666
    if (c->dstFormat == PIX_FMT_NV12)
1667
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1668
    else
1669
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1670

    
1671
    return srcSliceH;
1672
}
1673

    
1674
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1675
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1676
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1677

    
1678
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1679

    
1680
    return srcSliceH;
1681
}
1682

    
1683
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1684
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1685
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1686

    
1687
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1688

    
1689
    return srcSliceH;
1690
}
1691

    
1692
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1693
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1694
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1695

    
1696
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1697

    
1698
    return srcSliceH;
1699
}
1700

    
1701
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1702
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1703
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1704

    
1705
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1706

    
1707
    return srcSliceH;
1708
}
1709

    
1710
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1711
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1712
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
1713
    const enum PixelFormat srcFormat= c->srcFormat;
1714
    const enum PixelFormat dstFormat= c->dstFormat;
1715
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
1716
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
1717
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1718
    const int dstId= fmt_depth(dstFormat) >> 2;
1719
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1720

    
1721
    /* BGR -> BGR */
1722
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
1723
       || (isRGB(srcFormat) && isRGB(dstFormat))){
1724
        switch(srcId | (dstId<<4)){
1725
        case 0x34: conv= rgb16to15; break;
1726
        case 0x36: conv= rgb24to15; break;
1727
        case 0x38: conv= rgb32to15; break;
1728
        case 0x43: conv= rgb15to16; break;
1729
        case 0x46: conv= rgb24to16; break;
1730
        case 0x48: conv= rgb32to16; break;
1731
        case 0x63: conv= rgb15to24; break;
1732
        case 0x64: conv= rgb16to24; break;
1733
        case 0x68: conv= rgb32to24; break;
1734
        case 0x83: conv= rgb15to32; break;
1735
        case 0x84: conv= rgb16to32; break;
1736
        case 0x86: conv= rgb24to32; break;
1737
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1738
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1739
        }
1740
    }else if (  (isBGR(srcFormat) && isRGB(dstFormat))
1741
             || (isRGB(srcFormat) && isBGR(dstFormat))){
1742
        switch(srcId | (dstId<<4)){
1743
        case 0x33: conv= rgb15tobgr15; break;
1744
        case 0x34: conv= rgb16tobgr15; break;
1745
        case 0x36: conv= rgb24tobgr15; break;
1746
        case 0x38: conv= rgb32tobgr15; break;
1747
        case 0x43: conv= rgb15tobgr16; break;
1748
        case 0x44: conv= rgb16tobgr16; break;
1749
        case 0x46: conv= rgb24tobgr16; break;
1750
        case 0x48: conv= rgb32tobgr16; break;
1751
        case 0x63: conv= rgb15tobgr24; break;
1752
        case 0x64: conv= rgb16tobgr24; break;
1753
        case 0x66: conv= rgb24tobgr24; break;
1754
        case 0x68: conv= rgb32tobgr24; break;
1755
        case 0x83: conv= rgb15tobgr32; break;
1756
        case 0x84: conv= rgb16tobgr32; break;
1757
        case 0x86: conv= rgb24tobgr32; break;
1758
        case 0x88: conv= rgb32tobgr32; break;
1759
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1760
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1761
        }
1762
    }else{
1763
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1764
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1765
    }
1766

    
1767
    if(conv)
1768
    {
1769
        uint8_t *srcPtr= src[0];
1770
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
1771
            srcPtr += ALT32_CORR;
1772

    
1773
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1774
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1775
        else
1776
        {
1777
            int i;
1778
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1779

    
1780
            for (i=0; i<srcSliceH; i++)
1781
            {
1782
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
1783
                srcPtr+= srcStride[0];
1784
                dstPtr+= dstStride[0];
1785
            }
1786
        }
1787
    }
1788
    return srcSliceH;
1789
}
1790

    
1791
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1792
                              int srcSliceH, uint8_t* dst[], int dstStride[]){
1793

    
1794
    rgb24toyv12(
1795
        src[0],
1796
        dst[0]+ srcSliceY    *dstStride[0],
1797
        dst[1]+(srcSliceY>>1)*dstStride[1],
1798
        dst[2]+(srcSliceY>>1)*dstStride[2],
1799
        c->srcW, srcSliceH,
1800
        dstStride[0], dstStride[1], srcStride[0]);
1801
    return srcSliceH;
1802
}
1803

    
1804
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1805
                             int srcSliceH, uint8_t* dst[], int dstStride[]){
1806
    int i;
1807

    
1808
    /* copy Y */
1809
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1810
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1811
    else{
1812
        uint8_t *srcPtr= src[0];
1813
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1814

    
1815
        for (i=0; i<srcSliceH; i++)
1816
        {
1817
            memcpy(dstPtr, srcPtr, c->srcW);
1818
            srcPtr+= srcStride[0];
1819
            dstPtr+= dstStride[0];
1820
        }
1821
    }
1822

    
1823
    if (c->dstFormat==PIX_FMT_YUV420P){
1824
        planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
1825
        planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
1826
    }else{
1827
        planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
1828
        planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
1829
    }
1830
    return srcSliceH;
1831
}
1832

    
1833
/* unscaled copy like stuff (assumes nearly identical formats) */
1834
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1835
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1836
{
1837
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1838
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1839
    else
1840
    {
1841
        int i;
1842
        uint8_t *srcPtr= src[0];
1843
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1844
        int length=0;
1845

    
1846
        /* universal length finder */
1847
        while(length+c->srcW <= FFABS(dstStride[0])
1848
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1849
        assert(length!=0);
1850

    
1851
        for (i=0; i<srcSliceH; i++)
1852
        {
1853
            memcpy(dstPtr, srcPtr, length);
1854
            srcPtr+= srcStride[0];
1855
            dstPtr+= dstStride[0];
1856
        }
1857
    }
1858
    return srcSliceH;
1859
}
1860

    
1861
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1862
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1863
{
1864
    int plane;
1865
    for (plane=0; plane<3; plane++)
1866
    {
1867
        int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1868
        int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1869
        int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1870

    
1871
        if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
1872
        {
1873
            if (!isGray(c->dstFormat))
1874
                memset(dst[plane], 128, dstStride[plane]*height);
1875
        }
1876
        else
1877
        {
1878
            if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1879
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1880
            else
1881
            {
1882
                int i;
1883
                uint8_t *srcPtr= src[plane];
1884
                uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1885
                for (i=0; i<height; i++)
1886
                {
1887
                    memcpy(dstPtr, srcPtr, length);
1888
                    srcPtr+= srcStride[plane];
1889
                    dstPtr+= dstStride[plane];
1890
                }
1891
            }
1892
        }
1893
    }
1894
    return srcSliceH;
1895
}
1896

    
1897
static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1898
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
1899

    
1900
    int length= c->srcW;
1901
    int y=      srcSliceY;
1902
    int height= srcSliceH;
1903
    int i, j;
1904
    uint8_t *srcPtr= src[0];
1905
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1906

    
1907
    if (!isGray(c->dstFormat)){
1908
        int height= -((-srcSliceH)>>c->chrDstVSubSample);
1909
        memset(dst[1], 128, dstStride[1]*height);
1910
        memset(dst[2], 128, dstStride[2]*height);
1911
    }
1912
    if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;
1913
    for (i=0; i<height; i++)
1914
    {
1915
        for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1916
        srcPtr+= srcStride[0];
1917
        dstPtr+= dstStride[0];
1918
    }
1919
    return srcSliceH;
1920
}
1921

    
1922
static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1923
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
1924

    
1925
    int length= c->srcW;
1926
    int y=      srcSliceY;
1927
    int height= srcSliceH;
1928
    int i, j;
1929
    uint8_t *srcPtr= src[0];
1930
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1931
    for (i=0; i<height; i++)
1932
    {
1933
        for (j=0; j<length; j++)
1934
        {
1935
            dstPtr[j<<1] = srcPtr[j];
1936
            dstPtr[(j<<1)+1] = srcPtr[j];
1937
        }
1938
        srcPtr+= srcStride[0];
1939
        dstPtr+= dstStride[0];
1940
    }
1941
    return srcSliceH;
1942
}
1943

    
1944
static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1945
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
1946

    
1947
    int length= c->srcW;
1948
    int y=      srcSliceY;
1949
    int height= srcSliceH;
1950
    int i, j;
1951
    uint16_t *srcPtr= (uint16_t*)src[0];
1952
    uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);
1953
    for (i=0; i<height; i++)
1954
    {
1955
        for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);
1956
        srcPtr+= srcStride[0]/2;
1957
        dstPtr+= dstStride[0]/2;
1958
    }
1959
    return srcSliceH;
1960
}
1961

    
1962

    
1963
static void getSubSampleFactors(int *h, int *v, int format){
1964
    switch(format){
1965
    case PIX_FMT_UYVY422:
1966
    case PIX_FMT_YUYV422:
1967
        *h=1;
1968
        *v=0;
1969
        break;
1970
    case PIX_FMT_YUV420P:
1971
    case PIX_FMT_YUVA420P:
1972
    case PIX_FMT_GRAY16BE:
1973
    case PIX_FMT_GRAY16LE:
1974
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
1975
    case PIX_FMT_NV12:
1976
    case PIX_FMT_NV21:
1977
        *h=1;
1978
        *v=1;
1979
        break;
1980
    case PIX_FMT_YUV440P:
1981
        *h=0;
1982
        *v=1;
1983
        break;
1984
    case PIX_FMT_YUV410P:
1985
        *h=2;
1986
        *v=2;
1987
        break;
1988
    case PIX_FMT_YUV444P:
1989
        *h=0;
1990
        *v=0;
1991
        break;
1992
    case PIX_FMT_YUV422P:
1993
        *h=1;
1994
        *v=0;
1995
        break;
1996
    case PIX_FMT_YUV411P:
1997
        *h=2;
1998
        *v=0;
1999
        break;
2000
    default:
2001
        *h=0;
2002
        *v=0;
2003
        break;
2004
    }
2005
}
2006

    
2007
static uint16_t roundToInt16(int64_t f){
2008
    int r= (f + (1<<15))>>16;
2009
         if (r<-0x7FFF) return 0x8000;
2010
    else if (r> 0x7FFF) return 0x7FFF;
2011
    else                return r;
2012
}
2013

    
2014
/**
2015
 * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]
2016
 * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235
2017
 * @return -1 if not supported
2018
 */
2019
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2020
    int64_t crv =  inv_table[0];
2021
    int64_t cbu =  inv_table[1];
2022
    int64_t cgu = -inv_table[2];
2023
    int64_t cgv = -inv_table[3];
2024
    int64_t cy  = 1<<16;
2025
    int64_t oy  = 0;
2026

    
2027
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2028
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2029

    
2030
    c->brightness= brightness;
2031
    c->contrast  = contrast;
2032
    c->saturation= saturation;
2033
    c->srcRange  = srcRange;
2034
    c->dstRange  = dstRange;
2035
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return 0;
2036

    
2037
    c->uOffset=   0x0400040004000400LL;
2038
    c->vOffset=   0x0400040004000400LL;
2039

    
2040
    if (!srcRange){
2041
        cy= (cy*255) / 219;
2042
        oy= 16<<16;
2043
    }else{
2044
        crv= (crv*224) / 255;
2045
        cbu= (cbu*224) / 255;
2046
        cgu= (cgu*224) / 255;
2047
        cgv= (cgv*224) / 255;
2048
    }
2049

    
2050
    cy = (cy *contrast             )>>16;
2051
    crv= (crv*contrast * saturation)>>32;
2052
    cbu= (cbu*contrast * saturation)>>32;
2053
    cgu= (cgu*contrast * saturation)>>32;
2054
    cgv= (cgv*contrast * saturation)>>32;
2055

    
2056
    oy -= 256*brightness;
2057

    
2058
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2059
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2060
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2061
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2062
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2063
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2064

    
2065
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2066
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2067
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2068
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2069
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2070
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2071

    
2072
    yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2073
    //FIXME factorize
2074

    
2075
#ifdef COMPILE_ALTIVEC
2076
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2077
        yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
2078
#endif
2079
    return 0;
2080
}
2081

    
2082
/**
2083
 * @return -1 if not supported
2084
 */
2085
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2086
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2087

    
2088
    *inv_table = c->srcColorspaceTable;
2089
    *table     = c->dstColorspaceTable;
2090
    *srcRange  = c->srcRange;
2091
    *dstRange  = c->dstRange;
2092
    *brightness= c->brightness;
2093
    *contrast  = c->contrast;
2094
    *saturation= c->saturation;
2095

    
2096
    return 0;
2097
}
2098

    
2099
static int handle_jpeg(enum PixelFormat *format)
2100
{
2101
    switch (*format) {
2102
        case PIX_FMT_YUVJ420P:
2103
            *format = PIX_FMT_YUV420P;
2104
            return 1;
2105
        case PIX_FMT_YUVJ422P:
2106
            *format = PIX_FMT_YUV422P;
2107
            return 1;
2108
        case PIX_FMT_YUVJ444P:
2109
            *format = PIX_FMT_YUV444P;
2110
            return 1;
2111
        case PIX_FMT_YUVJ440P:
2112
            *format = PIX_FMT_YUV440P;
2113
            return 1;
2114
        default:
2115
            return 0;
2116
    }
2117
}
2118

    
2119
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2120
                           SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){
2121

    
2122
    SwsContext *c;
2123
    int i;
2124
    int usesVFilter, usesHFilter;
2125
    int unscaled, needsDither;
2126
    int srcRange, dstRange;
2127
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2128
#if defined(ARCH_X86)
2129
    if (flags & SWS_CPU_CAPS_MMX)
2130
        asm volatile("emms\n\t"::: "memory");
2131
#endif
2132

    
2133
#if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off
2134
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2135
#ifdef HAVE_MMX2
2136
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2137
#elif defined (HAVE_3DNOW)
2138
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2139
#elif defined (HAVE_MMX)
2140
    flags |= SWS_CPU_CAPS_MMX;
2141
#elif defined (HAVE_ALTIVEC)
2142
    flags |= SWS_CPU_CAPS_ALTIVEC;
2143
#elif defined (ARCH_BFIN)
2144
    flags |= SWS_CPU_CAPS_BFIN;
2145
#endif
2146
#endif /* RUNTIME_CPUDETECT */
2147
    if (clip_table[512] != 255) globalInit();
2148
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2149

    
2150
    unscaled = (srcW == dstW && srcH == dstH);
2151
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2152
        && (fmt_depth(dstFormat))<24
2153
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2154

    
2155
    srcRange = handle_jpeg(&srcFormat);
2156
    dstRange = handle_jpeg(&dstFormat);
2157

    
2158
    if (!isSupportedIn(srcFormat))
2159
    {
2160
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2161
        return NULL;
2162
    }
2163
    if (!isSupportedOut(dstFormat))
2164
    {
2165
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2166
        return NULL;
2167
    }
2168

    
2169
    i= flags & ( SWS_POINT
2170
                |SWS_AREA
2171
                |SWS_BILINEAR
2172
                |SWS_FAST_BILINEAR
2173
                |SWS_BICUBIC
2174
                |SWS_X
2175
                |SWS_GAUSS
2176
                |SWS_LANCZOS
2177
                |SWS_SINC
2178
                |SWS_SPLINE
2179
                |SWS_BICUBLIN);
2180
    if(!i || (i & (i-1)))
2181
    {
2182
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be choosen\n");
2183
        return NULL;
2184
    }
2185

    
2186
    /* sanity check */
2187
    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
2188
    {
2189
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2190
               srcW, srcH, dstW, dstH);
2191
        return NULL;
2192
    }
2193
    if(srcW > VOFW || dstW > VOFW){
2194
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile time max width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2195
        return NULL;
2196
    }
2197

    
2198
    if (!dstFilter) dstFilter= &dummyFilter;
2199
    if (!srcFilter) srcFilter= &dummyFilter;
2200

    
2201
    c= av_mallocz(sizeof(SwsContext));
2202

    
2203
    c->av_class = &sws_context_class;
2204
    c->srcW= srcW;
2205
    c->srcH= srcH;
2206
    c->dstW= dstW;
2207
    c->dstH= dstH;
2208
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2209
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2210
    c->flags= flags;
2211
    c->dstFormat= dstFormat;
2212
    c->srcFormat= srcFormat;
2213
    c->vRounder= 4* 0x0001000100010001ULL;
2214

    
2215
    usesHFilter= usesVFilter= 0;
2216
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2217
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2218
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2219
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2220
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2221
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2222
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2223
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2224

    
2225
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2226
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2227

    
2228
    // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
2229
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2230

    
2231
    // drop some chroma lines if the user wants it
2232
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2233
    c->chrSrcVSubSample+= c->vChrDrop;
2234

    
2235
    // drop every 2. pixel for chroma calculation unless user wants full chroma
2236
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2237
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2238
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2239
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2240
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2241
        c->chrSrcHSubSample=1;
2242

    
2243
    if (param){
2244
        c->param[0] = param[0];
2245
        c->param[1] = param[1];
2246
    }else{
2247
        c->param[0] =
2248
        c->param[1] = SWS_PARAM_DEFAULT;
2249
    }
2250

    
2251
    c->chrIntHSubSample= c->chrDstHSubSample;
2252
    c->chrIntVSubSample= c->chrSrcVSubSample;
2253

    
2254
    // Note the -((-x)>>y) is so that we always round toward +inf.
2255
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2256
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2257
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2258
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2259

    
2260
    sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2261

    
2262
    /* unscaled special Cases */
2263
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2264
    {
2265
        /* yv12_to_nv12 */
2266
        if (srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2267
        {
2268
            c->swScale= PlanarToNV12Wrapper;
2269
        }
2270
#ifdef CONFIG_GPL
2271
        /* yuv2bgr */
2272
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat))
2273
            && !(flags & SWS_ACCURATE_RND))
2274
        {
2275
            c->swScale= yuv2rgb_get_func_ptr(c);
2276
        }
2277
#endif
2278

    
2279
        if (srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P && !(flags & SWS_BITEXACT))
2280
        {
2281
            c->swScale= yvu9toyv12Wrapper;
2282
        }
2283

    
2284
        /* bgr24toYV12 */
2285
        if (srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P && !(flags & SWS_ACCURATE_RND))
2286
            c->swScale= bgr24toyv12Wrapper;
2287

    
2288
        /* rgb/bgr -> rgb/bgr (no dither needed forms) */
2289
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2290
           && (isBGR(dstFormat) || isRGB(dstFormat))
2291
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2292
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2293
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2294
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2295
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2296
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2297
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2298
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2299
                                             && dstFormat != PIX_FMT_RGB32_1
2300
                                             && dstFormat != PIX_FMT_BGR32_1
2301
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2302
             c->swScale= rgb2rgbWrapper;
2303

    
2304
        if (srcFormat == PIX_FMT_YUV422P)
2305
        {
2306
            if (dstFormat == PIX_FMT_YUYV422)
2307
                c->swScale= YUV422PToYuy2Wrapper;
2308
            else if (dstFormat == PIX_FMT_UYVY422)
2309
                c->swScale= YUV422PToUyvyWrapper;
2310
        }
2311

    
2312
        /* LQ converters if -sws 0 or -sws 4*/
2313
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2314
            /* yv12_to_yuy2 */
2315
            if (srcFormat == PIX_FMT_YUV420P)
2316
            {
2317
                if (dstFormat == PIX_FMT_YUYV422)
2318
                    c->swScale= PlanarToYuy2Wrapper;
2319
                else if (dstFormat == PIX_FMT_UYVY422)
2320
                    c->swScale= PlanarToUyvyWrapper;
2321
            }
2322
        }
2323

    
2324
#ifdef COMPILE_ALTIVEC
2325
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2326
            srcFormat == PIX_FMT_YUV420P) {
2327
          // unscaled YV12 -> packed YUV, we want speed
2328
          if (dstFormat == PIX_FMT_YUYV422)
2329
              c->swScale= yv12toyuy2_unscaled_altivec;
2330
          else if (dstFormat == PIX_FMT_UYVY422)
2331
              c->swScale= yv12touyvy_unscaled_altivec;
2332
        }
2333
#endif
2334

    
2335
        /* simple copy */
2336
        if (  srcFormat == dstFormat
2337
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2338
            || (isPlanarYUV(dstFormat) && isGray(srcFormat)))
2339
        {
2340
            if (isPacked(c->srcFormat))
2341
                c->swScale= packedCopy;
2342
            else /* Planar YUV or gray */
2343
                c->swScale= planarCopy;
2344
        }
2345

    
2346
        /* gray16{le,be} conversions */
2347
        if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))
2348
        {
2349
            c->swScale= gray16togray;
2350
        }
2351
        if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))
2352
        {
2353
            c->swScale= graytogray16;
2354
        }
2355
        if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
2356
        {
2357
            c->swScale= gray16swap;
2358
        }
2359

    
2360
#ifdef ARCH_BFIN
2361
        if (flags & SWS_CPU_CAPS_BFIN)
2362
            ff_bfin_get_unscaled_swscale (c);
2363
#endif
2364

    
2365
        if (c->swScale){
2366
            if (flags&SWS_PRINT_INFO)
2367
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2368
                                sws_format_name(srcFormat), sws_format_name(dstFormat));
2369
            return c;
2370
        }
2371
    }
2372

    
2373
    if (flags & SWS_CPU_CAPS_MMX2)
2374
    {
2375
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2376
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2377
        {
2378
            if (flags&SWS_PRINT_INFO)
2379
                av_log(c, AV_LOG_INFO, "output Width is not a multiple of 32 -> no MMX2 scaler\n");
2380
        }
2381
        if (usesHFilter) c->canMMX2BeUsed=0;
2382
    }
2383
    else
2384
        c->canMMX2BeUsed=0;
2385

    
2386
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2387
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2388

    
2389
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2390
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2391
    // n-2 is the last chrominance sample available
2392
    // this is not perfect, but no one should notice the difference, the more correct variant
2393
    // would be like the vertical one, but that would require some special code for the
2394
    // first and last pixel
2395
    if (flags&SWS_FAST_BILINEAR)
2396
    {
2397
        if (c->canMMX2BeUsed)
2398
        {
2399
            c->lumXInc+= 20;
2400
            c->chrXInc+= 20;
2401
        }
2402
        //we don't use the x86asm scaler if mmx is available
2403
        else if (flags & SWS_CPU_CAPS_MMX)
2404
        {
2405
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2406
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2407
        }
2408
    }
2409

    
2410
    /* precalculate horizontal scaler filter coefficients */
2411
    {
2412
        const int filterAlign=
2413
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2414
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2415
            1;
2416

    
2417
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2418
                   srcW      ,       dstW, filterAlign, 1<<14,
2419
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2420
                   srcFilter->lumH, dstFilter->lumH, c->param);
2421
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2422
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2423
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2424
                   srcFilter->chrH, dstFilter->chrH, c->param);
2425

    
2426
#define MAX_FUNNY_CODE_SIZE 10000
2427
#if defined(COMPILE_MMX2)
2428
// can't downscale !!!
2429
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2430
        {
2431
#ifdef MAP_ANONYMOUS
2432
            c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2433
            c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2434
#else
2435
            c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2436
            c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2437
#endif
2438

    
2439
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2440
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2441
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2442
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2443

    
2444
            initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2445
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2446
        }
2447
#endif /* defined(COMPILE_MMX2) */
2448
    } // Init Horizontal stuff
2449

    
2450

    
2451

    
2452
    /* precalculate vertical scaler filter coefficients */
2453
    {
2454
        const int filterAlign=
2455
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2456
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2457
            1;
2458

    
2459
        initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2460
                   srcH      ,        dstH, filterAlign, (1<<12),
2461
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2462
                   srcFilter->lumV, dstFilter->lumV, c->param);
2463
        initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2464
                   c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2465
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2466
                   srcFilter->chrV, dstFilter->chrV, c->param);
2467

    
2468
#ifdef HAVE_ALTIVEC
2469
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2470
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2471

    
2472
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2473
            int j;
2474
            short *p = (short *)&c->vYCoeffsBank[i];
2475
            for (j=0;j<8;j++)
2476
                p[j] = c->vLumFilter[i];
2477
        }
2478

    
2479
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2480
            int j;
2481
            short *p = (short *)&c->vCCoeffsBank[i];
2482
            for (j=0;j<8;j++)
2483
                p[j] = c->vChrFilter[i];
2484
        }
2485
#endif
2486
    }
2487

    
2488
    // Calculate Buffer Sizes so that they won't run out while handling these damn slices
2489
    c->vLumBufSize= c->vLumFilterSize;
2490
    c->vChrBufSize= c->vChrFilterSize;
2491
    for (i=0; i<dstH; i++)
2492
    {
2493
        int chrI= i*c->chrDstH / dstH;
2494
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2495
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2496

    
2497
        nextSlice>>= c->chrSrcVSubSample;
2498
        nextSlice<<= c->chrSrcVSubSample;
2499
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2500
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2501
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2502
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2503
    }
2504

    
2505
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2506
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2507
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2508
    //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)
2509
    /* align at 16 bytes for AltiVec */
2510
    for (i=0; i<c->vLumBufSize; i++)
2511
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2512
    for (i=0; i<c->vChrBufSize; i++)
2513
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2514

    
2515
    //try to avoid drawing green stuff between the right end and the stride end
2516
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2517

    
2518
    assert(2*VOFW == VOF);
2519

    
2520
    assert(c->chrDstH <= dstH);
2521

    
2522
    if (flags&SWS_PRINT_INFO)
2523
    {
2524
#ifdef DITHER1XBPP
2525
        const char *dither= " dithered";
2526
#else
2527
        const char *dither= "";
2528
#endif
2529
        if (flags&SWS_FAST_BILINEAR)
2530
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2531
        else if (flags&SWS_BILINEAR)
2532
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2533
        else if (flags&SWS_BICUBIC)
2534
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2535
        else if (flags&SWS_X)
2536
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2537
        else if (flags&SWS_POINT)
2538
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2539
        else if (flags&SWS_AREA)
2540
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2541
        else if (flags&SWS_BICUBLIN)
2542
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2543
        else if (flags&SWS_GAUSS)
2544
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2545
        else if (flags&SWS_SINC)
2546
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2547
        else if (flags&SWS_LANCZOS)
2548
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2549
        else if (flags&SWS_SPLINE)
2550
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2551
        else
2552
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2553

    
2554
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2555
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2556
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2557
        else
2558
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2559
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2560

    
2561
        if (flags & SWS_CPU_CAPS_MMX2)
2562
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2563
        else if (flags & SWS_CPU_CAPS_3DNOW)
2564
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2565
        else if (flags & SWS_CPU_CAPS_MMX)
2566
            av_log(c, AV_LOG_INFO, "using MMX\n");
2567
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2568
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2569
        else
2570
            av_log(c, AV_LOG_INFO, "using C\n");
2571
    }
2572

    
2573
    if (flags & SWS_PRINT_INFO)
2574
    {
2575
        if (flags & SWS_CPU_CAPS_MMX)
2576
        {
2577
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2578
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2579
            else
2580
            {
2581
                if (c->hLumFilterSize==4)
2582
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2583
                else if (c->hLumFilterSize==8)
2584
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2585
                else
2586
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2587

    
2588
                if (c->hChrFilterSize==4)
2589
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2590
                else if (c->hChrFilterSize==8)
2591
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2592
                else
2593
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2594
            }
2595
        }
2596
        else
2597
        {
2598
#if defined(ARCH_X86)
2599
            av_log(c, AV_LOG_VERBOSE, "using X86-Asm scaler for horizontal scaling\n");
2600
#else
2601
            if (flags & SWS_FAST_BILINEAR)
2602
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
2603
            else
2604
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
2605
#endif
2606
        }
2607
        if (isPlanarYUV(dstFormat))
2608
        {
2609
            if (c->vLumFilterSize==1)
2610
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2611
            else
2612
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2613
        }
2614
        else
2615
        {
2616
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
2617
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2618
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2619
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
2620
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2621
            else
2622
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2623
        }
2624

    
2625
        if (dstFormat==PIX_FMT_BGR24)
2626
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 Converter\n",
2627
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2628
        else if (dstFormat==PIX_FMT_RGB32)
2629
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2630
        else if (dstFormat==PIX_FMT_BGR565)
2631
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2632
        else if (dstFormat==PIX_FMT_BGR555)
2633
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2634

    
2635
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2636
    }
2637
    if (flags & SWS_PRINT_INFO)
2638
    {
2639
        av_log(c, AV_LOG_DEBUG, "Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2640
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2641
        av_log(c, AV_LOG_DEBUG, "Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2642
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2643
    }
2644

    
2645
    c->swScale= getSwsFunc(flags);
2646
    return c;
2647
}
2648

    
2649
/**
2650
 * swscale wrapper, so we don't need to export the SwsContext.
2651
 * assumes planar YUV to be in YUV order instead of YVU
2652
 */
2653
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2654
              int srcSliceH, uint8_t* dst[], int dstStride[]){
2655
    int i;
2656
    uint8_t* src2[4]= {src[0], src[1], src[2]};
2657
    uint32_t pal[256];
2658
    int use_pal=   c->srcFormat == PIX_FMT_PAL8
2659
                || c->srcFormat == PIX_FMT_BGR4_BYTE
2660
                || c->srcFormat == PIX_FMT_RGB4_BYTE
2661
                || c->srcFormat == PIX_FMT_BGR8
2662
                || c->srcFormat == PIX_FMT_RGB8;
2663

    
2664
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2665
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2666
        return 0;
2667
    }
2668
    if (c->sliceDir == 0) {
2669
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2670
    }
2671

    
2672
    if (use_pal){
2673
        for (i=0; i<256; i++){
2674
            int p, r, g, b,y,u,v;
2675
            if(c->srcFormat == PIX_FMT_PAL8){
2676
                p=((uint32_t*)(src[1]))[i];
2677
                r= (p>>16)&0xFF;
2678
                g= (p>> 8)&0xFF;
2679
                b=  p     &0xFF;
2680
            }else if(c->srcFormat == PIX_FMT_RGB8){
2681
                r= (i>>5    )*36;
2682
                g= ((i>>2)&7)*36;
2683
                b= (i&3     )*85;
2684
            }else if(c->srcFormat == PIX_FMT_BGR8){
2685
                b= (i>>6    )*85;
2686
                g= ((i>>3)&7)*36;
2687
                r= (i&7     )*36;
2688
            }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
2689
                r= (i>>3    )*255;
2690
                g= ((i>>1)&3)*85;
2691
                b= (i&1     )*255;
2692
            }else if(c->srcFormat == PIX_FMT_BGR4_BYTE){
2693
                b= (i>>3    )*255;
2694
                g= ((i>>1)&3)*85;
2695
                r= (i&1     )*255;
2696
            }
2697
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2698
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2699
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2700
            pal[i]= y + (u<<8) + (v<<16);
2701
        }
2702
        src2[1]= (uint8_t*)pal;
2703
    }
2704

    
2705
    // copy strides, so they can safely be modified
2706
    if (c->sliceDir == 1) {
2707
        // slices go from top to bottom
2708
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};
2709
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};
2710
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
2711
    } else {
2712
        // slices go from bottom to top => we flip the image internally
2713
        uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],
2714
                           dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],
2715
                           dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};
2716
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};
2717
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};
2718

    
2719
        src2[0] += (srcSliceH-1)*srcStride[0];
2720
        if (!use_pal)
2721
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2722
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2723

    
2724
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2725
    }
2726
}
2727

    
2728
/**
2729
 * swscale wrapper, so we don't need to export the SwsContext
2730
 */
2731
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2732
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
2733
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2734
}
2735

    
2736
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
2737
                                float lumaSharpen, float chromaSharpen,
2738
                                float chromaHShift, float chromaVShift,
2739
                                int verbose)
2740
{
2741
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
2742

    
2743
    if (lumaGBlur!=0.0){
2744
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
2745
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
2746
    }else{
2747
        filter->lumH= sws_getIdentityVec();
2748
        filter->lumV= sws_getIdentityVec();
2749
    }
2750

    
2751
    if (chromaGBlur!=0.0){
2752
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
2753
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
2754
    }else{
2755
        filter->chrH= sws_getIdentityVec();
2756
        filter->chrV= sws_getIdentityVec();
2757
    }
2758

    
2759
    if (chromaSharpen!=0.0){
2760
        SwsVector *id= sws_getIdentityVec();
2761
        sws_scaleVec(filter->chrH, -chromaSharpen);
2762
        sws_scaleVec(filter->chrV, -chromaSharpen);
2763
        sws_addVec(filter->chrH, id);
2764
        sws_addVec(filter->chrV, id);
2765
        sws_freeVec(id);
2766
    }
2767

    
2768
    if (lumaSharpen!=0.0){
2769
        SwsVector *id= sws_getIdentityVec();
2770
        sws_scaleVec(filter->lumH, -lumaSharpen);
2771
        sws_scaleVec(filter->lumV, -lumaSharpen);
2772
        sws_addVec(filter->lumH, id);
2773
        sws_addVec(filter->lumV, id);
2774
        sws_freeVec(id);
2775
    }
2776

    
2777
    if (chromaHShift != 0.0)
2778
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
2779

    
2780
    if (chromaVShift != 0.0)
2781
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
2782

    
2783
    sws_normalizeVec(filter->chrH, 1.0);
2784
    sws_normalizeVec(filter->chrV, 1.0);
2785
    sws_normalizeVec(filter->lumH, 1.0);
2786
    sws_normalizeVec(filter->lumV, 1.0);
2787

    
2788
    if (verbose) sws_printVec(filter->chrH);
2789
    if (verbose) sws_printVec(filter->lumH);
2790

    
2791
    return filter;
2792
}
2793

    
2794
/**
2795
 * returns a normalized gaussian curve used to filter stuff
2796
 * quality=3 is high quality, lowwer is lowwer quality
2797
 */
2798
SwsVector *sws_getGaussianVec(double variance, double quality){
2799
    const int length= (int)(variance*quality + 0.5) | 1;
2800
    int i;
2801
    double *coeff= av_malloc(length*sizeof(double));
2802
    double middle= (length-1)*0.5;
2803
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2804

    
2805
    vec->coeff= coeff;
2806
    vec->length= length;
2807

    
2808
    for (i=0; i<length; i++)
2809
    {
2810
        double dist= i-middle;
2811
        coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
2812
    }
2813

    
2814
    sws_normalizeVec(vec, 1.0);
2815

    
2816
    return vec;
2817
}
2818

    
2819
SwsVector *sws_getConstVec(double c, int length){
2820
    int i;
2821
    double *coeff= av_malloc(length*sizeof(double));
2822
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2823

    
2824
    vec->coeff= coeff;
2825
    vec->length= length;
2826

    
2827
    for (i=0; i<length; i++)
2828
        coeff[i]= c;
2829

    
2830
    return vec;
2831
}
2832

    
2833

    
2834
SwsVector *sws_getIdentityVec(void){
2835
    return sws_getConstVec(1.0, 1);
2836
}
2837

    
2838
double sws_dcVec(SwsVector *a){
2839
    int i;
2840
    double sum=0;
2841

    
2842
    for (i=0; i<a->length; i++)
2843
        sum+= a->coeff[i];
2844

    
2845
    return sum;
2846
}
2847

    
2848
void sws_scaleVec(SwsVector *a, double scalar){
2849
    int i;
2850

    
2851
    for (i=0; i<a->length; i++)
2852
        a->coeff[i]*= scalar;
2853
}
2854

    
2855
void sws_normalizeVec(SwsVector *a, double height){
2856
    sws_scaleVec(a, height/sws_dcVec(a));
2857
}
2858

    
2859
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
2860
    int length= a->length + b->length - 1;
2861
    double *coeff= av_malloc(length*sizeof(double));
2862
    int i, j;
2863
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2864

    
2865
    vec->coeff= coeff;
2866
    vec->length= length;
2867

    
2868
    for (i=0; i<length; i++) coeff[i]= 0.0;
2869

    
2870
    for (i=0; i<a->length; i++)
2871
    {
2872
        for (j=0; j<b->length; j++)
2873
        {
2874
            coeff[i+j]+= a->coeff[i]*b->coeff[j];
2875
        }
2876
    }
2877

    
2878
    return vec;
2879
}
2880

    
2881
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
2882
    int length= FFMAX(a->length, b->length);
2883
    double *coeff= av_malloc(length*sizeof(double));
2884
    int i;
2885
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2886

    
2887
    vec->coeff= coeff;
2888
    vec->length= length;
2889

    
2890
    for (i=0; i<length; i++) coeff[i]= 0.0;
2891

    
2892
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2893
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
2894

    
2895
    return vec;
2896
}
2897

    
2898
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
2899
    int length= FFMAX(a->length, b->length);
2900
    double *coeff= av_malloc(length*sizeof(double));
2901
    int i;
2902
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2903

    
2904
    vec->coeff= coeff;
2905
    vec->length= length;
2906

    
2907
    for (i=0; i<length; i++) coeff[i]= 0.0;
2908

    
2909
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2910
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
2911

    
2912
    return vec;
2913
}
2914

    
2915
/* shift left / or right if "shift" is negative */
2916
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
2917
    int length= a->length + FFABS(shift)*2;
2918
    double *coeff= av_malloc(length*sizeof(double));
2919
    int i;
2920
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2921

    
2922
    vec->coeff= coeff;
2923
    vec->length= length;
2924

    
2925
    for (i=0; i<length; i++) coeff[i]= 0.0;
2926

    
2927
    for (i=0; i<a->length; i++)
2928
    {
2929
        coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
2930
    }
2931

    
2932
    return vec;
2933
}
2934

    
2935
void sws_shiftVec(SwsVector *a, int shift){
2936
    SwsVector *shifted= sws_getShiftedVec(a, shift);
2937
    av_free(a->coeff);
2938
    a->coeff= shifted->coeff;
2939
    a->length= shifted->length;
2940
    av_free(shifted);
2941
}
2942

    
2943
void sws_addVec(SwsVector *a, SwsVector *b){
2944
    SwsVector *sum= sws_sumVec(a, b);
2945
    av_free(a->coeff);
2946
    a->coeff= sum->coeff;
2947
    a->length= sum->length;
2948
    av_free(sum);
2949
}
2950

    
2951
void sws_subVec(SwsVector *a, SwsVector *b){
2952
    SwsVector *diff= sws_diffVec(a, b);
2953
    av_free(a->coeff);
2954
    a->coeff= diff->coeff;
2955
    a->length= diff->length;
2956
    av_free(diff);
2957
}
2958

    
2959
void sws_convVec(SwsVector *a, SwsVector *b){
2960
    SwsVector *conv= sws_getConvVec(a, b);
2961
    av_free(a->coeff);
2962
    a->coeff= conv->coeff;
2963
    a->length= conv->length;
2964
    av_free(conv);
2965
}
2966

    
2967
SwsVector *sws_cloneVec(SwsVector *a){
2968
    double *coeff= av_malloc(a->length*sizeof(double));
2969
    int i;
2970
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2971

    
2972
    vec->coeff= coeff;
2973
    vec->length= a->length;
2974

    
2975
    for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
2976

    
2977
    return vec;
2978
}
2979

    
2980
void sws_printVec(SwsVector *a){
2981
    int i;
2982
    double max=0;
2983
    double min=0;
2984
    double range;
2985

    
2986
    for (i=0; i<a->length; i++)
2987
        if (a->coeff[i]>max) max= a->coeff[i];
2988

    
2989
    for (i=0; i<a->length; i++)
2990
        if (a->coeff[i]<min) min= a->coeff[i];
2991

    
2992
    range= max - min;
2993

    
2994
    for (i=0; i<a->length; i++)
2995
    {
2996
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
2997
        av_log(NULL, AV_LOG_DEBUG, "%1.3f ", a->coeff[i]);
2998
        for (;x>0; x--) av_log(NULL, AV_LOG_DEBUG, " ");
2999
        av_log(NULL, AV_LOG_DEBUG, "|\n");
3000
    }
3001
}
3002

    
3003
void sws_freeVec(SwsVector *a){
3004
    if (!a) return;
3005
    av_freep(&a->coeff);
3006
    a->length=0;
3007
    av_free(a);
3008
}
3009

    
3010
void sws_freeFilter(SwsFilter *filter){
3011
    if (!filter) return;
3012

    
3013
    if (filter->lumH) sws_freeVec(filter->lumH);
3014
    if (filter->lumV) sws_freeVec(filter->lumV);
3015
    if (filter->chrH) sws_freeVec(filter->chrH);
3016
    if (filter->chrV) sws_freeVec(filter->chrV);
3017
    av_free(filter);
3018
}
3019

    
3020

    
3021
void sws_freeContext(SwsContext *c){
3022
    int i;
3023
    if (!c) return;
3024

    
3025
    if (c->lumPixBuf)
3026
    {
3027
        for (i=0; i<c->vLumBufSize; i++)
3028
            av_freep(&c->lumPixBuf[i]);
3029
        av_freep(&c->lumPixBuf);
3030
    }
3031

    
3032
    if (c->chrPixBuf)
3033
    {
3034
        for (i=0; i<c->vChrBufSize; i++)
3035
            av_freep(&c->chrPixBuf[i]);
3036
        av_freep(&c->chrPixBuf);
3037
    }
3038

    
3039
    av_freep(&c->vLumFilter);
3040
    av_freep(&c->vChrFilter);
3041
    av_freep(&c->hLumFilter);
3042
    av_freep(&c->hChrFilter);
3043
#ifdef HAVE_ALTIVEC
3044
    av_freep(&c->vYCoeffsBank);
3045
    av_freep(&c->vCCoeffsBank);
3046
#endif
3047

    
3048
    av_freep(&c->vLumFilterPos);
3049
    av_freep(&c->vChrFilterPos);
3050
    av_freep(&c->hLumFilterPos);
3051
    av_freep(&c->hChrFilterPos);
3052

    
3053
#if defined(ARCH_X86) && defined(CONFIG_GPL)
3054
#ifdef MAP_ANONYMOUS
3055
    if (c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);
3056
    if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
3057
#else
3058
    av_free(c->funnyYCode);
3059
    av_free(c->funnyUVCode);
3060
#endif
3061
    c->funnyYCode=NULL;
3062
    c->funnyUVCode=NULL;
3063
#endif /* defined(ARCH_X86) */
3064

    
3065
    av_freep(&c->lumMmx2Filter);
3066
    av_freep(&c->chrMmx2Filter);
3067
    av_freep(&c->lumMmx2FilterPos);
3068
    av_freep(&c->chrMmx2FilterPos);
3069
    av_freep(&c->yuvTable);
3070

    
3071
    av_free(c);
3072
}
3073

    
3074
/**
3075
 * Checks if context is valid or reallocs a new one instead.
3076
 * If context is NULL, just calls sws_getContext() to get a new one.
3077
 * Otherwise, checks if the parameters are the same already saved in context.
3078
 * If that is the case, returns the current context.
3079
 * Otherwise, frees context and gets a new one.
3080
 *
3081
 * Be warned that srcFilter, dstFilter are not checked, they are
3082
 * asumed to remain valid.
3083
 */
3084
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3085
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3086
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3087
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)
3088
{
3089
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3090

    
3091
    if (!param)
3092
        param = default_param;
3093

    
3094
    if (context) {
3095
        if (context->srcW != srcW || context->srcH != srcH ||
3096
            context->srcFormat != srcFormat ||
3097
            context->dstW != dstW || context->dstH != dstH ||
3098
            context->dstFormat != dstFormat || context->flags != flags ||
3099
            context->param[0] != param[0] || context->param[1] != param[1])
3100
        {
3101
            sws_freeContext(context);
3102
            context = NULL;
3103
        }
3104
    }
3105
    if (!context) {
3106
        return sws_getContext(srcW, srcH, srcFormat,
3107
                              dstW, dstH, dstFormat, flags,
3108
                              srcFilter, dstFilter, param);
3109
    }
3110
    return context;
3111
}
3112