Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ 3164d25e

History | View | Annotate | Download (115 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
#if 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
#if HAVE_VIRTUALALLOC
72
#define WIN32_LEAN_AND_MEAN
73
#include <windows.h>
74
#endif
75
#include "swscale.h"
76
#include "swscale_internal.h"
77
#include "rgb2rgb.h"
78
#include "libavutil/x86_cpu.h"
79
#include "libavutil/bswap.h"
80

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

    
86
#undef MOVNTQ
87
#undef PAVGB
88

    
89
//#undef HAVE_MMX2
90
//#define HAVE_AMD3DNOW
91
//#undef HAVE_MMX
92
//#undef ARCH_X86
93
//#define WORDS_BIGENDIAN
94
#define DITHER1XBPP
95

    
96
#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
97

    
98
#define RET 0xC3 //near return opcode for x86
99

    
100
#ifdef M_PI
101
#define PI M_PI
102
#else
103
#define PI 3.14159265358979323846
104
#endif
105

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

    
171
#define RGB2YUV_SHIFT 15
172
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
173
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
174
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
175
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
176
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
177
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
178
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
179
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
180
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
181

    
182
extern const int32_t ff_yuv2rgb_coeffs[8][4];
183

    
184
static const double rgb2yuv_table[8][9]={
185
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
186
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
187
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
188
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
189
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
190
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
191
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
192
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
193
};
194

    
195
/*
196
NOTES
197
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
198

199
TODO
200
more intelligent misalignment avoidance for the horizontal scaler
201
write special vertical cubic upscale version
202
optimize C code (YV12 / minmax)
203
add support for packed pixel YUV input & output
204
add support for Y8 output
205
optimize BGR24 & BGR32
206
add BGR4 output support
207
write special BGR->BGR scaler
208
*/
209

    
210
#if ARCH_X86 && CONFIG_GPL
211
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
212
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
213
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
214
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
215
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
216
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
217
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
218
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
219

    
220
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
221
        0x0103010301030103LL,
222
        0x0200020002000200LL,};
223

    
224
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
225
        0x0602060206020602LL,
226
        0x0004000400040004LL,};
227

    
228
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
229
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
230
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
231
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
232
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
233
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
234

    
235
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
236
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
237
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
238

    
239
#ifdef FAST_BGR2YV12
240
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
241
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
242
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
243
#else
244
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
245
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
246
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
247
#endif /* FAST_BGR2YV12 */
248
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
249
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
250
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
251

    
252
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
253
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
254
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
255
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
256
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
257

    
258
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
259
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
260
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
261
};
262

    
263
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
264

    
265
#endif /* ARCH_X86 && CONFIG_GPL */
266

    
267
// clipping helper table for C implementations:
268
static unsigned char clip_table[768];
269

    
270
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
271

    
272
static const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={
273
{  1,   3,   1,   3,   1,   3,   1,   3, },
274
{  2,   0,   2,   0,   2,   0,   2,   0, },
275
};
276

    
277
static const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={
278
{  6,   2,   6,   2,   6,   2,   6,   2, },
279
{  0,   4,   0,   4,   0,   4,   0,   4, },
280
};
281

    
282
const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={
283
{ 17,   9,  23,  15,  16,   8,  22,  14, },
284
{  5,  29,   3,  27,   4,  28,   2,  26, },
285
{ 21,  13,  19,  11,  20,  12,  18,  10, },
286
{  0,  24,   6,  30,   1,  25,   7,  31, },
287
{ 16,   8,  22,  14,  17,   9,  23,  15, },
288
{  4,  28,   2,  26,   5,  29,   3,  27, },
289
{ 20,  12,  18,  10,  21,  13,  19,  11, },
290
{  1,  25,   7,  31,   0,  24,   6,  30, },
291
};
292

    
293
#if 0
294
const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={
295
{  0,  48,  12,  60,   3,  51,  15,  63, },
296
{ 32,  16,  44,  28,  35,  19,  47,  31, },
297
{  8,  56,   4,  52,  11,  59,   7,  55, },
298
{ 40,  24,  36,  20,  43,  27,  39,  23, },
299
{  2,  50,  14,  62,   1,  49,  13,  61, },
300
{ 34,  18,  46,  30,  33,  17,  45,  29, },
301
{ 10,  58,   6,  54,   9,  57,   5,  53, },
302
{ 42,  26,  38,  22,  41,  25,  37,  21, },
303
};
304
#endif
305

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

    
317
#if 0
318
const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={
319
{ 68,  36,  92,  60,  66,  34,  90,  58, },
320
{ 20, 116,  12, 108,  18, 114,  10, 106, },
321
{ 84,  52,  76,  44,  82,  50,  74,  42, },
322
{  0,  96,  24, 120,   6, 102,  30, 126, },
323
{ 64,  32,  88,  56,  70,  38,  94,  62, },
324
{ 16, 112,   8, 104,  22, 118,  14, 110, },
325
{ 80,  48,  72,  40,  86,  54,  78,  46, },
326
{  4, 100,  28, 124,   2,  98,  26, 122, },
327
};
328
#endif
329

    
330
#if 1
331
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
332
{117,  62, 158, 103, 113,  58, 155, 100, },
333
{ 34, 199,  21, 186,  31, 196,  17, 182, },
334
{144,  89, 131,  76, 141,  86, 127,  72, },
335
{  0, 165,  41, 206,  10, 175,  52, 217, },
336
{110,  55, 151,  96, 120,  65, 162, 107, },
337
{ 28, 193,  14, 179,  38, 203,  24, 189, },
338
{138,  83, 124,  69, 148,  93, 134,  79, },
339
{  7, 172,  48, 213,   3, 168,  45, 210, },
340
};
341
#elif 1
342
// tries to correct a gamma of 1.5
343
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
344
{  0, 143,  18, 200,   2, 156,  25, 215, },
345
{ 78,  28, 125,  64,  89,  36, 138,  74, },
346
{ 10, 180,   3, 161,  16, 195,   8, 175, },
347
{109,  51,  93,  38, 121,  60, 105,  47, },
348
{  1, 152,  23, 210,   0, 147,  20, 205, },
349
{ 85,  33, 134,  71,  81,  30, 130,  67, },
350
{ 14, 190,   6, 171,  12, 185,   5, 166, },
351
{117,  57, 101,  44, 113,  54,  97,  41, },
352
};
353
#elif 1
354
// tries to correct a gamma of 2.0
355
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
356
{  0, 124,   8, 193,   0, 140,  12, 213, },
357
{ 55,  14, 104,  42,  66,  19, 119,  52, },
358
{  3, 168,   1, 145,   6, 187,   3, 162, },
359
{ 86,  31,  70,  21,  99,  39,  82,  28, },
360
{  0, 134,  11, 206,   0, 129,   9, 200, },
361
{ 62,  17, 114,  48,  58,  16, 109,  45, },
362
{  5, 181,   2, 157,   4, 175,   1, 151, },
363
{ 95,  36,  78,  26,  90,  34,  74,  24, },
364
};
365
#else
366
// tries to correct a gamma of 2.5
367
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
368
{  0, 107,   3, 187,   0, 125,   6, 212, },
369
{ 39,   7,  86,  28,  49,  11, 102,  36, },
370
{  1, 158,   0, 131,   3, 180,   1, 151, },
371
{ 68,  19,  52,  12,  81,  25,  64,  17, },
372
{  0, 119,   5, 203,   0, 113,   4, 195, },
373
{ 45,   9,  96,  33,  42,   8,  91,  30, },
374
{  2, 172,   1, 144,   2, 165,   0, 137, },
375
{ 77,  23,  60,  15,  72,  21,  56,  14, },
376
};
377
#endif
378

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

    
475
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
476
                               int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
477
                               int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
478
{
479
    //FIXME Optimize (just quickly written not optimized..)
480
    int i;
481
    for (i=0; i<dstW; i++)
482
    {
483
        int val=1<<18;
484
        int j;
485
        for (j=0; j<lumFilterSize; j++)
486
            val += lumSrc[j][i] * lumFilter[j];
487

    
488
        dest[i]= av_clip_uint8(val>>19);
489
    }
490

    
491
    if (uDest)
492
        for (i=0; i<chrDstW; i++)
493
        {
494
            int u=1<<18;
495
            int v=1<<18;
496
            int j;
497
            for (j=0; j<chrFilterSize; j++)
498
            {
499
                u += chrSrc[j][i] * chrFilter[j];
500
                v += chrSrc[j][i + VOFW] * chrFilter[j];
501
            }
502

    
503
            uDest[i]= av_clip_uint8(u>>19);
504
            vDest[i]= av_clip_uint8(v>>19);
505
        }
506

    
507
    if (CONFIG_SWSCALE_ALPHA && aDest)
508
        for (i=0; i<dstW; i++){
509
            int val=1<<18;
510
            int j;
511
            for (j=0; j<lumFilterSize; j++)
512
                val += alpSrc[j][i] * lumFilter[j];
513

    
514
            aDest[i]= av_clip_uint8(val>>19);
515
        }
516

    
517
}
518

    
519
static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
520
                                int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
521
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
522
{
523
    //FIXME Optimize (just quickly written not optimized..)
524
    int i;
525
    for (i=0; i<dstW; i++)
526
    {
527
        int val=1<<18;
528
        int j;
529
        for (j=0; j<lumFilterSize; j++)
530
            val += lumSrc[j][i] * lumFilter[j];
531

    
532
        dest[i]= av_clip_uint8(val>>19);
533
    }
534

    
535
    if (!uDest)
536
        return;
537

    
538
    if (dstFormat == PIX_FMT_NV12)
539
        for (i=0; i<chrDstW; i++)
540
        {
541
            int u=1<<18;
542
            int v=1<<18;
543
            int j;
544
            for (j=0; j<chrFilterSize; j++)
545
            {
546
                u += chrSrc[j][i] * chrFilter[j];
547
                v += chrSrc[j][i + VOFW] * chrFilter[j];
548
            }
549

    
550
            uDest[2*i]= av_clip_uint8(u>>19);
551
            uDest[2*i+1]= av_clip_uint8(v>>19);
552
        }
553
    else
554
        for (i=0; i<chrDstW; i++)
555
        {
556
            int u=1<<18;
557
            int v=1<<18;
558
            int j;
559
            for (j=0; j<chrFilterSize; j++)
560
            {
561
                u += chrSrc[j][i] * chrFilter[j];
562
                v += chrSrc[j][i + VOFW] * chrFilter[j];
563
            }
564

    
565
            uDest[2*i]= av_clip_uint8(v>>19);
566
            uDest[2*i+1]= av_clip_uint8(u>>19);
567
        }
568
}
569

    
570
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
571
    for (i=0; i<(dstW>>1); i++){\
572
        int j;\
573
        int Y1 = 1<<18;\
574
        int Y2 = 1<<18;\
575
        int U  = 1<<18;\
576
        int V  = 1<<18;\
577
        int av_unused A1, A2;\
578
        type av_unused *r, *b, *g;\
579
        const int i2= 2*i;\
580
        \
581
        for (j=0; j<lumFilterSize; j++)\
582
        {\
583
            Y1 += lumSrc[j][i2] * lumFilter[j];\
584
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
585
        }\
586
        for (j=0; j<chrFilterSize; j++)\
587
        {\
588
            U += chrSrc[j][i] * chrFilter[j];\
589
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
590
        }\
591
        Y1>>=19;\
592
        Y2>>=19;\
593
        U >>=19;\
594
        V >>=19;\
595
        if (alpha){\
596
            A1 = 1<<18;\
597
            A2 = 1<<18;\
598
            for (j=0; j<lumFilterSize; j++){\
599
                A1 += alpSrc[j][i2  ] * lumFilter[j];\
600
                A2 += alpSrc[j][i2+1] * lumFilter[j];\
601
            }\
602
            A1>>=19;\
603
            A2>>=19;\
604
        }\
605

    
606
#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
607
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
608
        if ((Y1|Y2|U|V)&256)\
609
        {\
610
            if (Y1>255)   Y1=255; \
611
            else if (Y1<0)Y1=0;   \
612
            if (Y2>255)   Y2=255; \
613
            else if (Y2<0)Y2=0;   \
614
            if (U>255)    U=255;  \
615
            else if (U<0) U=0;    \
616
            if (V>255)    V=255;  \
617
            else if (V<0) V=0;    \
618
        }\
619
        if (alpha && ((A1|A2)&256)){\
620
            A1=av_clip_uint8(A1);\
621
            A2=av_clip_uint8(A2);\
622
        }
623

    
624
#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
625
    for (i=0; i<dstW; i++){\
626
        int j;\
627
        int Y = 0;\
628
        int U = -128<<19;\
629
        int V = -128<<19;\
630
        int av_unused A;\
631
        int R,G,B;\
632
        \
633
        for (j=0; j<lumFilterSize; j++){\
634
            Y += lumSrc[j][i     ] * lumFilter[j];\
635
        }\
636
        for (j=0; j<chrFilterSize; j++){\
637
            U += chrSrc[j][i     ] * chrFilter[j];\
638
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
639
        }\
640
        Y >>=10;\
641
        U >>=10;\
642
        V >>=10;\
643
        if (alpha){\
644
            A = rnd;\
645
            for (j=0; j<lumFilterSize; j++)\
646
                A += alpSrc[j][i     ] * lumFilter[j];\
647
            A >>=19;\
648
            if (A&256)\
649
                A = av_clip_uint8(A);\
650
        }\
651

    
652
#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
653
    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
654
        Y-= c->yuv2rgb_y_offset;\
655
        Y*= c->yuv2rgb_y_coeff;\
656
        Y+= rnd;\
657
        R= Y + V*c->yuv2rgb_v2r_coeff;\
658
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
659
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
660
        if ((R|G|B)&(0xC0000000)){\
661
            if (R>=(256<<22))   R=(256<<22)-1; \
662
            else if (R<0)R=0;   \
663
            if (G>=(256<<22))   G=(256<<22)-1; \
664
            else if (G<0)G=0;   \
665
            if (B>=(256<<22))   B=(256<<22)-1; \
666
            else if (B<0)B=0;   \
667
        }\
668

    
669

    
670
#define YSCALE_YUV_2_GRAY16_C \
671
    for (i=0; i<(dstW>>1); i++){\
672
        int j;\
673
        int Y1 = 1<<18;\
674
        int Y2 = 1<<18;\
675
        int U  = 1<<18;\
676
        int V  = 1<<18;\
677
        \
678
        const int i2= 2*i;\
679
        \
680
        for (j=0; j<lumFilterSize; j++)\
681
        {\
682
            Y1 += lumSrc[j][i2] * lumFilter[j];\
683
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
684
        }\
685
        Y1>>=11;\
686
        Y2>>=11;\
687
        if ((Y1|Y2|U|V)&65536)\
688
        {\
689
            if (Y1>65535)   Y1=65535; \
690
            else if (Y1<0)Y1=0;   \
691
            if (Y2>65535)   Y2=65535; \
692
            else if (Y2<0)Y2=0;   \
693
        }
694

    
695
#define YSCALE_YUV_2_RGBX_C(type,alpha) \
696
    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
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_PACKED2_C(type,alpha)   \
702
    for (i=0; i<(dstW>>1); i++){ \
703
        const int i2= 2*i;       \
704
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
705
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
706
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
707
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
708
        type av_unused *r, *b, *g;                                    \
709
        int av_unused A1, A2;                                         \
710
        if (alpha){\
711
            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
712
            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
713
        }\
714

    
715
#define YSCALE_YUV_2_GRAY16_2_C   \
716
    for (i=0; i<(dstW>>1); i++){ \
717
        const int i2= 2*i;       \
718
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
719
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
720

    
721
#define YSCALE_YUV_2_RGB2_C(type,alpha) \
722
    YSCALE_YUV_2_PACKED2_C(type,alpha)\
723
    r = (type *)c->table_rV[V];\
724
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
725
    b = (type *)c->table_bU[U];\
726

    
727
#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
728
    for (i=0; i<(dstW>>1); i++){\
729
        const int i2= 2*i;\
730
        int Y1= buf0[i2  ]>>7;\
731
        int Y2= buf0[i2+1]>>7;\
732
        int U= (uvbuf1[i     ])>>7;\
733
        int V= (uvbuf1[i+VOFW])>>7;\
734
        type av_unused *r, *b, *g;\
735
        int av_unused A1, A2;\
736
        if (alpha){\
737
            A1= abuf0[i2  ]>>7;\
738
            A2= abuf0[i2+1]>>7;\
739
        }\
740

    
741
#define YSCALE_YUV_2_GRAY16_1_C \
742
    for (i=0; i<(dstW>>1); i++){\
743
        const int i2= 2*i;\
744
        int Y1= buf0[i2  ]<<1;\
745
        int Y2= buf0[i2+1]<<1;\
746

    
747
#define YSCALE_YUV_2_RGB1_C(type,alpha) \
748
    YSCALE_YUV_2_PACKED1_C(type,alpha)\
749
    r = (type *)c->table_rV[V];\
750
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
751
    b = (type *)c->table_bU[U];\
752

    
753
#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
754
    for (i=0; i<(dstW>>1); i++){\
755
        const int i2= 2*i;\
756
        int Y1= buf0[i2  ]>>7;\
757
        int Y2= buf0[i2+1]>>7;\
758
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
759
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
760
        type av_unused *r, *b, *g;\
761
        int av_unused A1, A2;\
762
        if (alpha){\
763
            A1= abuf0[i2  ]>>7;\
764
            A2= abuf0[i2+1]>>7;\
765
        }\
766

    
767
#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
768
    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
769
    r = (type *)c->table_rV[V];\
770
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
771
    b = (type *)c->table_bU[U];\
772

    
773
#define YSCALE_YUV_2_MONO2_C \
774
    const uint8_t * const d128=dither_8x8_220[y&7];\
775
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
776
    for (i=0; i<dstW-7; i+=8){\
777
        int acc;\
778
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
779
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
780
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
781
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
782
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
783
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
784
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
785
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
786
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
787
        dest++;\
788
    }\
789

    
790

    
791
#define YSCALE_YUV_2_MONOX_C \
792
    const uint8_t * const d128=dither_8x8_220[y&7];\
793
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
794
    int acc=0;\
795
    for (i=0; i<dstW-1; i+=2){\
796
        int j;\
797
        int Y1=1<<18;\
798
        int Y2=1<<18;\
799
\
800
        for (j=0; j<lumFilterSize; j++)\
801
        {\
802
            Y1 += lumSrc[j][i] * lumFilter[j];\
803
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
804
        }\
805
        Y1>>=19;\
806
        Y2>>=19;\
807
        if ((Y1|Y2)&256)\
808
        {\
809
            if (Y1>255)   Y1=255;\
810
            else if (Y1<0)Y1=0;\
811
            if (Y2>255)   Y2=255;\
812
            else if (Y2<0)Y2=0;\
813
        }\
814
        acc+= acc + g[Y1+d128[(i+0)&7]];\
815
        acc+= acc + g[Y2+d128[(i+1)&7]];\
816
        if ((i&7)==6){\
817
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
818
            dest++;\
819
        }\
820
    }
821

    
822

    
823
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
824
    switch(c->dstFormat)\
825
    {\
826
    case PIX_FMT_RGBA:\
827
    case PIX_FMT_BGRA:\
828
        if (CONFIG_SMALL){\
829
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
830
            func(uint32_t,needAlpha)\
831
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
832
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
833
            }\
834
        }else{\
835
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
836
                func(uint32_t,1)\
837
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
838
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
839
                }\
840
            }else{\
841
                func(uint32_t,0)\
842
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
843
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
844
                }\
845
            }\
846
        }\
847
        break;\
848
    case PIX_FMT_ARGB:\
849
    case PIX_FMT_ABGR:\
850
        if (CONFIG_SMALL){\
851
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
852
            func(uint32_t,needAlpha)\
853
                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
854
                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
855
            }\
856
        }else{\
857
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
858
                func(uint32_t,1)\
859
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
860
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
861
                }\
862
            }else{\
863
                func(uint32_t,0)\
864
                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
865
                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
866
                }\
867
            }\
868
        }                \
869
        break;\
870
    case PIX_FMT_RGB24:\
871
        func(uint8_t,0)\
872
            ((uint8_t*)dest)[0]= r[Y1];\
873
            ((uint8_t*)dest)[1]= g[Y1];\
874
            ((uint8_t*)dest)[2]= b[Y1];\
875
            ((uint8_t*)dest)[3]= r[Y2];\
876
            ((uint8_t*)dest)[4]= g[Y2];\
877
            ((uint8_t*)dest)[5]= b[Y2];\
878
            dest+=6;\
879
        }\
880
        break;\
881
    case PIX_FMT_BGR24:\
882
        func(uint8_t,0)\
883
            ((uint8_t*)dest)[0]= b[Y1];\
884
            ((uint8_t*)dest)[1]= g[Y1];\
885
            ((uint8_t*)dest)[2]= r[Y1];\
886
            ((uint8_t*)dest)[3]= b[Y2];\
887
            ((uint8_t*)dest)[4]= g[Y2];\
888
            ((uint8_t*)dest)[5]= r[Y2];\
889
            dest+=6;\
890
        }\
891
        break;\
892
    case PIX_FMT_RGB565:\
893
    case PIX_FMT_BGR565:\
894
        {\
895
            const int dr1= dither_2x2_8[y&1    ][0];\
896
            const int dg1= dither_2x2_4[y&1    ][0];\
897
            const int db1= dither_2x2_8[(y&1)^1][0];\
898
            const int dr2= dither_2x2_8[y&1    ][1];\
899
            const int dg2= dither_2x2_4[y&1    ][1];\
900
            const int db2= dither_2x2_8[(y&1)^1][1];\
901
            func(uint16_t,0)\
902
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
903
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
904
            }\
905
        }\
906
        break;\
907
    case PIX_FMT_RGB555:\
908
    case PIX_FMT_BGR555:\
909
        {\
910
            const int dr1= dither_2x2_8[y&1    ][0];\
911
            const int dg1= dither_2x2_8[y&1    ][1];\
912
            const int db1= dither_2x2_8[(y&1)^1][0];\
913
            const int dr2= dither_2x2_8[y&1    ][1];\
914
            const int dg2= dither_2x2_8[y&1    ][0];\
915
            const int db2= dither_2x2_8[(y&1)^1][1];\
916
            func(uint16_t,0)\
917
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
918
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
919
            }\
920
        }\
921
        break;\
922
    case PIX_FMT_RGB8:\
923
    case PIX_FMT_BGR8:\
924
        {\
925
            const uint8_t * const d64= dither_8x8_73[y&7];\
926
            const uint8_t * const d32= dither_8x8_32[y&7];\
927
            func(uint8_t,0)\
928
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
929
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
930
            }\
931
        }\
932
        break;\
933
    case PIX_FMT_RGB4:\
934
    case PIX_FMT_BGR4:\
935
        {\
936
            const uint8_t * const d64= dither_8x8_73 [y&7];\
937
            const uint8_t * const d128=dither_8x8_220[y&7];\
938
            func(uint8_t,0)\
939
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
940
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
941
            }\
942
        }\
943
        break;\
944
    case PIX_FMT_RGB4_BYTE:\
945
    case PIX_FMT_BGR4_BYTE:\
946
        {\
947
            const uint8_t * const d64= dither_8x8_73 [y&7];\
948
            const uint8_t * const d128=dither_8x8_220[y&7];\
949
            func(uint8_t,0)\
950
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
951
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
952
            }\
953
        }\
954
        break;\
955
    case PIX_FMT_MONOBLACK:\
956
    case PIX_FMT_MONOWHITE:\
957
        {\
958
            func_monoblack\
959
        }\
960
        break;\
961
    case PIX_FMT_YUYV422:\
962
        func2\
963
            ((uint8_t*)dest)[2*i2+0]= Y1;\
964
            ((uint8_t*)dest)[2*i2+1]= U;\
965
            ((uint8_t*)dest)[2*i2+2]= Y2;\
966
            ((uint8_t*)dest)[2*i2+3]= V;\
967
        }                \
968
        break;\
969
    case PIX_FMT_UYVY422:\
970
        func2\
971
            ((uint8_t*)dest)[2*i2+0]= U;\
972
            ((uint8_t*)dest)[2*i2+1]= Y1;\
973
            ((uint8_t*)dest)[2*i2+2]= V;\
974
            ((uint8_t*)dest)[2*i2+3]= Y2;\
975
        }                \
976
        break;\
977
    case PIX_FMT_GRAY16BE:\
978
        func_g16\
979
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
980
            ((uint8_t*)dest)[2*i2+1]= Y1;\
981
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
982
            ((uint8_t*)dest)[2*i2+3]= Y2;\
983
        }                \
984
        break;\
985
    case PIX_FMT_GRAY16LE:\
986
        func_g16\
987
            ((uint8_t*)dest)[2*i2+0]= Y1;\
988
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
989
            ((uint8_t*)dest)[2*i2+2]= Y2;\
990
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
991
        }                \
992
        break;\
993
    }\
994

    
995

    
996
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
997
                                  int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
998
                                  int16_t **alpSrc, uint8_t *dest, int dstW, int y)
999
{
1000
    int i;
1001
    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
1002
}
1003

    
1004
static inline void yuv2rgbXinC_full(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
1005
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
1006
                                    int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1007
{
1008
    int i;
1009
    int step= fmt_depth(c->dstFormat)/8;
1010
    int aidx= 3;
1011

    
1012
    switch(c->dstFormat){
1013
    case PIX_FMT_ARGB:
1014
        dest++;
1015
        aidx= 0;
1016
    case PIX_FMT_RGB24:
1017
        aidx--;
1018
    case PIX_FMT_RGBA:
1019
        if (CONFIG_SMALL){
1020
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1021
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1022
                dest[aidx]= needAlpha ? A : 255;
1023
                dest[0]= R>>22;
1024
                dest[1]= G>>22;
1025
                dest[2]= B>>22;
1026
                dest+= step;
1027
            }
1028
        }else{
1029
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1030
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1031
                    dest[aidx]= A;
1032
                    dest[0]= R>>22;
1033
                    dest[1]= G>>22;
1034
                    dest[2]= B>>22;
1035
                    dest+= step;
1036
                }
1037
            }else{
1038
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1039
                    dest[aidx]= 255;
1040
                    dest[0]= R>>22;
1041
                    dest[1]= G>>22;
1042
                    dest[2]= B>>22;
1043
                    dest+= step;
1044
                }
1045
            }
1046
        }
1047
        break;
1048
    case PIX_FMT_ABGR:
1049
        dest++;
1050
        aidx= 0;
1051
    case PIX_FMT_BGR24:
1052
        aidx--;
1053
    case PIX_FMT_BGRA:
1054
        if (CONFIG_SMALL){
1055
            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1056
            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1057
                dest[aidx]= needAlpha ? A : 255;
1058
                dest[0]= B>>22;
1059
                dest[1]= G>>22;
1060
                dest[2]= R>>22;
1061
                dest+= step;
1062
            }
1063
        }else{
1064
            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1065
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1066
                    dest[aidx]= A;
1067
                    dest[0]= B>>22;
1068
                    dest[1]= G>>22;
1069
                    dest[2]= R>>22;
1070
                    dest+= step;
1071
                }
1072
            }else{
1073
                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1074
                    dest[aidx]= 255;
1075
                    dest[0]= B>>22;
1076
                    dest[1]= G>>22;
1077
                    dest[2]= R>>22;
1078
                    dest+= step;
1079
                }
1080
            }
1081
        }
1082
        break;
1083
    default:
1084
        assert(0);
1085
    }
1086
}
1087

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

    
1097
//Note: we have C, X86, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1098
//Plain C versions
1099
#if !HAVE_MMX || defined (RUNTIME_CPUDETECT) || !CONFIG_GPL
1100
#define COMPILE_C
1101
#endif
1102

    
1103
#if ARCH_PPC
1104
#if (HAVE_ALTIVEC || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
1105
#undef COMPILE_C
1106
#define COMPILE_ALTIVEC
1107
#endif
1108
#endif //ARCH_PPC
1109

    
1110
#if ARCH_X86
1111

    
1112
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
1113
#define COMPILE_MMX
1114
#endif
1115

    
1116
#if (HAVE_MMX2 || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
1117
#define COMPILE_MMX2
1118
#endif
1119

    
1120
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
1121
#define COMPILE_3DNOW
1122
#endif
1123
#endif //ARCH_X86
1124

    
1125
#undef HAVE_MMX
1126
#undef HAVE_MMX2
1127
#undef HAVE_AMD3DNOW
1128
#undef HAVE_ALTIVEC
1129
#define HAVE_MMX 0
1130
#define HAVE_MMX2 0
1131
#define HAVE_AMD3DNOW 0
1132
#define HAVE_ALTIVEC 0
1133

    
1134
#ifdef COMPILE_C
1135
#define RENAME(a) a ## _C
1136
#include "swscale_template.c"
1137
#endif
1138

    
1139
#ifdef COMPILE_ALTIVEC
1140
#undef RENAME
1141
#undef HAVE_ALTIVEC
1142
#define HAVE_ALTIVEC 1
1143
#define RENAME(a) a ## _altivec
1144
#include "swscale_template.c"
1145
#endif
1146

    
1147
#if ARCH_X86
1148

    
1149
//x86 versions
1150
/*
1151
#undef RENAME
1152
#undef HAVE_MMX
1153
#undef HAVE_MMX2
1154
#undef HAVE_AMD3DNOW
1155
#define ARCH_X86
1156
#define RENAME(a) a ## _X86
1157
#include "swscale_template.c"
1158
*/
1159
//MMX versions
1160
#ifdef COMPILE_MMX
1161
#undef RENAME
1162
#undef HAVE_MMX
1163
#undef HAVE_MMX2
1164
#undef HAVE_AMD3DNOW
1165
#define HAVE_MMX 1
1166
#define HAVE_MMX2 0
1167
#define HAVE_AMD3DNOW 0
1168
#define RENAME(a) a ## _MMX
1169
#include "swscale_template.c"
1170
#endif
1171

    
1172
//MMX2 versions
1173
#ifdef COMPILE_MMX2
1174
#undef RENAME
1175
#undef HAVE_MMX
1176
#undef HAVE_MMX2
1177
#undef HAVE_AMD3DNOW
1178
#define HAVE_MMX 1
1179
#define HAVE_MMX2 1
1180
#define HAVE_AMD3DNOW 0
1181
#define RENAME(a) a ## _MMX2
1182
#include "swscale_template.c"
1183
#endif
1184

    
1185
//3DNOW versions
1186
#ifdef COMPILE_3DNOW
1187
#undef RENAME
1188
#undef HAVE_MMX
1189
#undef HAVE_MMX2
1190
#undef HAVE_AMD3DNOW
1191
#define HAVE_MMX 1
1192
#define HAVE_MMX2 0
1193
#define HAVE_AMD3DNOW 1
1194
#define RENAME(a) a ## _3DNow
1195
#include "swscale_template.c"
1196
#endif
1197

    
1198
#endif //ARCH_X86
1199

    
1200
// minor note: the HAVE_xyz are messed up after this line so don't use them
1201

    
1202
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1203
{
1204
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1205
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
1206
    else                return getSplineCoeff(        0.0,
1207
                                             b+ 2.0*c + 3.0*d,
1208
                                                    c + 3.0*d,
1209
                                            -b- 3.0*c - 6.0*d,
1210
                                            dist-1.0);
1211
}
1212

    
1213
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1214
                             int srcW, int dstW, int filterAlign, int one, int flags,
1215
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1216
{
1217
    int i;
1218
    int filterSize;
1219
    int filter2Size;
1220
    int minFilterSize;
1221
    int64_t *filter=NULL;
1222
    int64_t *filter2=NULL;
1223
    const int64_t fone= 1LL<<54;
1224
    int ret= -1;
1225
#if ARCH_X86
1226
    if (flags & SWS_CPU_CAPS_MMX)
1227
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1228
#endif
1229

    
1230
    // NOTE: the +1 is for the MMX scaler which reads over the end
1231
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1232

    
1233
    if (FFABS(xInc - 0x10000) <10) // unscaled
1234
    {
1235
        int i;
1236
        filterSize= 1;
1237
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1238

    
1239
        for (i=0; i<dstW; i++)
1240
        {
1241
            filter[i*filterSize]= fone;
1242
            (*filterPos)[i]=i;
1243
        }
1244

    
1245
    }
1246
    else if (flags&SWS_POINT) // lame looking point sampling mode
1247
    {
1248
        int i;
1249
        int xDstInSrc;
1250
        filterSize= 1;
1251
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1252

    
1253
        xDstInSrc= xInc/2 - 0x8000;
1254
        for (i=0; i<dstW; i++)
1255
        {
1256
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1257

    
1258
            (*filterPos)[i]= xx;
1259
            filter[i]= fone;
1260
            xDstInSrc+= xInc;
1261
        }
1262
    }
1263
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1264
    {
1265
        int i;
1266
        int xDstInSrc;
1267
        if      (flags&SWS_BICUBIC) filterSize= 4;
1268
        else if (flags&SWS_X      ) filterSize= 4;
1269
        else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA
1270
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1271

    
1272
        xDstInSrc= xInc/2 - 0x8000;
1273
        for (i=0; i<dstW; i++)
1274
        {
1275
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1276
            int j;
1277

    
1278
            (*filterPos)[i]= xx;
1279
                //bilinear upscale / linear interpolate / area averaging
1280
                for (j=0; j<filterSize; j++)
1281
                {
1282
                    int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1283
                    if (coeff<0) coeff=0;
1284
                    filter[i*filterSize + j]= coeff;
1285
                    xx++;
1286
                }
1287
            xDstInSrc+= xInc;
1288
        }
1289
    }
1290
    else
1291
    {
1292
        int xDstInSrc;
1293
        int sizeFactor;
1294

    
1295
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1296
        else if (flags&SWS_X)            sizeFactor=  8;
1297
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1298
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1299
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1300
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1301
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1302
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1303
        else {
1304
            sizeFactor= 0; //GCC warning killer
1305
            assert(0);
1306
        }
1307

    
1308
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1309
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1310

    
1311
        if (filterSize > srcW-2) filterSize=srcW-2;
1312

    
1313
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1314

    
1315
        xDstInSrc= xInc - 0x10000;
1316
        for (i=0; i<dstW; i++)
1317
        {
1318
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1319
            int j;
1320
            (*filterPos)[i]= xx;
1321
            for (j=0; j<filterSize; j++)
1322
            {
1323
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1324
                double floatd;
1325
                int64_t coeff;
1326

    
1327
                if (xInc > 1<<16)
1328
                    d= d*dstW/srcW;
1329
                floatd= d * (1.0/(1<<30));
1330

    
1331
                if (flags & SWS_BICUBIC)
1332
                {
1333
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1334
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1335
                    int64_t dd = ( d*d)>>30;
1336
                    int64_t ddd= (dd*d)>>30;
1337

    
1338
                    if      (d < 1LL<<30)
1339
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1340
                    else if (d < 1LL<<31)
1341
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1342
                    else
1343
                        coeff=0.0;
1344
                    coeff *= fone>>(30+24);
1345
                }
1346
/*                else if (flags & SWS_X)
1347
                {
1348
                    double p= param ? param*0.01 : 0.3;
1349
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1350
                    coeff*= pow(2.0, - p*d*d);
1351
                }*/
1352
                else if (flags & SWS_X)
1353
                {
1354
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1355
                    double c;
1356

    
1357
                    if (floatd<1.0)
1358
                        c = cos(floatd*PI);
1359
                    else
1360
                        c=-1.0;
1361
                    if (c<0.0)      c= -pow(-c, A);
1362
                    else            c=  pow( c, A);
1363
                    coeff= (c*0.5 + 0.5)*fone;
1364
                }
1365
                else if (flags & SWS_AREA)
1366
                {
1367
                    int64_t d2= d - (1<<29);
1368
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1369
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1370
                    else coeff=0.0;
1371
                    coeff *= fone>>(30+16);
1372
                }
1373
                else if (flags & SWS_GAUSS)
1374
                {
1375
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1376
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1377
                }
1378
                else if (flags & SWS_SINC)
1379
                {
1380
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1381
                }
1382
                else if (flags & SWS_LANCZOS)
1383
                {
1384
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1385
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1386
                    if (floatd>p) coeff=0;
1387
                }
1388
                else if (flags & SWS_BILINEAR)
1389
                {
1390
                    coeff= (1<<30) - d;
1391
                    if (coeff<0) coeff=0;
1392
                    coeff *= fone >> 30;
1393
                }
1394
                else if (flags & SWS_SPLINE)
1395
                {
1396
                    double p=-2.196152422706632;
1397
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1398
                }
1399
                else {
1400
                    coeff= 0.0; //GCC warning killer
1401
                    assert(0);
1402
                }
1403

    
1404
                filter[i*filterSize + j]= coeff;
1405
                xx++;
1406
            }
1407
            xDstInSrc+= 2*xInc;
1408
        }
1409
    }
1410

    
1411
    /* apply src & dst Filter to filter -> filter2
1412
       av_free(filter);
1413
    */
1414
    assert(filterSize>0);
1415
    filter2Size= filterSize;
1416
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1417
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1418
    assert(filter2Size>0);
1419
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1420

    
1421
    for (i=0; i<dstW; i++)
1422
    {
1423
        int j, k;
1424

    
1425
        if(srcFilter){
1426
            for (k=0; k<srcFilter->length; k++){
1427
                for (j=0; j<filterSize; j++)
1428
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1429
            }
1430
        }else{
1431
            for (j=0; j<filterSize; j++)
1432
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1433
        }
1434
        //FIXME dstFilter
1435

    
1436
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1437
    }
1438
    av_freep(&filter);
1439

    
1440
    /* try to reduce the filter-size (step1 find size and shift left) */
1441
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1442
    minFilterSize= 0;
1443
    for (i=dstW-1; i>=0; i--)
1444
    {
1445
        int min= filter2Size;
1446
        int j;
1447
        int64_t cutOff=0.0;
1448

    
1449
        /* get rid off near zero elements on the left by shifting left */
1450
        for (j=0; j<filter2Size; j++)
1451
        {
1452
            int k;
1453
            cutOff += FFABS(filter2[i*filter2Size]);
1454

    
1455
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1456

    
1457
            /* preserve monotonicity because the core can't handle the filter otherwise */
1458
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1459

    
1460
            // move filter coefficients left
1461
            for (k=1; k<filter2Size; k++)
1462
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1463
            filter2[i*filter2Size + k - 1]= 0;
1464
            (*filterPos)[i]++;
1465
        }
1466

    
1467
        cutOff=0;
1468
        /* count near zeros on the right */
1469
        for (j=filter2Size-1; j>0; j--)
1470
        {
1471
            cutOff += FFABS(filter2[i*filter2Size + j]);
1472

    
1473
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1474
            min--;
1475
        }
1476

    
1477
        if (min>minFilterSize) minFilterSize= min;
1478
    }
1479

    
1480
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1481
        // we can handle the special case 4,
1482
        // so we don't want to go to the full 8
1483
        if (minFilterSize < 5)
1484
            filterAlign = 4;
1485

    
1486
        // We really don't want to waste our time
1487
        // doing useless computation, so fall back on
1488
        // the scalar C code for very small filters.
1489
        // Vectorizing is worth it only if you have a
1490
        // decent-sized vector.
1491
        if (minFilterSize < 3)
1492
            filterAlign = 1;
1493
    }
1494

    
1495
    if (flags & SWS_CPU_CAPS_MMX) {
1496
        // special case for unscaled vertical filtering
1497
        if (minFilterSize == 1 && filterAlign == 2)
1498
            filterAlign= 1;
1499
    }
1500

    
1501
    assert(minFilterSize > 0);
1502
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1503
    assert(filterSize > 0);
1504
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1505
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1506
        goto error;
1507
    *outFilterSize= filterSize;
1508

    
1509
    if (flags&SWS_PRINT_INFO)
1510
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1511
    /* try to reduce the filter-size (step2 reduce it) */
1512
    for (i=0; i<dstW; i++)
1513
    {
1514
        int j;
1515

    
1516
        for (j=0; j<filterSize; j++)
1517
        {
1518
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1519
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1520
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1521
                filter[i*filterSize + j]= 0;
1522
        }
1523
    }
1524

    
1525

    
1526
    //FIXME try to align filterPos if possible
1527

    
1528
    //fix borders
1529
    for (i=0; i<dstW; i++)
1530
    {
1531
        int j;
1532
        if ((*filterPos)[i] < 0)
1533
        {
1534
            // move filter coefficients left to compensate for filterPos
1535
            for (j=1; j<filterSize; j++)
1536
            {
1537
                int left= FFMAX(j + (*filterPos)[i], 0);
1538
                filter[i*filterSize + left] += filter[i*filterSize + j];
1539
                filter[i*filterSize + j]=0;
1540
            }
1541
            (*filterPos)[i]= 0;
1542
        }
1543

    
1544
        if ((*filterPos)[i] + filterSize > srcW)
1545
        {
1546
            int shift= (*filterPos)[i] + filterSize - srcW;
1547
            // move filter coefficients right to compensate for filterPos
1548
            for (j=filterSize-2; j>=0; j--)
1549
            {
1550
                int right= FFMIN(j + shift, filterSize-1);
1551
                filter[i*filterSize +right] += filter[i*filterSize +j];
1552
                filter[i*filterSize +j]=0;
1553
            }
1554
            (*filterPos)[i]= srcW - filterSize;
1555
        }
1556
    }
1557

    
1558
    // Note the +1 is for the MMX scaler which reads over the end
1559
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1560
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1561

    
1562
    /* normalize & store in outFilter */
1563
    for (i=0; i<dstW; i++)
1564
    {
1565
        int j;
1566
        int64_t error=0;
1567
        int64_t sum=0;
1568

    
1569
        for (j=0; j<filterSize; j++)
1570
        {
1571
            sum+= filter[i*filterSize + j];
1572
        }
1573
        sum= (sum + one/2)/ one;
1574
        for (j=0; j<*outFilterSize; j++)
1575
        {
1576
            int64_t v= filter[i*filterSize + j] + error;
1577
            int intV= ROUNDED_DIV(v, sum);
1578
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1579
            error= v - intV*sum;
1580
        }
1581
    }
1582

    
1583
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1584
    for (i=0; i<*outFilterSize; i++)
1585
    {
1586
        int j= dstW*(*outFilterSize);
1587
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1588
    }
1589

    
1590
    ret=0;
1591
error:
1592
    av_free(filter);
1593
    av_free(filter2);
1594
    return ret;
1595
}
1596

    
1597
#ifdef COMPILE_MMX2
1598
static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1599
{
1600
    uint8_t *fragmentA;
1601
    x86_reg imm8OfPShufW1A;
1602
    x86_reg imm8OfPShufW2A;
1603
    x86_reg fragmentLengthA;
1604
    uint8_t *fragmentB;
1605
    x86_reg imm8OfPShufW1B;
1606
    x86_reg imm8OfPShufW2B;
1607
    x86_reg fragmentLengthB;
1608
    int fragmentPos;
1609

    
1610
    int xpos, i;
1611

    
1612
    // create an optimized horizontal scaling routine
1613

    
1614
    //code fragment
1615

    
1616
    __asm__ volatile(
1617
        "jmp                         9f                 \n\t"
1618
    // Begin
1619
        "0:                                             \n\t"
1620
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1621
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1622
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1623
        "punpcklbw                %%mm7, %%mm1          \n\t"
1624
        "punpcklbw                %%mm7, %%mm0          \n\t"
1625
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1626
        "1:                                             \n\t"
1627
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1628
        "2:                                             \n\t"
1629
        "psubw                    %%mm1, %%mm0          \n\t"
1630
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1631
        "pmullw                   %%mm3, %%mm0          \n\t"
1632
        "psllw                       $7, %%mm1          \n\t"
1633
        "paddw                    %%mm1, %%mm0          \n\t"
1634

    
1635
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1636

    
1637
        "add                         $8, %%"REG_a"      \n\t"
1638
    // End
1639
        "9:                                             \n\t"
1640
//        "int $3                                         \n\t"
1641
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1642
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1643
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1644
        "dec                         %1                 \n\t"
1645
        "dec                         %2                 \n\t"
1646
        "sub                         %0, %1             \n\t"
1647
        "sub                         %0, %2             \n\t"
1648
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1649
        "sub                         %0, %3             \n\t"
1650

    
1651

    
1652
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1653
        "=r" (fragmentLengthA)
1654
    );
1655

    
1656
    __asm__ volatile(
1657
        "jmp                         9f                 \n\t"
1658
    // Begin
1659
        "0:                                             \n\t"
1660
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1661
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1662
        "punpcklbw                %%mm7, %%mm0          \n\t"
1663
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1664
        "1:                                             \n\t"
1665
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1666
        "2:                                             \n\t"
1667
        "psubw                    %%mm1, %%mm0          \n\t"
1668
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1669
        "pmullw                   %%mm3, %%mm0          \n\t"
1670
        "psllw                       $7, %%mm1          \n\t"
1671
        "paddw                    %%mm1, %%mm0          \n\t"
1672

    
1673
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1674

    
1675
        "add                         $8, %%"REG_a"      \n\t"
1676
    // End
1677
        "9:                                             \n\t"
1678
//        "int                       $3                   \n\t"
1679
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1680
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1681
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1682
        "dec                         %1                 \n\t"
1683
        "dec                         %2                 \n\t"
1684
        "sub                         %0, %1             \n\t"
1685
        "sub                         %0, %2             \n\t"
1686
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1687
        "sub                         %0, %3             \n\t"
1688

    
1689

    
1690
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1691
        "=r" (fragmentLengthB)
1692
    );
1693

    
1694
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1695
    fragmentPos=0;
1696

    
1697
    for (i=0; i<dstW/numSplits; i++)
1698
    {
1699
        int xx=xpos>>16;
1700

    
1701
        if ((i&3) == 0)
1702
        {
1703
            int a=0;
1704
            int b=((xpos+xInc)>>16) - xx;
1705
            int c=((xpos+xInc*2)>>16) - xx;
1706
            int d=((xpos+xInc*3)>>16) - xx;
1707

    
1708
            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1709
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1710
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1711
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1712
            filterPos[i/2]= xx;
1713

    
1714
            if (d+1<4)
1715
            {
1716
                int maxShift= 3-(d+1);
1717
                int shift=0;
1718

    
1719
                memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1720

    
1721
                funnyCode[fragmentPos + imm8OfPShufW1B]=
1722
                    (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1723
                funnyCode[fragmentPos + imm8OfPShufW2B]=
1724
                    a | (b<<2) | (c<<4) | (d<<6);
1725

    
1726
                if (i+3>=dstW) shift=maxShift; //avoid overread
1727
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1728

    
1729
                if (shift && i>=shift)
1730
                {
1731
                    funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1732
                    funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1733
                    filterPos[i/2]-=shift;
1734
                }
1735

    
1736
                fragmentPos+= fragmentLengthB;
1737
            }
1738
            else
1739
            {
1740
                int maxShift= 3-d;
1741
                int shift=0;
1742

    
1743
                memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
1744

    
1745
                funnyCode[fragmentPos + imm8OfPShufW1A]=
1746
                funnyCode[fragmentPos + imm8OfPShufW2A]=
1747
                    a | (b<<2) | (c<<4) | (d<<6);
1748

    
1749
                if (i+4>=dstW) shift=maxShift; //avoid overread
1750
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
1751

    
1752
                if (shift && i>=shift)
1753
                {
1754
                    funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1755
                    funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1756
                    filterPos[i/2]-=shift;
1757
                }
1758

    
1759
                fragmentPos+= fragmentLengthA;
1760
            }
1761

    
1762
            funnyCode[fragmentPos]= RET;
1763
        }
1764
        xpos+=xInc;
1765
    }
1766
    filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1767
}
1768
#endif /* COMPILE_MMX2 */
1769

    
1770
static void globalInit(void){
1771
    // generating tables:
1772
    int i;
1773
    for (i=0; i<768; i++){
1774
        int c= av_clip_uint8(i-256);
1775
        clip_table[i]=c;
1776
    }
1777
}
1778

    
1779
static SwsFunc getSwsFunc(int flags){
1780

    
1781
#if defined(RUNTIME_CPUDETECT) && CONFIG_GPL
1782
#if ARCH_X86
1783
    // ordered per speed fastest first
1784
    if (flags & SWS_CPU_CAPS_MMX2)
1785
        return swScale_MMX2;
1786
    else if (flags & SWS_CPU_CAPS_3DNOW)
1787
        return swScale_3DNow;
1788
    else if (flags & SWS_CPU_CAPS_MMX)
1789
        return swScale_MMX;
1790
    else
1791
        return swScale_C;
1792

    
1793
#else
1794
#if ARCH_PPC
1795
    if (flags & SWS_CPU_CAPS_ALTIVEC)
1796
        return swScale_altivec;
1797
    else
1798
        return swScale_C;
1799
#endif
1800
    return swScale_C;
1801
#endif /* ARCH_X86 */
1802
#else //RUNTIME_CPUDETECT
1803
#if   HAVE_MMX2
1804
    return swScale_MMX2;
1805
#elif HAVE_AMD3DNOW
1806
    return swScale_3DNow;
1807
#elif HAVE_MMX
1808
    return swScale_MMX;
1809
#elif HAVE_ALTIVEC
1810
    return swScale_altivec;
1811
#else
1812
    return swScale_C;
1813
#endif
1814
#endif //!RUNTIME_CPUDETECT
1815
}
1816

    
1817
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1818
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1819
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1820
    /* Copy Y plane */
1821
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1822
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1823
    else
1824
    {
1825
        int i;
1826
        uint8_t *srcPtr= src[0];
1827
        uint8_t *dstPtr= dst;
1828
        for (i=0; i<srcSliceH; i++)
1829
        {
1830
            memcpy(dstPtr, srcPtr, c->srcW);
1831
            srcPtr+= srcStride[0];
1832
            dstPtr+= dstStride[0];
1833
        }
1834
    }
1835
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1836
    if (c->dstFormat == PIX_FMT_NV12)
1837
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1838
    else
1839
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1840

    
1841
    return srcSliceH;
1842
}
1843

    
1844
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1845
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1846
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1847

    
1848
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1849

    
1850
    return srcSliceH;
1851
}
1852

    
1853
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1854
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1855
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1856

    
1857
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1858

    
1859
    return srcSliceH;
1860
}
1861

    
1862
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1863
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1864
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1865

    
1866
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1867

    
1868
    return srcSliceH;
1869
}
1870

    
1871
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1872
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1873
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1874

    
1875
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1876

    
1877
    return srcSliceH;
1878
}
1879

    
1880
static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1881
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1882
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1883
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1884
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1885

    
1886
    yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1887

    
1888
    return srcSliceH;
1889
}
1890

    
1891
static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1892
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1893
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1894
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1895
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1896

    
1897
    yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1898

    
1899
    return srcSliceH;
1900
}
1901

    
1902
static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1903
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1904
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1905
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1906
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1907

    
1908
    uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1909

    
1910
    return srcSliceH;
1911
}
1912

    
1913
static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1914
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1915
    uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1916
    uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1917
    uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1918

    
1919
    uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1920

    
1921
    return srcSliceH;
1922
}
1923

    
1924
static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1925
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
1926
    const enum PixelFormat srcFormat= c->srcFormat;
1927
    const enum PixelFormat dstFormat= c->dstFormat;
1928
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1929
                 const uint8_t *palette)=NULL;
1930
    int i;
1931
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1932
    uint8_t *srcPtr= src[0];
1933

    
1934
    if (!usePal(srcFormat))
1935
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1936
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1937

    
1938
    switch(dstFormat){
1939
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
1940
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
1941
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1942
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1943
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
1944
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
1945
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1946
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1947
    }
1948

    
1949

    
1950
    for (i=0; i<srcSliceH; i++) {
1951
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1952
        srcPtr+= srcStride[0];
1953
        dstPtr+= dstStride[0];
1954
    }
1955

    
1956
    return srcSliceH;
1957
}
1958

    
1959
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1960
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1961
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
1962
    const enum PixelFormat srcFormat= c->srcFormat;
1963
    const enum PixelFormat dstFormat= c->dstFormat;
1964
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
1965
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
1966
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1967
    const int dstId= fmt_depth(dstFormat) >> 2;
1968
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1969

    
1970
    /* BGR -> BGR */
1971
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
1972
       || (isRGB(srcFormat) && isRGB(dstFormat))){
1973
        switch(srcId | (dstId<<4)){
1974
        case 0x34: conv= rgb16to15; break;
1975
        case 0x36: conv= rgb24to15; break;
1976
        case 0x38: conv= rgb32to15; break;
1977
        case 0x43: conv= rgb15to16; break;
1978
        case 0x46: conv= rgb24to16; break;
1979
        case 0x48: conv= rgb32to16; break;
1980
        case 0x63: conv= rgb15to24; break;
1981
        case 0x64: conv= rgb16to24; break;
1982
        case 0x68: conv= rgb32to24; break;
1983
        case 0x83: conv= rgb15to32; break;
1984
        case 0x84: conv= rgb16to32; break;
1985
        case 0x86: conv= rgb24to32; break;
1986
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1987
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1988
        }
1989
    }else if (  (isBGR(srcFormat) && isRGB(dstFormat))
1990
             || (isRGB(srcFormat) && isBGR(dstFormat))){
1991
        switch(srcId | (dstId<<4)){
1992
        case 0x33: conv= rgb15tobgr15; break;
1993
        case 0x34: conv= rgb16tobgr15; break;
1994
        case 0x36: conv= rgb24tobgr15; break;
1995
        case 0x38: conv= rgb32tobgr15; break;
1996
        case 0x43: conv= rgb15tobgr16; break;
1997
        case 0x44: conv= rgb16tobgr16; break;
1998
        case 0x46: conv= rgb24tobgr16; break;
1999
        case 0x48: conv= rgb32tobgr16; break;
2000
        case 0x63: conv= rgb15tobgr24; break;
2001
        case 0x64: conv= rgb16tobgr24; break;
2002
        case 0x66: conv= rgb24tobgr24; break;
2003
        case 0x68: conv= rgb32tobgr24; break;
2004
        case 0x83: conv= rgb15tobgr32; break;
2005
        case 0x84: conv= rgb16tobgr32; break;
2006
        case 0x86: conv= rgb24tobgr32; break;
2007
        case 0x88: conv= rgb32tobgr32; break;
2008
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2009
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2010
        }
2011
    }else{
2012
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2013
               sws_format_name(srcFormat), sws_format_name(dstFormat));
2014
    }
2015

    
2016
    if(conv)
2017
    {
2018
        uint8_t *srcPtr= src[0];
2019
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2020
            srcPtr += ALT32_CORR;
2021

    
2022
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2023
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2024
        else
2025
        {
2026
            int i;
2027
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2028

    
2029
            for (i=0; i<srcSliceH; i++)
2030
            {
2031
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
2032
                srcPtr+= srcStride[0];
2033
                dstPtr+= dstStride[0];
2034
            }
2035
        }
2036
    }
2037
    return srcSliceH;
2038
}
2039

    
2040
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2041
                              int srcSliceH, uint8_t* dst[], int dstStride[]){
2042

    
2043
    rgb24toyv12(
2044
        src[0],
2045
        dst[0]+ srcSliceY    *dstStride[0],
2046
        dst[1]+(srcSliceY>>1)*dstStride[1],
2047
        dst[2]+(srcSliceY>>1)*dstStride[2],
2048
        c->srcW, srcSliceH,
2049
        dstStride[0], dstStride[1], srcStride[0]);
2050
    return srcSliceH;
2051
}
2052

    
2053
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2054
                             int srcSliceH, uint8_t* dst[], int dstStride[]){
2055
    int i;
2056

    
2057
    /* copy Y */
2058
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2059
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2060
    else{
2061
        uint8_t *srcPtr= src[0];
2062
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2063

    
2064
        for (i=0; i<srcSliceH; i++)
2065
        {
2066
            memcpy(dstPtr, srcPtr, c->srcW);
2067
            srcPtr+= srcStride[0];
2068
            dstPtr+= dstStride[0];
2069
        }
2070
    }
2071

    
2072
    if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){
2073
        planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
2074
        planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
2075
    }else{
2076
        planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
2077
        planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
2078
    }
2079
    return srcSliceH;
2080
}
2081

    
2082
/* unscaled copy like stuff (assumes nearly identical formats) */
2083
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2084
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2085
{
2086
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2087
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2088
    else
2089
    {
2090
        int i;
2091
        uint8_t *srcPtr= src[0];
2092
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2093
        int length=0;
2094

    
2095
        /* universal length finder */
2096
        while(length+c->srcW <= FFABS(dstStride[0])
2097
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2098
        assert(length!=0);
2099

    
2100
        for (i=0; i<srcSliceH; i++)
2101
        {
2102
            memcpy(dstPtr, srcPtr, length);
2103
            srcPtr+= srcStride[0];
2104
            dstPtr+= dstStride[0];
2105
        }
2106
    }
2107
    return srcSliceH;
2108
}
2109

    
2110
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2111
                      int srcSliceH, uint8_t* dst[], int dstStride[])
2112
{
2113
    int plane;
2114
    for (plane=0; plane<4; plane++)
2115
    {
2116
        int length= (plane==0 || plane==3) ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
2117
        int y=      (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2118
        int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
2119

    
2120
        if (dst[plane] && !src[plane])
2121
            fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
2122
        else
2123
        {
2124
            if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2125
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2126
            else
2127
            {
2128
                int i;
2129
                uint8_t *srcPtr= src[plane];
2130
                uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2131
                for (i=0; i<height; i++)
2132
                {
2133
                    memcpy(dstPtr, srcPtr, length);
2134
                    srcPtr+= srcStride[plane];
2135
                    dstPtr+= dstStride[plane];
2136
                }
2137
            }
2138
        }
2139
    }
2140
    return srcSliceH;
2141
}
2142

    
2143
static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2144
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
2145

    
2146
    int length= c->srcW;
2147
    int y=      srcSliceY;
2148
    int height= srcSliceH;
2149
    int i, j;
2150
    uint8_t *srcPtr= src[0];
2151
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
2152

    
2153
    if (!isGray(c->dstFormat)){
2154
        int height= -((-srcSliceH)>>c->chrDstVSubSample);
2155
        memset(dst[1], 128, dstStride[1]*height);
2156
        memset(dst[2], 128, dstStride[2]*height);
2157
    }
2158
    if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;
2159
    for (i=0; i<height; i++)
2160
    {
2161
        for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2162
        srcPtr+= srcStride[0];
2163
        dstPtr+= dstStride[0];
2164
    }
2165
    return srcSliceH;
2166
}
2167

    
2168
static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2169
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
2170

    
2171
    int length= c->srcW;
2172
    int y=      srcSliceY;
2173
    int height= srcSliceH;
2174
    int i, j;
2175
    uint8_t *srcPtr= src[0];
2176
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
2177
    for (i=0; i<height; i++)
2178
    {
2179
        for (j=0; j<length; j++)
2180
        {
2181
            dstPtr[j<<1] = srcPtr[j];
2182
            dstPtr[(j<<1)+1] = srcPtr[j];
2183
        }
2184
        srcPtr+= srcStride[0];
2185
        dstPtr+= dstStride[0];
2186
    }
2187
    return srcSliceH;
2188
}
2189

    
2190
static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2191
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
2192

    
2193
    int length= c->srcW;
2194
    int y=      srcSliceY;
2195
    int height= srcSliceH;
2196
    int i, j;
2197
    uint16_t *srcPtr= (uint16_t*)src[0];
2198
    uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);
2199
    for (i=0; i<height; i++)
2200
    {
2201
        for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);
2202
        srcPtr+= srcStride[0]/2;
2203
        dstPtr+= dstStride[0]/2;
2204
    }
2205
    return srcSliceH;
2206
}
2207

    
2208

    
2209
static void getSubSampleFactors(int *h, int *v, int format){
2210
    switch(format){
2211
    case PIX_FMT_UYVY422:
2212
    case PIX_FMT_YUYV422:
2213
        *h=1;
2214
        *v=0;
2215
        break;
2216
    case PIX_FMT_YUV420P:
2217
    case PIX_FMT_YUVA420P:
2218
    case PIX_FMT_GRAY16BE:
2219
    case PIX_FMT_GRAY16LE:
2220
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2221
    case PIX_FMT_NV12:
2222
    case PIX_FMT_NV21:
2223
        *h=1;
2224
        *v=1;
2225
        break;
2226
    case PIX_FMT_YUV440P:
2227
        *h=0;
2228
        *v=1;
2229
        break;
2230
    case PIX_FMT_YUV410P:
2231
        *h=2;
2232
        *v=2;
2233
        break;
2234
    case PIX_FMT_YUV444P:
2235
        *h=0;
2236
        *v=0;
2237
        break;
2238
    case PIX_FMT_YUV422P:
2239
        *h=1;
2240
        *v=0;
2241
        break;
2242
    case PIX_FMT_YUV411P:
2243
        *h=2;
2244
        *v=0;
2245
        break;
2246
    default:
2247
        *h=0;
2248
        *v=0;
2249
        break;
2250
    }
2251
}
2252

    
2253
static uint16_t roundToInt16(int64_t f){
2254
    int r= (f + (1<<15))>>16;
2255
         if (r<-0x7FFF) return 0x8000;
2256
    else if (r> 0x7FFF) return 0x7FFF;
2257
    else                return r;
2258
}
2259

    
2260
/**
2261
 * @param inv_table the yuv2rgb coefficients, normally ff_yuv2rgb_coeffs[x]
2262
 * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235
2263
 * @return -1 if not supported
2264
 */
2265
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2266
    int64_t crv =  inv_table[0];
2267
    int64_t cbu =  inv_table[1];
2268
    int64_t cgu = -inv_table[2];
2269
    int64_t cgv = -inv_table[3];
2270
    int64_t cy  = 1<<16;
2271
    int64_t oy  = 0;
2272

    
2273
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2274
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2275

    
2276
    c->brightness= brightness;
2277
    c->contrast  = contrast;
2278
    c->saturation= saturation;
2279
    c->srcRange  = srcRange;
2280
    c->dstRange  = dstRange;
2281
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return 0;
2282

    
2283
    c->uOffset=   0x0400040004000400LL;
2284
    c->vOffset=   0x0400040004000400LL;
2285

    
2286
    if (!srcRange){
2287
        cy= (cy*255) / 219;
2288
        oy= 16<<16;
2289
    }else{
2290
        crv= (crv*224) / 255;
2291
        cbu= (cbu*224) / 255;
2292
        cgu= (cgu*224) / 255;
2293
        cgv= (cgv*224) / 255;
2294
    }
2295

    
2296
    cy = (cy *contrast             )>>16;
2297
    crv= (crv*contrast * saturation)>>32;
2298
    cbu= (cbu*contrast * saturation)>>32;
2299
    cgu= (cgu*contrast * saturation)>>32;
2300
    cgv= (cgv*contrast * saturation)>>32;
2301

    
2302
    oy -= 256*brightness;
2303

    
2304
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2305
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2306
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2307
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2308
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2309
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2310

    
2311
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2312
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2313
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2314
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2315
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2316
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2317

    
2318
    ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2319
    //FIXME factorize
2320

    
2321
#ifdef COMPILE_ALTIVEC
2322
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2323
        ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
2324
#endif
2325
    return 0;
2326
}
2327

    
2328
/**
2329
 * @return -1 if not supported
2330
 */
2331
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2332
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2333

    
2334
    *inv_table = c->srcColorspaceTable;
2335
    *table     = c->dstColorspaceTable;
2336
    *srcRange  = c->srcRange;
2337
    *dstRange  = c->dstRange;
2338
    *brightness= c->brightness;
2339
    *contrast  = c->contrast;
2340
    *saturation= c->saturation;
2341

    
2342
    return 0;
2343
}
2344

    
2345
static int handle_jpeg(enum PixelFormat *format)
2346
{
2347
    switch (*format) {
2348
        case PIX_FMT_YUVJ420P:
2349
            *format = PIX_FMT_YUV420P;
2350
            return 1;
2351
        case PIX_FMT_YUVJ422P:
2352
            *format = PIX_FMT_YUV422P;
2353
            return 1;
2354
        case PIX_FMT_YUVJ444P:
2355
            *format = PIX_FMT_YUV444P;
2356
            return 1;
2357
        case PIX_FMT_YUVJ440P:
2358
            *format = PIX_FMT_YUV440P;
2359
            return 1;
2360
        default:
2361
            return 0;
2362
    }
2363
}
2364

    
2365
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2366
                           SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){
2367

    
2368
    SwsContext *c;
2369
    int i;
2370
    int usesVFilter, usesHFilter;
2371
    int unscaled, needsDither;
2372
    int srcRange, dstRange;
2373
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2374
#if ARCH_X86
2375
    if (flags & SWS_CPU_CAPS_MMX)
2376
        __asm__ volatile("emms\n\t"::: "memory");
2377
#endif
2378

    
2379
#if !defined(RUNTIME_CPUDETECT) || !CONFIG_GPL //ensure that the flags match the compiled variant if cpudetect is off
2380
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2381
#if   HAVE_MMX2
2382
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2383
#elif HAVE_AMD3DNOW
2384
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2385
#elif HAVE_MMX
2386
    flags |= SWS_CPU_CAPS_MMX;
2387
#elif HAVE_ALTIVEC
2388
    flags |= SWS_CPU_CAPS_ALTIVEC;
2389
#elif ARCH_BFIN
2390
    flags |= SWS_CPU_CAPS_BFIN;
2391
#endif
2392
#endif /* RUNTIME_CPUDETECT */
2393
    if (clip_table[512] != 255) globalInit();
2394
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2395

    
2396
    unscaled = (srcW == dstW && srcH == dstH);
2397
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2398
        && (fmt_depth(dstFormat))<24
2399
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2400

    
2401
    srcRange = handle_jpeg(&srcFormat);
2402
    dstRange = handle_jpeg(&dstFormat);
2403

    
2404
    if (!isSupportedIn(srcFormat))
2405
    {
2406
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2407
        return NULL;
2408
    }
2409
    if (!isSupportedOut(dstFormat))
2410
    {
2411
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2412
        return NULL;
2413
    }
2414

    
2415
    i= flags & ( SWS_POINT
2416
                |SWS_AREA
2417
                |SWS_BILINEAR
2418
                |SWS_FAST_BILINEAR
2419
                |SWS_BICUBIC
2420
                |SWS_X
2421
                |SWS_GAUSS
2422
                |SWS_LANCZOS
2423
                |SWS_SINC
2424
                |SWS_SPLINE
2425
                |SWS_BICUBLIN);
2426
    if(!i || (i & (i-1)))
2427
    {
2428
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2429
        return NULL;
2430
    }
2431

    
2432
    /* sanity check */
2433
    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
2434
    {
2435
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2436
               srcW, srcH, dstW, dstH);
2437
        return NULL;
2438
    }
2439
    if(srcW > VOFW || dstW > VOFW){
2440
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2441
        return NULL;
2442
    }
2443

    
2444
    if (!dstFilter) dstFilter= &dummyFilter;
2445
    if (!srcFilter) srcFilter= &dummyFilter;
2446

    
2447
    c= av_mallocz(sizeof(SwsContext));
2448

    
2449
    c->av_class = &sws_context_class;
2450
    c->srcW= srcW;
2451
    c->srcH= srcH;
2452
    c->dstW= dstW;
2453
    c->dstH= dstH;
2454
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2455
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2456
    c->flags= flags;
2457
    c->dstFormat= dstFormat;
2458
    c->srcFormat= srcFormat;
2459
    c->vRounder= 4* 0x0001000100010001ULL;
2460

    
2461
    usesHFilter= usesVFilter= 0;
2462
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2463
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2464
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2465
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2466
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2467
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2468
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2469
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2470

    
2471
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2472
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2473

    
2474
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2475
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2476

    
2477
    // drop some chroma lines if the user wants it
2478
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2479
    c->chrSrcVSubSample+= c->vChrDrop;
2480

    
2481
    // drop every other pixel for chroma calculation unless user wants full chroma
2482
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2483
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2484
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2485
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2486
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2487
        c->chrSrcHSubSample=1;
2488

    
2489
    if (param){
2490
        c->param[0] = param[0];
2491
        c->param[1] = param[1];
2492
    }else{
2493
        c->param[0] =
2494
        c->param[1] = SWS_PARAM_DEFAULT;
2495
    }
2496

    
2497
    c->chrIntHSubSample= c->chrDstHSubSample;
2498
    c->chrIntVSubSample= c->chrSrcVSubSample;
2499

    
2500
    // Note the -((-x)>>y) is so that we always round toward +inf.
2501
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2502
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2503
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2504
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2505

    
2506
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2507

    
2508
    /* unscaled special cases */
2509
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2510
    {
2511
        /* yv12_to_nv12 */
2512
        if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2513
        {
2514
            c->swScale= PlanarToNV12Wrapper;
2515
        }
2516
        /* yuv2bgr */
2517
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2518
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1))
2519
        {
2520
            c->swScale= ff_yuv2rgb_get_func_ptr(c);
2521
        }
2522

    
2523
        if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))
2524
        {
2525
            c->swScale= yvu9toyv12Wrapper;
2526
        }
2527

    
2528
        /* bgr24toYV12 */
2529
        if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2530
            c->swScale= bgr24toyv12Wrapper;
2531

    
2532
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2533
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2534
           && (isBGR(dstFormat) || isRGB(dstFormat))
2535
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2536
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2537
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2538
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2539
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2540
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2541
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2542
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2543
                                             && dstFormat != PIX_FMT_RGB32_1
2544
                                             && dstFormat != PIX_FMT_BGR32_1
2545
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2546
             c->swScale= rgb2rgbWrapper;
2547

    
2548
        if ((usePal(srcFormat) && (
2549
                 dstFormat == PIX_FMT_RGB32   ||
2550
                 dstFormat == PIX_FMT_RGB32_1 ||
2551
                 dstFormat == PIX_FMT_RGB24   ||
2552
                 dstFormat == PIX_FMT_BGR32   ||
2553
                 dstFormat == PIX_FMT_BGR32_1 ||
2554
                 dstFormat == PIX_FMT_BGR24)))
2555
             c->swScale= pal2rgbWrapper;
2556

    
2557
        if (srcFormat == PIX_FMT_YUV422P)
2558
        {
2559
            if (dstFormat == PIX_FMT_YUYV422)
2560
                c->swScale= YUV422PToYuy2Wrapper;
2561
            else if (dstFormat == PIX_FMT_UYVY422)
2562
                c->swScale= YUV422PToUyvyWrapper;
2563
        }
2564

    
2565
        /* LQ converters if -sws 0 or -sws 4*/
2566
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2567
            /* yv12_to_yuy2 */
2568
            if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)
2569
            {
2570
                if (dstFormat == PIX_FMT_YUYV422)
2571
                    c->swScale= PlanarToYuy2Wrapper;
2572
                else if (dstFormat == PIX_FMT_UYVY422)
2573
                    c->swScale= PlanarToUyvyWrapper;
2574
            }
2575

    
2576
            if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV420P)
2577
                c->swScale= YUYV2YUV420Wrapper;
2578
            if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV420P)
2579
                c->swScale= UYVY2YUV420Wrapper;
2580
        }
2581
        if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2582
            c->swScale= YUYV2YUV422Wrapper;
2583
        if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2584
            c->swScale= UYVY2YUV422Wrapper;
2585

    
2586
#ifdef COMPILE_ALTIVEC
2587
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2588
            !(c->flags & SWS_BITEXACT) &&
2589
            srcFormat == PIX_FMT_YUV420P) {
2590
          // unscaled YV12 -> packed YUV, we want speed
2591
          if (dstFormat == PIX_FMT_YUYV422)
2592
              c->swScale= yv12toyuy2_unscaled_altivec;
2593
          else if (dstFormat == PIX_FMT_UYVY422)
2594
              c->swScale= yv12touyvy_unscaled_altivec;
2595
        }
2596
#endif
2597

    
2598
        /* simple copy */
2599
        if (  srcFormat == dstFormat
2600
            || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2601
            || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2602
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2603
            || (isPlanarYUV(dstFormat) && isGray(srcFormat)))
2604
        {
2605
            if (isPacked(c->srcFormat))
2606
                c->swScale= packedCopy;
2607
            else /* Planar YUV or gray */
2608
                c->swScale= planarCopy;
2609
        }
2610

    
2611
        /* gray16{le,be} conversions */
2612
        if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))
2613
        {
2614
            c->swScale= gray16togray;
2615
        }
2616
        if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))
2617
        {
2618
            c->swScale= graytogray16;
2619
        }
2620
        if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
2621
        {
2622
            c->swScale= gray16swap;
2623
        }
2624

    
2625
#if ARCH_BFIN
2626
        if (flags & SWS_CPU_CAPS_BFIN)
2627
            ff_bfin_get_unscaled_swscale (c);
2628
#endif
2629

    
2630
        if (c->swScale){
2631
            if (flags&SWS_PRINT_INFO)
2632
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2633
                                sws_format_name(srcFormat), sws_format_name(dstFormat));
2634
            return c;
2635
        }
2636
    }
2637

    
2638
    if (flags & SWS_CPU_CAPS_MMX2)
2639
    {
2640
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2641
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2642
        {
2643
            if (flags&SWS_PRINT_INFO)
2644
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2645
        }
2646
        if (usesHFilter) c->canMMX2BeUsed=0;
2647
    }
2648
    else
2649
        c->canMMX2BeUsed=0;
2650

    
2651
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2652
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2653

    
2654
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2655
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2656
    // n-2 is the last chrominance sample available
2657
    // this is not perfect, but no one should notice the difference, the more correct variant
2658
    // would be like the vertical one, but that would require some special code for the
2659
    // first and last pixel
2660
    if (flags&SWS_FAST_BILINEAR)
2661
    {
2662
        if (c->canMMX2BeUsed)
2663
        {
2664
            c->lumXInc+= 20;
2665
            c->chrXInc+= 20;
2666
        }
2667
        //we don't use the x86 asm scaler if MMX is available
2668
        else if (flags & SWS_CPU_CAPS_MMX)
2669
        {
2670
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2671
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2672
        }
2673
    }
2674

    
2675
    /* precalculate horizontal scaler filter coefficients */
2676
    {
2677
        const int filterAlign=
2678
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2679
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2680
            1;
2681

    
2682
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2683
                   srcW      ,       dstW, filterAlign, 1<<14,
2684
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2685
                   srcFilter->lumH, dstFilter->lumH, c->param);
2686
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2687
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2688
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2689
                   srcFilter->chrH, dstFilter->chrH, c->param);
2690

    
2691
#define MAX_FUNNY_CODE_SIZE 10000
2692
#if defined(COMPILE_MMX2)
2693
// can't downscale !!!
2694
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2695
        {
2696
#ifdef MAP_ANONYMOUS
2697
            c->funnyYCode  = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2698
            c->funnyUVCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2699
#elif HAVE_VIRTUALALLOC
2700
            c->funnyYCode  = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2701
            c->funnyUVCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2702
#else
2703
            c->funnyYCode  = av_malloc(MAX_FUNNY_CODE_SIZE);
2704
            c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2705
#endif
2706

    
2707
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2708
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2709
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2710
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2711

    
2712
            initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2713
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2714
        }
2715
#endif /* defined(COMPILE_MMX2) */
2716
    } // initialize horizontal stuff
2717

    
2718

    
2719

    
2720
    /* precalculate vertical scaler filter coefficients */
2721
    {
2722
        const int filterAlign=
2723
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2724
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2725
            1;
2726

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

    
2736
#if HAVE_ALTIVEC
2737
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2738
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2739

    
2740
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2741
            int j;
2742
            short *p = (short *)&c->vYCoeffsBank[i];
2743
            for (j=0;j<8;j++)
2744
                p[j] = c->vLumFilter[i];
2745
        }
2746

    
2747
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2748
            int j;
2749
            short *p = (short *)&c->vCCoeffsBank[i];
2750
            for (j=0;j<8;j++)
2751
                p[j] = c->vChrFilter[i];
2752
        }
2753
#endif
2754
    }
2755

    
2756
    // calculate buffer sizes so that they won't run out while handling these damn slices
2757
    c->vLumBufSize= c->vLumFilterSize;
2758
    c->vChrBufSize= c->vChrFilterSize;
2759
    for (i=0; i<dstH; i++)
2760
    {
2761
        int chrI= i*c->chrDstH / dstH;
2762
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2763
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2764

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

    
2773
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2774
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2775
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2776
    if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2777
        c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2778
    //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)
2779
    /* align at 16 bytes for AltiVec */
2780
    for (i=0; i<c->vLumBufSize; i++)
2781
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2782
    for (i=0; i<c->vChrBufSize; i++)
2783
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2784
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2785
        for (i=0; i<c->vLumBufSize; i++)
2786
            c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2787

    
2788
    //try to avoid drawing green stuff between the right end and the stride end
2789
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2790

    
2791
    assert(2*VOFW == VOF);
2792

    
2793
    assert(c->chrDstH <= dstH);
2794

    
2795
    if (flags&SWS_PRINT_INFO)
2796
    {
2797
#ifdef DITHER1XBPP
2798
        const char *dither= " dithered";
2799
#else
2800
        const char *dither= "";
2801
#endif
2802
        if (flags&SWS_FAST_BILINEAR)
2803
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2804
        else if (flags&SWS_BILINEAR)
2805
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2806
        else if (flags&SWS_BICUBIC)
2807
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2808
        else if (flags&SWS_X)
2809
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2810
        else if (flags&SWS_POINT)
2811
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2812
        else if (flags&SWS_AREA)
2813
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2814
        else if (flags&SWS_BICUBLIN)
2815
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2816
        else if (flags&SWS_GAUSS)
2817
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2818
        else if (flags&SWS_SINC)
2819
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2820
        else if (flags&SWS_LANCZOS)
2821
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2822
        else if (flags&SWS_SPLINE)
2823
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2824
        else
2825
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2826

    
2827
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2828
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2829
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2830
        else
2831
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2832
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2833

    
2834
        if (flags & SWS_CPU_CAPS_MMX2)
2835
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2836
        else if (flags & SWS_CPU_CAPS_3DNOW)
2837
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2838
        else if (flags & SWS_CPU_CAPS_MMX)
2839
            av_log(c, AV_LOG_INFO, "using MMX\n");
2840
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2841
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2842
        else
2843
            av_log(c, AV_LOG_INFO, "using C\n");
2844
    }
2845

    
2846
    if (flags & SWS_PRINT_INFO)
2847
    {
2848
        if (flags & SWS_CPU_CAPS_MMX)
2849
        {
2850
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2851
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2852
            else
2853
            {
2854
                if (c->hLumFilterSize==4)
2855
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2856
                else if (c->hLumFilterSize==8)
2857
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2858
                else
2859
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2860

    
2861
                if (c->hChrFilterSize==4)
2862
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2863
                else if (c->hChrFilterSize==8)
2864
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2865
                else
2866
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2867
            }
2868
        }
2869
        else
2870
        {
2871
#if ARCH_X86
2872
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
2873
#else
2874
            if (flags & SWS_FAST_BILINEAR)
2875
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
2876
            else
2877
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
2878
#endif
2879
        }
2880
        if (isPlanarYUV(dstFormat))
2881
        {
2882
            if (c->vLumFilterSize==1)
2883
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2884
            else
2885
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2886
        }
2887
        else
2888
        {
2889
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
2890
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2891
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2892
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
2893
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2894
            else
2895
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2896
        }
2897

    
2898
        if (dstFormat==PIX_FMT_BGR24)
2899
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
2900
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2901
        else if (dstFormat==PIX_FMT_RGB32)
2902
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2903
        else if (dstFormat==PIX_FMT_BGR565)
2904
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2905
        else if (dstFormat==PIX_FMT_BGR555)
2906
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2907

    
2908
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2909
    }
2910
    if (flags & SWS_PRINT_INFO)
2911
    {
2912
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2913
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2914
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2915
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2916
    }
2917

    
2918
    c->swScale= getSwsFunc(flags);
2919
    return c;
2920
}
2921

    
2922
/**
2923
 * swscale wrapper, so we don't need to export the SwsContext.
2924
 * Assumes planar YUV to be in YUV order instead of YVU.
2925
 */
2926
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2927
              int srcSliceH, uint8_t* dst[], int dstStride[]){
2928
    int i;
2929
    uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
2930

    
2931
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2932
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2933
        return 0;
2934
    }
2935
    if (c->sliceDir == 0) {
2936
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2937
    }
2938

    
2939
    if (usePal(c->srcFormat)){
2940
        for (i=0; i<256; i++){
2941
            int p, r, g, b,y,u,v;
2942
            if(c->srcFormat == PIX_FMT_PAL8){
2943
                p=((uint32_t*)(src[1]))[i];
2944
                r= (p>>16)&0xFF;
2945
                g= (p>> 8)&0xFF;
2946
                b=  p     &0xFF;
2947
            }else if(c->srcFormat == PIX_FMT_RGB8){
2948
                r= (i>>5    )*36;
2949
                g= ((i>>2)&7)*36;
2950
                b= (i&3     )*85;
2951
            }else if(c->srcFormat == PIX_FMT_BGR8){
2952
                b= (i>>6    )*85;
2953
                g= ((i>>3)&7)*36;
2954
                r= (i&7     )*36;
2955
            }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
2956
                r= (i>>3    )*255;
2957
                g= ((i>>1)&3)*85;
2958
                b= (i&1     )*255;
2959
            }else {
2960
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2961
                b= (i>>3    )*255;
2962
                g= ((i>>1)&3)*85;
2963
                r= (i&1     )*255;
2964
            }
2965
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2966
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2967
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2968
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
2969

    
2970

    
2971
            switch(c->dstFormat) {
2972
            case PIX_FMT_BGR32:
2973
#ifndef WORDS_BIGENDIAN
2974
            case PIX_FMT_RGB24:
2975
#endif
2976
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
2977
                break;
2978
            case PIX_FMT_BGR32_1:
2979
#ifdef  WORDS_BIGENDIAN
2980
            case PIX_FMT_BGR24:
2981
#endif
2982
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2983
                break;
2984
            case PIX_FMT_RGB32_1:
2985
#ifdef  WORDS_BIGENDIAN
2986
            case PIX_FMT_RGB24:
2987
#endif
2988
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2989
                break;
2990
            case PIX_FMT_RGB32:
2991
#ifndef WORDS_BIGENDIAN
2992
            case PIX_FMT_BGR24:
2993
#endif
2994
            default:
2995
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
2996
            }
2997
        }
2998
    }
2999

    
3000
    // copy strides, so they can safely be modified
3001
    if (c->sliceDir == 1) {
3002
        // slices go from top to bottom
3003
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3004
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3005
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
3006
    } else {
3007
        // slices go from bottom to top => we flip the image internally
3008
        uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],
3009
                           dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],
3010
                           dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2],
3011
                           dst[3] + (c->dstH-1)*dstStride[3]};
3012
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3013
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3014

    
3015
        src2[0] += (srcSliceH-1)*srcStride[0];
3016
        if (!usePal(c->srcFormat))
3017
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3018
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3019
        src2[3] += (srcSliceH-1)*srcStride[3];
3020

    
3021
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3022
    }
3023
}
3024

    
3025
#if LIBSWSCALE_VERSION_MAJOR < 1
3026
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3027
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
3028
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3029
}
3030
#endif
3031

    
3032
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3033
                                float lumaSharpen, float chromaSharpen,
3034
                                float chromaHShift, float chromaVShift,
3035
                                int verbose)
3036
{
3037
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3038

    
3039
    if (lumaGBlur!=0.0){
3040
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3041
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3042
    }else{
3043
        filter->lumH= sws_getIdentityVec();
3044
        filter->lumV= sws_getIdentityVec();
3045
    }
3046

    
3047
    if (chromaGBlur!=0.0){
3048
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3049
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3050
    }else{
3051
        filter->chrH= sws_getIdentityVec();
3052
        filter->chrV= sws_getIdentityVec();
3053
    }
3054

    
3055
    if (chromaSharpen!=0.0){
3056
        SwsVector *id= sws_getIdentityVec();
3057
        sws_scaleVec(filter->chrH, -chromaSharpen);
3058
        sws_scaleVec(filter->chrV, -chromaSharpen);
3059
        sws_addVec(filter->chrH, id);
3060
        sws_addVec(filter->chrV, id);
3061
        sws_freeVec(id);
3062
    }
3063

    
3064
    if (lumaSharpen!=0.0){
3065
        SwsVector *id= sws_getIdentityVec();
3066
        sws_scaleVec(filter->lumH, -lumaSharpen);
3067
        sws_scaleVec(filter->lumV, -lumaSharpen);
3068
        sws_addVec(filter->lumH, id);
3069
        sws_addVec(filter->lumV, id);
3070
        sws_freeVec(id);
3071
    }
3072

    
3073
    if (chromaHShift != 0.0)
3074
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3075

    
3076
    if (chromaVShift != 0.0)
3077
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3078

    
3079
    sws_normalizeVec(filter->chrH, 1.0);
3080
    sws_normalizeVec(filter->chrV, 1.0);
3081
    sws_normalizeVec(filter->lumH, 1.0);
3082
    sws_normalizeVec(filter->lumV, 1.0);
3083

    
3084
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3085
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3086

    
3087
    return filter;
3088
}
3089

    
3090
SwsVector *sws_getGaussianVec(double variance, double quality){
3091
    const int length= (int)(variance*quality + 0.5) | 1;
3092
    int i;
3093
    double *coeff= av_malloc(length*sizeof(double));
3094
    double middle= (length-1)*0.5;
3095
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3096

    
3097
    vec->coeff= coeff;
3098
    vec->length= length;
3099

    
3100
    for (i=0; i<length; i++)
3101
    {
3102
        double dist= i-middle;
3103
        coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3104
    }
3105

    
3106
    sws_normalizeVec(vec, 1.0);
3107

    
3108
    return vec;
3109
}
3110

    
3111
SwsVector *sws_getConstVec(double c, int length){
3112
    int i;
3113
    double *coeff= av_malloc(length*sizeof(double));
3114
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3115

    
3116
    vec->coeff= coeff;
3117
    vec->length= length;
3118

    
3119
    for (i=0; i<length; i++)
3120
        coeff[i]= c;
3121

    
3122
    return vec;
3123
}
3124

    
3125

    
3126
SwsVector *sws_getIdentityVec(void){
3127
    return sws_getConstVec(1.0, 1);
3128
}
3129

    
3130
double sws_dcVec(SwsVector *a){
3131
    int i;
3132
    double sum=0;
3133

    
3134
    for (i=0; i<a->length; i++)
3135
        sum+= a->coeff[i];
3136

    
3137
    return sum;
3138
}
3139

    
3140
void sws_scaleVec(SwsVector *a, double scalar){
3141
    int i;
3142

    
3143
    for (i=0; i<a->length; i++)
3144
        a->coeff[i]*= scalar;
3145
}
3146

    
3147
void sws_normalizeVec(SwsVector *a, double height){
3148
    sws_scaleVec(a, height/sws_dcVec(a));
3149
}
3150

    
3151
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
3152
    int length= a->length + b->length - 1;
3153
    double *coeff= av_malloc(length*sizeof(double));
3154
    int i, j;
3155
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3156

    
3157
    vec->coeff= coeff;
3158
    vec->length= length;
3159

    
3160
    for (i=0; i<length; i++) coeff[i]= 0.0;
3161

    
3162
    for (i=0; i<a->length; i++)
3163
    {
3164
        for (j=0; j<b->length; j++)
3165
        {
3166
            coeff[i+j]+= a->coeff[i]*b->coeff[j];
3167
        }
3168
    }
3169

    
3170
    return vec;
3171
}
3172

    
3173
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
3174
    int length= FFMAX(a->length, b->length);
3175
    double *coeff= av_malloc(length*sizeof(double));
3176
    int i;
3177
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3178

    
3179
    vec->coeff= coeff;
3180
    vec->length= length;
3181

    
3182
    for (i=0; i<length; i++) coeff[i]= 0.0;
3183

    
3184
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3185
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3186

    
3187
    return vec;
3188
}
3189

    
3190
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
3191
    int length= FFMAX(a->length, b->length);
3192
    double *coeff= av_malloc(length*sizeof(double));
3193
    int i;
3194
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3195

    
3196
    vec->coeff= coeff;
3197
    vec->length= length;
3198

    
3199
    for (i=0; i<length; i++) coeff[i]= 0.0;
3200

    
3201
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3202
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3203

    
3204
    return vec;
3205
}
3206

    
3207
/* shift left / or right if "shift" is negative */
3208
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
3209
    int length= a->length + FFABS(shift)*2;
3210
    double *coeff= av_malloc(length*sizeof(double));
3211
    int i;
3212
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3213

    
3214
    vec->coeff= coeff;
3215
    vec->length= length;
3216

    
3217
    for (i=0; i<length; i++) coeff[i]= 0.0;
3218

    
3219
    for (i=0; i<a->length; i++)
3220
    {
3221
        coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3222
    }
3223

    
3224
    return vec;
3225
}
3226

    
3227
void sws_shiftVec(SwsVector *a, int shift){
3228
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3229
    av_free(a->coeff);
3230
    a->coeff= shifted->coeff;
3231
    a->length= shifted->length;
3232
    av_free(shifted);
3233
}
3234

    
3235
void sws_addVec(SwsVector *a, SwsVector *b){
3236
    SwsVector *sum= sws_sumVec(a, b);
3237
    av_free(a->coeff);
3238
    a->coeff= sum->coeff;
3239
    a->length= sum->length;
3240
    av_free(sum);
3241
}
3242

    
3243
void sws_subVec(SwsVector *a, SwsVector *b){
3244
    SwsVector *diff= sws_diffVec(a, b);
3245
    av_free(a->coeff);
3246
    a->coeff= diff->coeff;
3247
    a->length= diff->length;
3248
    av_free(diff);
3249
}
3250

    
3251
void sws_convVec(SwsVector *a, SwsVector *b){
3252
    SwsVector *conv= sws_getConvVec(a, b);
3253
    av_free(a->coeff);
3254
    a->coeff= conv->coeff;
3255
    a->length= conv->length;
3256
    av_free(conv);
3257
}
3258

    
3259
SwsVector *sws_cloneVec(SwsVector *a){
3260
    double *coeff= av_malloc(a->length*sizeof(double));
3261
    int i;
3262
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3263

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

    
3267
    for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
3268

    
3269
    return vec;
3270
}
3271

    
3272
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){
3273
    int i;
3274
    double max=0;
3275
    double min=0;
3276
    double range;
3277

    
3278
    for (i=0; i<a->length; i++)
3279
        if (a->coeff[i]>max) max= a->coeff[i];
3280

    
3281
    for (i=0; i<a->length; i++)
3282
        if (a->coeff[i]<min) min= a->coeff[i];
3283

    
3284
    range= max - min;
3285

    
3286
    for (i=0; i<a->length; i++)
3287
    {
3288
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3289
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3290
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3291
        av_log(log_ctx, log_level, "|\n");
3292
    }
3293
}
3294

    
3295
#if LIBSWSCALE_VERSION_MAJOR < 1
3296
void sws_printVec(SwsVector *a){
3297
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3298
}
3299
#endif
3300

    
3301
void sws_freeVec(SwsVector *a){
3302
    if (!a) return;
3303
    av_freep(&a->coeff);
3304
    a->length=0;
3305
    av_free(a);
3306
}
3307

    
3308
void sws_freeFilter(SwsFilter *filter){
3309
    if (!filter) return;
3310

    
3311
    if (filter->lumH) sws_freeVec(filter->lumH);
3312
    if (filter->lumV) sws_freeVec(filter->lumV);
3313
    if (filter->chrH) sws_freeVec(filter->chrH);
3314
    if (filter->chrV) sws_freeVec(filter->chrV);
3315
    av_free(filter);
3316
}
3317

    
3318

    
3319
void sws_freeContext(SwsContext *c){
3320
    int i;
3321
    if (!c) return;
3322

    
3323
    if (c->lumPixBuf)
3324
    {
3325
        for (i=0; i<c->vLumBufSize; i++)
3326
            av_freep(&c->lumPixBuf[i]);
3327
        av_freep(&c->lumPixBuf);
3328
    }
3329

    
3330
    if (c->chrPixBuf)
3331
    {
3332
        for (i=0; i<c->vChrBufSize; i++)
3333
            av_freep(&c->chrPixBuf[i]);
3334
        av_freep(&c->chrPixBuf);
3335
    }
3336

    
3337
    if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
3338
        for (i=0; i<c->vLumBufSize; i++)
3339
            av_freep(&c->alpPixBuf[i]);
3340
        av_freep(&c->alpPixBuf);
3341
    }
3342

    
3343
    av_freep(&c->vLumFilter);
3344
    av_freep(&c->vChrFilter);
3345
    av_freep(&c->hLumFilter);
3346
    av_freep(&c->hChrFilter);
3347
#if HAVE_ALTIVEC
3348
    av_freep(&c->vYCoeffsBank);
3349
    av_freep(&c->vCCoeffsBank);
3350
#endif
3351

    
3352
    av_freep(&c->vLumFilterPos);
3353
    av_freep(&c->vChrFilterPos);
3354
    av_freep(&c->hLumFilterPos);
3355
    av_freep(&c->hChrFilterPos);
3356

    
3357
#if ARCH_X86 && CONFIG_GPL
3358
#ifdef MAP_ANONYMOUS
3359
    if (c->funnyYCode ) munmap(c->funnyYCode , MAX_FUNNY_CODE_SIZE);
3360
    if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
3361
#elif HAVE_VIRTUALALLOC
3362
    if (c->funnyYCode ) VirtualFree(c->funnyYCode , MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3363
    if (c->funnyUVCode) VirtualFree(c->funnyUVCode, MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3364
#else
3365
    av_free(c->funnyYCode );
3366
    av_free(c->funnyUVCode);
3367
#endif
3368
    c->funnyYCode=NULL;
3369
    c->funnyUVCode=NULL;
3370
#endif /* ARCH_X86 && CONFIG_GPL */
3371

    
3372
    av_freep(&c->lumMmx2Filter);
3373
    av_freep(&c->chrMmx2Filter);
3374
    av_freep(&c->lumMmx2FilterPos);
3375
    av_freep(&c->chrMmx2FilterPos);
3376
    av_freep(&c->yuvTable);
3377

    
3378
    av_free(c);
3379
}
3380

    
3381
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3382
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3383
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3384
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)
3385
{
3386
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3387

    
3388
    if (!param)
3389
        param = default_param;
3390

    
3391
    if (context) {
3392
        if (context->srcW != srcW || context->srcH != srcH ||
3393
            context->srcFormat != srcFormat ||
3394
            context->dstW != dstW || context->dstH != dstH ||
3395
            context->dstFormat != dstFormat || context->flags != flags ||
3396
            context->param[0] != param[0] || context->param[1] != param[1])
3397
        {
3398
            sws_freeContext(context);
3399
            context = NULL;
3400
        }
3401
    }
3402
    if (!context) {
3403
        return sws_getContext(srcW, srcH, srcFormat,
3404
                              dstW, dstH, dstFormat, flags,
3405
                              srcFilter, dstFilter, param);
3406
    }
3407
    return context;
3408
}
3409