Statistics
| Branch: | Revision:

ffmpeg / libswscale / swscale.c @ e76709d8

History | View | Annotate | Download (107 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
#include "swscale.h"
72
#include "swscale_internal.h"
73
#include "rgb2rgb.h"
74
#include "libavutil/x86_cpu.h"
75
#include "libavutil/bswap.h"
76

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

    
82
#undef MOVNTQ
83
#undef PAVGB
84

    
85
//#undef HAVE_MMX2
86
//#define HAVE_AMD3DNOW
87
//#undef HAVE_MMX
88
//#undef ARCH_X86
89
//#define WORDS_BIGENDIAN
90
#define DITHER1XBPP
91

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

    
94
#define RET 0xC3 //near return opcode for x86
95

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

    
102
#define isSupportedIn(x)    (       \
103
           (x)==PIX_FMT_YUV420P     \
104
        || (x)==PIX_FMT_YUVA420P    \
105
        || (x)==PIX_FMT_YUYV422     \
106
        || (x)==PIX_FMT_UYVY422     \
107
        || (x)==PIX_FMT_RGB32       \
108
        || (x)==PIX_FMT_RGB32_1     \
109
        || (x)==PIX_FMT_BGR24       \
110
        || (x)==PIX_FMT_BGR565      \
111
        || (x)==PIX_FMT_BGR555      \
112
        || (x)==PIX_FMT_BGR32       \
113
        || (x)==PIX_FMT_BGR32_1     \
114
        || (x)==PIX_FMT_RGB24       \
115
        || (x)==PIX_FMT_RGB565      \
116
        || (x)==PIX_FMT_RGB555      \
117
        || (x)==PIX_FMT_GRAY8       \
118
        || (x)==PIX_FMT_YUV410P     \
119
        || (x)==PIX_FMT_YUV440P     \
120
        || (x)==PIX_FMT_GRAY16BE    \
121
        || (x)==PIX_FMT_GRAY16LE    \
122
        || (x)==PIX_FMT_YUV444P     \
123
        || (x)==PIX_FMT_YUV422P     \
124
        || (x)==PIX_FMT_YUV411P     \
125
        || (x)==PIX_FMT_PAL8        \
126
        || (x)==PIX_FMT_BGR8        \
127
        || (x)==PIX_FMT_RGB8        \
128
        || (x)==PIX_FMT_BGR4_BYTE   \
129
        || (x)==PIX_FMT_RGB4_BYTE   \
130
        || (x)==PIX_FMT_YUV440P     \
131
        || (x)==PIX_FMT_MONOWHITE   \
132
        || (x)==PIX_FMT_MONOBLACK   \
133
    )
134
#define isSupportedOut(x)   (       \
135
           (x)==PIX_FMT_YUV420P     \
136
        || (x)==PIX_FMT_YUYV422     \
137
        || (x)==PIX_FMT_UYVY422     \
138
        || (x)==PIX_FMT_YUV444P     \
139
        || (x)==PIX_FMT_YUV422P     \
140
        || (x)==PIX_FMT_YUV411P     \
141
        || isRGB(x)                 \
142
        || isBGR(x)                 \
143
        || (x)==PIX_FMT_NV12        \
144
        || (x)==PIX_FMT_NV21        \
145
        || (x)==PIX_FMT_GRAY16BE    \
146
        || (x)==PIX_FMT_GRAY16LE    \
147
        || (x)==PIX_FMT_GRAY8       \
148
        || (x)==PIX_FMT_YUV410P     \
149
        || (x)==PIX_FMT_YUV440P     \
150
    )
151
#define isPacked(x)         (       \
152
           (x)==PIX_FMT_PAL8        \
153
        || (x)==PIX_FMT_YUYV422     \
154
        || (x)==PIX_FMT_UYVY422     \
155
        || isRGB(x)                 \
156
        || isBGR(x)                 \
157
    )
158
#define usePal(x)           (       \
159
           (x)==PIX_FMT_PAL8        \
160
        || (x)==PIX_FMT_BGR4_BYTE   \
161
        || (x)==PIX_FMT_RGB4_BYTE   \
162
        || (x)==PIX_FMT_BGR8        \
163
        || (x)==PIX_FMT_RGB8        \
164
    )
165

    
166
#define RGB2YUV_SHIFT 15
167
#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
168
#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
169
#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
170
#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
171
#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
172
#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
173
#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
174
#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
175
#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
176

    
177
extern const int32_t ff_yuv2rgb_coeffs[8][4];
178

    
179
static const double rgb2yuv_table[8][9]={
180
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
181
    {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
182
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
183
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
184
    {0.59  , 0.11  , 0.30  , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
185
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
186
    {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
187
    {0.701 , 0.087 , 0.212 , -0.384, 0.5  -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
188
};
189

    
190
/*
191
NOTES
192
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
193

194
TODO
195
more intelligent misalignment avoidance for the horizontal scaler
196
write special vertical cubic upscale version
197
optimize C code (YV12 / minmax)
198
add support for packed pixel YUV input & output
199
add support for Y8 output
200
optimize BGR24 & BGR32
201
add BGR4 output support
202
write special BGR->BGR scaler
203
*/
204

    
205
#if ARCH_X86 && CONFIG_GPL
206
DECLARE_ASM_CONST(8, uint64_t, bF8)=       0xF8F8F8F8F8F8F8F8LL;
207
DECLARE_ASM_CONST(8, uint64_t, bFC)=       0xFCFCFCFCFCFCFCFCLL;
208
DECLARE_ASM_CONST(8, uint64_t, w10)=       0x0010001000100010LL;
209
DECLARE_ASM_CONST(8, uint64_t, w02)=       0x0002000200020002LL;
210
DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
211
DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
212
DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
213
DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
214

    
215
const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
216
        0x0103010301030103LL,
217
        0x0200020002000200LL,};
218

    
219
const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
220
        0x0602060206020602LL,
221
        0x0004000400040004LL,};
222

    
223
DECLARE_ASM_CONST(8, uint64_t, b16Mask)=   0x001F001F001F001FLL;
224
DECLARE_ASM_CONST(8, uint64_t, g16Mask)=   0x07E007E007E007E0LL;
225
DECLARE_ASM_CONST(8, uint64_t, r16Mask)=   0xF800F800F800F800LL;
226
DECLARE_ASM_CONST(8, uint64_t, b15Mask)=   0x001F001F001F001FLL;
227
DECLARE_ASM_CONST(8, uint64_t, g15Mask)=   0x03E003E003E003E0LL;
228
DECLARE_ASM_CONST(8, uint64_t, r15Mask)=   0x7C007C007C007C00LL;
229

    
230
DECLARE_ALIGNED(8, const uint64_t, ff_M24A)         = 0x00FF0000FF0000FFLL;
231
DECLARE_ALIGNED(8, const uint64_t, ff_M24B)         = 0xFF0000FF0000FF00LL;
232
DECLARE_ALIGNED(8, const uint64_t, ff_M24C)         = 0x0000FF0000FF0000LL;
233

    
234
#ifdef FAST_BGR2YV12
235
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000000210041000DULL;
236
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000FFEEFFDC0038ULL;
237
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00000038FFD2FFF8ULL;
238
#else
239
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff)   = 0x000020E540830C8BULL;
240
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff)   = 0x0000ED0FDAC23831ULL;
241
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff)   = 0x00003831D0E6F6EAULL;
242
#endif /* FAST_BGR2YV12 */
243
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset)  = 0x1010101010101010ULL;
244
DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
245
DECLARE_ALIGNED(8, const uint64_t, ff_w1111)        = 0x0001000100010001ULL;
246

    
247
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
248
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
249
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
250
DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
251
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
252

    
253
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
254
    {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
255
    {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
256
};
257

    
258
DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
259

    
260
#endif /* ARCH_X86 && CONFIG_GPL */
261

    
262
// clipping helper table for C implementations:
263
static unsigned char clip_table[768];
264

    
265
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
266

    
267
static const uint8_t  __attribute__((aligned(8))) dither_2x2_4[2][8]={
268
{  1,   3,   1,   3,   1,   3,   1,   3, },
269
{  2,   0,   2,   0,   2,   0,   2,   0, },
270
};
271

    
272
static const uint8_t  __attribute__((aligned(8))) dither_2x2_8[2][8]={
273
{  6,   2,   6,   2,   6,   2,   6,   2, },
274
{  0,   4,   0,   4,   0,   4,   0,   4, },
275
};
276

    
277
const uint8_t  __attribute__((aligned(8))) dither_8x8_32[8][8]={
278
{ 17,   9,  23,  15,  16,   8,  22,  14, },
279
{  5,  29,   3,  27,   4,  28,   2,  26, },
280
{ 21,  13,  19,  11,  20,  12,  18,  10, },
281
{  0,  24,   6,  30,   1,  25,   7,  31, },
282
{ 16,   8,  22,  14,  17,   9,  23,  15, },
283
{  4,  28,   2,  26,   5,  29,   3,  27, },
284
{ 20,  12,  18,  10,  21,  13,  19,  11, },
285
{  1,  25,   7,  31,   0,  24,   6,  30, },
286
};
287

    
288
#if 0
289
const uint8_t  __attribute__((aligned(8))) dither_8x8_64[8][8]={
290
{  0,  48,  12,  60,   3,  51,  15,  63, },
291
{ 32,  16,  44,  28,  35,  19,  47,  31, },
292
{  8,  56,   4,  52,  11,  59,   7,  55, },
293
{ 40,  24,  36,  20,  43,  27,  39,  23, },
294
{  2,  50,  14,  62,   1,  49,  13,  61, },
295
{ 34,  18,  46,  30,  33,  17,  45,  29, },
296
{ 10,  58,   6,  54,   9,  57,   5,  53, },
297
{ 42,  26,  38,  22,  41,  25,  37,  21, },
298
};
299
#endif
300

    
301
const uint8_t  __attribute__((aligned(8))) dither_8x8_73[8][8]={
302
{  0,  55,  14,  68,   3,  58,  17,  72, },
303
{ 37,  18,  50,  32,  40,  22,  54,  35, },
304
{  9,  64,   5,  59,  13,  67,   8,  63, },
305
{ 46,  27,  41,  23,  49,  31,  44,  26, },
306
{  2,  57,  16,  71,   1,  56,  15,  70, },
307
{ 39,  21,  52,  34,  38,  19,  51,  33, },
308
{ 11,  66,   7,  62,  10,  65,   6,  60, },
309
{ 48,  30,  43,  25,  47,  29,  42,  24, },
310
};
311

    
312
#if 0
313
const uint8_t  __attribute__((aligned(8))) dither_8x8_128[8][8]={
314
{ 68,  36,  92,  60,  66,  34,  90,  58, },
315
{ 20, 116,  12, 108,  18, 114,  10, 106, },
316
{ 84,  52,  76,  44,  82,  50,  74,  42, },
317
{  0,  96,  24, 120,   6, 102,  30, 126, },
318
{ 64,  32,  88,  56,  70,  38,  94,  62, },
319
{ 16, 112,   8, 104,  22, 118,  14, 110, },
320
{ 80,  48,  72,  40,  86,  54,  78,  46, },
321
{  4, 100,  28, 124,   2,  98,  26, 122, },
322
};
323
#endif
324

    
325
#if 1
326
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
327
{117,  62, 158, 103, 113,  58, 155, 100, },
328
{ 34, 199,  21, 186,  31, 196,  17, 182, },
329
{144,  89, 131,  76, 141,  86, 127,  72, },
330
{  0, 165,  41, 206,  10, 175,  52, 217, },
331
{110,  55, 151,  96, 120,  65, 162, 107, },
332
{ 28, 193,  14, 179,  38, 203,  24, 189, },
333
{138,  83, 124,  69, 148,  93, 134,  79, },
334
{  7, 172,  48, 213,   3, 168,  45, 210, },
335
};
336
#elif 1
337
// tries to correct a gamma of 1.5
338
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
339
{  0, 143,  18, 200,   2, 156,  25, 215, },
340
{ 78,  28, 125,  64,  89,  36, 138,  74, },
341
{ 10, 180,   3, 161,  16, 195,   8, 175, },
342
{109,  51,  93,  38, 121,  60, 105,  47, },
343
{  1, 152,  23, 210,   0, 147,  20, 205, },
344
{ 85,  33, 134,  71,  81,  30, 130,  67, },
345
{ 14, 190,   6, 171,  12, 185,   5, 166, },
346
{117,  57, 101,  44, 113,  54,  97,  41, },
347
};
348
#elif 1
349
// tries to correct a gamma of 2.0
350
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
351
{  0, 124,   8, 193,   0, 140,  12, 213, },
352
{ 55,  14, 104,  42,  66,  19, 119,  52, },
353
{  3, 168,   1, 145,   6, 187,   3, 162, },
354
{ 86,  31,  70,  21,  99,  39,  82,  28, },
355
{  0, 134,  11, 206,   0, 129,   9, 200, },
356
{ 62,  17, 114,  48,  58,  16, 109,  45, },
357
{  5, 181,   2, 157,   4, 175,   1, 151, },
358
{ 95,  36,  78,  26,  90,  34,  74,  24, },
359
};
360
#else
361
// tries to correct a gamma of 2.5
362
const uint8_t  __attribute__((aligned(8))) dither_8x8_220[8][8]={
363
{  0, 107,   3, 187,   0, 125,   6, 212, },
364
{ 39,   7,  86,  28,  49,  11, 102,  36, },
365
{  1, 158,   0, 131,   3, 180,   1, 151, },
366
{ 68,  19,  52,  12,  81,  25,  64,  17, },
367
{  0, 119,   5, 203,   0, 113,   4, 195, },
368
{ 45,   9,  96,  33,  42,   8,  91,  30, },
369
{  2, 172,   1, 144,   2, 165,   0, 137, },
370
{ 77,  23,  60,  15,  72,  21,  56,  14, },
371
};
372
#endif
373

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

    
470
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
471
                               int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
472
                               uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
473
{
474
    //FIXME Optimize (just quickly written not optimized..)
475
    int i;
476
    for (i=0; i<dstW; i++)
477
    {
478
        int val=1<<18;
479
        int j;
480
        for (j=0; j<lumFilterSize; j++)
481
            val += lumSrc[j][i] * lumFilter[j];
482

    
483
        dest[i]= av_clip_uint8(val>>19);
484
    }
485

    
486
    if (uDest)
487
        for (i=0; i<chrDstW; i++)
488
        {
489
            int u=1<<18;
490
            int v=1<<18;
491
            int j;
492
            for (j=0; j<chrFilterSize; j++)
493
            {
494
                u += chrSrc[j][i] * chrFilter[j];
495
                v += chrSrc[j][i + VOFW] * chrFilter[j];
496
            }
497

    
498
            uDest[i]= av_clip_uint8(u>>19);
499
            vDest[i]= av_clip_uint8(v>>19);
500
        }
501
}
502

    
503
static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
504
                                int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
505
                                uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
506
{
507
    //FIXME Optimize (just quickly written not optimized..)
508
    int i;
509
    for (i=0; i<dstW; i++)
510
    {
511
        int val=1<<18;
512
        int j;
513
        for (j=0; j<lumFilterSize; j++)
514
            val += lumSrc[j][i] * lumFilter[j];
515

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

    
519
    if (!uDest)
520
        return;
521

    
522
    if (dstFormat == PIX_FMT_NV12)
523
        for (i=0; i<chrDstW; i++)
524
        {
525
            int u=1<<18;
526
            int v=1<<18;
527
            int j;
528
            for (j=0; j<chrFilterSize; j++)
529
            {
530
                u += chrSrc[j][i] * chrFilter[j];
531
                v += chrSrc[j][i + VOFW] * chrFilter[j];
532
            }
533

    
534
            uDest[2*i]= av_clip_uint8(u>>19);
535
            uDest[2*i+1]= av_clip_uint8(v>>19);
536
        }
537
    else
538
        for (i=0; i<chrDstW; i++)
539
        {
540
            int u=1<<18;
541
            int v=1<<18;
542
            int j;
543
            for (j=0; j<chrFilterSize; j++)
544
            {
545
                u += chrSrc[j][i] * chrFilter[j];
546
                v += chrSrc[j][i + VOFW] * chrFilter[j];
547
            }
548

    
549
            uDest[2*i]= av_clip_uint8(v>>19);
550
            uDest[2*i+1]= av_clip_uint8(u>>19);
551
        }
552
}
553

    
554
#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type) \
555
    for (i=0; i<(dstW>>1); i++){\
556
        int j;\
557
        int Y1 = 1<<18;\
558
        int Y2 = 1<<18;\
559
        int U  = 1<<18;\
560
        int V  = 1<<18;\
561
        type av_unused *r, *b, *g;\
562
        const int i2= 2*i;\
563
        \
564
        for (j=0; j<lumFilterSize; j++)\
565
        {\
566
            Y1 += lumSrc[j][i2] * lumFilter[j];\
567
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
568
        }\
569
        for (j=0; j<chrFilterSize; j++)\
570
        {\
571
            U += chrSrc[j][i] * chrFilter[j];\
572
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
573
        }\
574
        Y1>>=19;\
575
        Y2>>=19;\
576
        U >>=19;\
577
        V >>=19;\
578

    
579
#define YSCALE_YUV_2_PACKEDX_C(type) \
580
        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type)\
581
        if ((Y1|Y2|U|V)&256)\
582
        {\
583
            if (Y1>255)   Y1=255; \
584
            else if (Y1<0)Y1=0;   \
585
            if (Y2>255)   Y2=255; \
586
            else if (Y2<0)Y2=0;   \
587
            if (U>255)    U=255;  \
588
            else if (U<0) U=0;    \
589
            if (V>255)    V=255;  \
590
            else if (V<0) V=0;    \
591
        }
592

    
593
#define YSCALE_YUV_2_PACKEDX_FULL_C \
594
    for (i=0; i<dstW; i++){\
595
        int j;\
596
        int Y = 0;\
597
        int U = -128<<19;\
598
        int V = -128<<19;\
599
        int R,G,B;\
600
        \
601
        for (j=0; j<lumFilterSize; j++){\
602
            Y += lumSrc[j][i     ] * lumFilter[j];\
603
        }\
604
        for (j=0; j<chrFilterSize; j++){\
605
            U += chrSrc[j][i     ] * chrFilter[j];\
606
            V += chrSrc[j][i+VOFW] * chrFilter[j];\
607
        }\
608
        Y >>=10;\
609
        U >>=10;\
610
        V >>=10;\
611

    
612
#define YSCALE_YUV_2_RGBX_FULL_C(rnd) \
613
    YSCALE_YUV_2_PACKEDX_FULL_C\
614
        Y-= c->yuv2rgb_y_offset;\
615
        Y*= c->yuv2rgb_y_coeff;\
616
        Y+= rnd;\
617
        R= Y + V*c->yuv2rgb_v2r_coeff;\
618
        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
619
        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
620
        if ((R|G|B)&(0xC0000000)){\
621
            if (R>=(256<<22))   R=(256<<22)-1; \
622
            else if (R<0)R=0;   \
623
            if (G>=(256<<22))   G=(256<<22)-1; \
624
            else if (G<0)G=0;   \
625
            if (B>=(256<<22))   B=(256<<22)-1; \
626
            else if (B<0)B=0;   \
627
        }\
628

    
629

    
630
#define YSCALE_YUV_2_GRAY16_C \
631
    for (i=0; i<(dstW>>1); i++){\
632
        int j;\
633
        int Y1 = 1<<18;\
634
        int Y2 = 1<<18;\
635
        int U  = 1<<18;\
636
        int V  = 1<<18;\
637
        \
638
        const int i2= 2*i;\
639
        \
640
        for (j=0; j<lumFilterSize; j++)\
641
        {\
642
            Y1 += lumSrc[j][i2] * lumFilter[j];\
643
            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
644
        }\
645
        Y1>>=11;\
646
        Y2>>=11;\
647
        if ((Y1|Y2|U|V)&65536)\
648
        {\
649
            if (Y1>65535)   Y1=65535; \
650
            else if (Y1<0)Y1=0;   \
651
            if (Y2>65535)   Y2=65535; \
652
            else if (Y2<0)Y2=0;   \
653
        }
654

    
655
#define YSCALE_YUV_2_RGBX_C(type) \
656
    YSCALE_YUV_2_PACKEDX_C(type)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
657
    r = (type *)c->table_rV[V];   \
658
    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
659
    b = (type *)c->table_bU[U];   \
660

    
661
#define YSCALE_YUV_2_PACKED2_C   \
662
    for (i=0; i<(dstW>>1); i++){ \
663
        const int i2= 2*i;       \
664
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
665
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
666
        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;  \
667
        int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19;  \
668

    
669
#define YSCALE_YUV_2_GRAY16_2_C   \
670
    for (i=0; i<(dstW>>1); i++){ \
671
        const int i2= 2*i;       \
672
        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
673
        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;           \
674

    
675
#define YSCALE_YUV_2_RGB2_C(type) \
676
    YSCALE_YUV_2_PACKED2_C\
677
    type *r, *b, *g;\
678
    r = (type *)c->table_rV[V];\
679
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
680
    b = (type *)c->table_bU[U];\
681

    
682
#define YSCALE_YUV_2_PACKED1_C \
683
    for (i=0; i<(dstW>>1); i++){\
684
        const int i2= 2*i;\
685
        int Y1= buf0[i2  ]>>7;\
686
        int Y2= buf0[i2+1]>>7;\
687
        int U= (uvbuf1[i     ])>>7;\
688
        int V= (uvbuf1[i+VOFW])>>7;\
689

    
690
#define YSCALE_YUV_2_GRAY16_1_C \
691
    for (i=0; i<(dstW>>1); i++){\
692
        const int i2= 2*i;\
693
        int Y1= buf0[i2  ]<<1;\
694
        int Y2= buf0[i2+1]<<1;\
695

    
696
#define YSCALE_YUV_2_RGB1_C(type) \
697
    YSCALE_YUV_2_PACKED1_C\
698
    type *r, *b, *g;\
699
    r = (type *)c->table_rV[V];\
700
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
701
    b = (type *)c->table_bU[U];\
702

    
703
#define YSCALE_YUV_2_PACKED1B_C \
704
    for (i=0; i<(dstW>>1); i++){\
705
        const int i2= 2*i;\
706
        int Y1= buf0[i2  ]>>7;\
707
        int Y2= buf0[i2+1]>>7;\
708
        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
709
        int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
710

    
711
#define YSCALE_YUV_2_RGB1B_C(type) \
712
    YSCALE_YUV_2_PACKED1B_C\
713
    type *r, *b, *g;\
714
    r = (type *)c->table_rV[V];\
715
    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
716
    b = (type *)c->table_bU[U];\
717

    
718
#define YSCALE_YUV_2_MONO2_C \
719
    const uint8_t * const d128=dither_8x8_220[y&7];\
720
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
721
    for (i=0; i<dstW-7; i+=8){\
722
        int acc;\
723
        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
724
        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
725
        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
726
        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
727
        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
728
        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
729
        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
730
        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
731
        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
732
        dest++;\
733
    }\
734

    
735

    
736
#define YSCALE_YUV_2_MONOX_C \
737
    const uint8_t * const d128=dither_8x8_220[y&7];\
738
    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
739
    int acc=0;\
740
    for (i=0; i<dstW-1; i+=2){\
741
        int j;\
742
        int Y1=1<<18;\
743
        int Y2=1<<18;\
744
\
745
        for (j=0; j<lumFilterSize; j++)\
746
        {\
747
            Y1 += lumSrc[j][i] * lumFilter[j];\
748
            Y2 += lumSrc[j][i+1] * lumFilter[j];\
749
        }\
750
        Y1>>=19;\
751
        Y2>>=19;\
752
        if ((Y1|Y2)&256)\
753
        {\
754
            if (Y1>255)   Y1=255;\
755
            else if (Y1<0)Y1=0;\
756
            if (Y2>255)   Y2=255;\
757
            else if (Y2<0)Y2=0;\
758
        }\
759
        acc+= acc + g[Y1+d128[(i+0)&7]];\
760
        acc+= acc + g[Y2+d128[(i+1)&7]];\
761
        if ((i&7)==6){\
762
            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
763
            dest++;\
764
        }\
765
    }
766

    
767

    
768
#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
769
    switch(c->dstFormat)\
770
    {\
771
    case PIX_FMT_RGB32:\
772
    case PIX_FMT_BGR32:\
773
    case PIX_FMT_RGB32_1:\
774
    case PIX_FMT_BGR32_1:\
775
        func(uint32_t)\
776
            ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
777
            ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
778
        }                \
779
        break;\
780
    case PIX_FMT_RGB24:\
781
        func(uint8_t)\
782
            ((uint8_t*)dest)[0]= r[Y1];\
783
            ((uint8_t*)dest)[1]= g[Y1];\
784
            ((uint8_t*)dest)[2]= b[Y1];\
785
            ((uint8_t*)dest)[3]= r[Y2];\
786
            ((uint8_t*)dest)[4]= g[Y2];\
787
            ((uint8_t*)dest)[5]= b[Y2];\
788
            dest+=6;\
789
        }\
790
        break;\
791
    case PIX_FMT_BGR24:\
792
        func(uint8_t)\
793
            ((uint8_t*)dest)[0]= b[Y1];\
794
            ((uint8_t*)dest)[1]= g[Y1];\
795
            ((uint8_t*)dest)[2]= r[Y1];\
796
            ((uint8_t*)dest)[3]= b[Y2];\
797
            ((uint8_t*)dest)[4]= g[Y2];\
798
            ((uint8_t*)dest)[5]= r[Y2];\
799
            dest+=6;\
800
        }\
801
        break;\
802
    case PIX_FMT_RGB565:\
803
    case PIX_FMT_BGR565:\
804
        {\
805
            const int dr1= dither_2x2_8[y&1    ][0];\
806
            const int dg1= dither_2x2_4[y&1    ][0];\
807
            const int db1= dither_2x2_8[(y&1)^1][0];\
808
            const int dr2= dither_2x2_8[y&1    ][1];\
809
            const int dg2= dither_2x2_4[y&1    ][1];\
810
            const int db2= dither_2x2_8[(y&1)^1][1];\
811
            func(uint16_t)\
812
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
813
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
814
            }\
815
        }\
816
        break;\
817
    case PIX_FMT_RGB555:\
818
    case PIX_FMT_BGR555:\
819
        {\
820
            const int dr1= dither_2x2_8[y&1    ][0];\
821
            const int dg1= dither_2x2_8[y&1    ][1];\
822
            const int db1= dither_2x2_8[(y&1)^1][0];\
823
            const int dr2= dither_2x2_8[y&1    ][1];\
824
            const int dg2= dither_2x2_8[y&1    ][0];\
825
            const int db2= dither_2x2_8[(y&1)^1][1];\
826
            func(uint16_t)\
827
                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
828
                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
829
            }\
830
        }\
831
        break;\
832
    case PIX_FMT_RGB8:\
833
    case PIX_FMT_BGR8:\
834
        {\
835
            const uint8_t * const d64= dither_8x8_73[y&7];\
836
            const uint8_t * const d32= dither_8x8_32[y&7];\
837
            func(uint8_t)\
838
                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
839
                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
840
            }\
841
        }\
842
        break;\
843
    case PIX_FMT_RGB4:\
844
    case PIX_FMT_BGR4:\
845
        {\
846
            const uint8_t * const d64= dither_8x8_73 [y&7];\
847
            const uint8_t * const d128=dither_8x8_220[y&7];\
848
            func(uint8_t)\
849
                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
850
                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
851
            }\
852
        }\
853
        break;\
854
    case PIX_FMT_RGB4_BYTE:\
855
    case PIX_FMT_BGR4_BYTE:\
856
        {\
857
            const uint8_t * const d64= dither_8x8_73 [y&7];\
858
            const uint8_t * const d128=dither_8x8_220[y&7];\
859
            func(uint8_t)\
860
                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
861
                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
862
            }\
863
        }\
864
        break;\
865
    case PIX_FMT_MONOBLACK:\
866
    case PIX_FMT_MONOWHITE:\
867
        {\
868
            func_monoblack\
869
        }\
870
        break;\
871
    case PIX_FMT_YUYV422:\
872
        func2\
873
            ((uint8_t*)dest)[2*i2+0]= Y1;\
874
            ((uint8_t*)dest)[2*i2+1]= U;\
875
            ((uint8_t*)dest)[2*i2+2]= Y2;\
876
            ((uint8_t*)dest)[2*i2+3]= V;\
877
        }                \
878
        break;\
879
    case PIX_FMT_UYVY422:\
880
        func2\
881
            ((uint8_t*)dest)[2*i2+0]= U;\
882
            ((uint8_t*)dest)[2*i2+1]= Y1;\
883
            ((uint8_t*)dest)[2*i2+2]= V;\
884
            ((uint8_t*)dest)[2*i2+3]= Y2;\
885
        }                \
886
        break;\
887
    case PIX_FMT_GRAY16BE:\
888
        func_g16\
889
            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
890
            ((uint8_t*)dest)[2*i2+1]= Y1;\
891
            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
892
            ((uint8_t*)dest)[2*i2+3]= Y2;\
893
        }                \
894
        break;\
895
    case PIX_FMT_GRAY16LE:\
896
        func_g16\
897
            ((uint8_t*)dest)[2*i2+0]= Y1;\
898
            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
899
            ((uint8_t*)dest)[2*i2+2]= Y2;\
900
            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
901
        }                \
902
        break;\
903
    }\
904

    
905

    
906
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
907
                                  int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
908
                                  uint8_t *dest, int dstW, int y)
909
{
910
    int i;
911
    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
912
}
913

    
914
static inline void yuv2rgbXinC_full(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
915
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
916
                                    uint8_t *dest, int dstW, int y)
917
{
918
    int i;
919
    int step= fmt_depth(c->dstFormat)/8;
920
    int aidx= 3;
921

    
922
    switch(c->dstFormat){
923
    case PIX_FMT_ARGB:
924
        dest++;
925
        aidx= 0;
926
    case PIX_FMT_RGB24:
927
        aidx--;
928
    case PIX_FMT_RGBA:
929
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
930
            dest[aidx]= 0;
931
            dest[0]= R>>22;
932
            dest[1]= G>>22;
933
            dest[2]= B>>22;
934
            dest+= step;
935
        }
936
        break;
937
    case PIX_FMT_ABGR:
938
        dest++;
939
        aidx= 0;
940
    case PIX_FMT_BGR24:
941
        aidx--;
942
    case PIX_FMT_BGRA:
943
        YSCALE_YUV_2_RGBX_FULL_C(1<<21)
944
            dest[aidx]= 0;
945
            dest[0]= B>>22;
946
            dest[1]= G>>22;
947
            dest[2]= R>>22;
948
            dest+= step;
949
        }
950
        break;
951
    default:
952
        assert(0);
953
    }
954
}
955

    
956
//Note: we have C, X86, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
957
//Plain C versions
958
#if !HAVE_MMX || defined (RUNTIME_CPUDETECT) || !CONFIG_GPL
959
#define COMPILE_C
960
#endif
961

    
962
#if ARCH_PPC
963
#if (HAVE_ALTIVEC || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
964
#undef COMPILE_C
965
#define COMPILE_ALTIVEC
966
#endif
967
#endif //ARCH_PPC
968

    
969
#if ARCH_X86
970

    
971
#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
972
#define COMPILE_MMX
973
#endif
974

    
975
#if (HAVE_MMX2 || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
976
#define COMPILE_MMX2
977
#endif
978

    
979
#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && CONFIG_GPL
980
#define COMPILE_3DNOW
981
#endif
982
#endif //ARCH_X86
983

    
984
#undef HAVE_MMX
985
#undef HAVE_MMX2
986
#undef HAVE_AMD3DNOW
987
#undef HAVE_ALTIVEC
988
#define HAVE_MMX 0
989
#define HAVE_MMX2 0
990
#define HAVE_AMD3DNOW 0
991
#define HAVE_ALTIVEC 0
992

    
993
#ifdef COMPILE_C
994
#define RENAME(a) a ## _C
995
#include "swscale_template.c"
996
#endif
997

    
998
#ifdef COMPILE_ALTIVEC
999
#undef RENAME
1000
#undef HAVE_ALTIVEC
1001
#define HAVE_ALTIVEC 1
1002
#define RENAME(a) a ## _altivec
1003
#include "swscale_template.c"
1004
#endif
1005

    
1006
#if ARCH_X86
1007

    
1008
//x86 versions
1009
/*
1010
#undef RENAME
1011
#undef HAVE_MMX
1012
#undef HAVE_MMX2
1013
#undef HAVE_AMD3DNOW
1014
#define ARCH_X86
1015
#define RENAME(a) a ## _X86
1016
#include "swscale_template.c"
1017
*/
1018
//MMX versions
1019
#ifdef COMPILE_MMX
1020
#undef RENAME
1021
#undef HAVE_MMX
1022
#undef HAVE_MMX2
1023
#undef HAVE_AMD3DNOW
1024
#define HAVE_MMX 1
1025
#define HAVE_MMX2 0
1026
#define HAVE_AMD3DNOW 0
1027
#define RENAME(a) a ## _MMX
1028
#include "swscale_template.c"
1029
#endif
1030

    
1031
//MMX2 versions
1032
#ifdef COMPILE_MMX2
1033
#undef RENAME
1034
#undef HAVE_MMX
1035
#undef HAVE_MMX2
1036
#undef HAVE_AMD3DNOW
1037
#define HAVE_MMX 1
1038
#define HAVE_MMX2 1
1039
#define HAVE_AMD3DNOW 0
1040
#define RENAME(a) a ## _MMX2
1041
#include "swscale_template.c"
1042
#endif
1043

    
1044
//3DNOW versions
1045
#ifdef COMPILE_3DNOW
1046
#undef RENAME
1047
#undef HAVE_MMX
1048
#undef HAVE_MMX2
1049
#undef HAVE_AMD3DNOW
1050
#define HAVE_MMX 1
1051
#define HAVE_MMX2 0
1052
#define HAVE_AMD3DNOW 1
1053
#define RENAME(a) a ## _3DNow
1054
#include "swscale_template.c"
1055
#endif
1056

    
1057
#endif //ARCH_X86
1058

    
1059
// minor note: the HAVE_xyz are messed up after this line so don't use them
1060

    
1061
static double getSplineCoeff(double a, double b, double c, double d, double dist)
1062
{
1063
//    printf("%f %f %f %f %f\n", a,b,c,d,dist);
1064
    if (dist<=1.0)      return ((d*dist + c)*dist + b)*dist +a;
1065
    else                return getSplineCoeff(        0.0,
1066
                                             b+ 2.0*c + 3.0*d,
1067
                                                    c + 3.0*d,
1068
                                            -b- 3.0*c - 6.0*d,
1069
                                            dist-1.0);
1070
}
1071

    
1072
static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1073
                             int srcW, int dstW, int filterAlign, int one, int flags,
1074
                             SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1075
{
1076
    int i;
1077
    int filterSize;
1078
    int filter2Size;
1079
    int minFilterSize;
1080
    int64_t *filter=NULL;
1081
    int64_t *filter2=NULL;
1082
    const int64_t fone= 1LL<<54;
1083
    int ret= -1;
1084
#if ARCH_X86
1085
    if (flags & SWS_CPU_CAPS_MMX)
1086
        __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1087
#endif
1088

    
1089
    // NOTE: the +1 is for the MMX scaler which reads over the end
1090
    *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1091

    
1092
    if (FFABS(xInc - 0x10000) <10) // unscaled
1093
    {
1094
        int i;
1095
        filterSize= 1;
1096
        filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1097

    
1098
        for (i=0; i<dstW; i++)
1099
        {
1100
            filter[i*filterSize]= fone;
1101
            (*filterPos)[i]=i;
1102
        }
1103

    
1104
    }
1105
    else if (flags&SWS_POINT) // lame looking point sampling mode
1106
    {
1107
        int i;
1108
        int xDstInSrc;
1109
        filterSize= 1;
1110
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1111

    
1112
        xDstInSrc= xInc/2 - 0x8000;
1113
        for (i=0; i<dstW; i++)
1114
        {
1115
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1116

    
1117
            (*filterPos)[i]= xx;
1118
            filter[i]= fone;
1119
            xDstInSrc+= xInc;
1120
        }
1121
    }
1122
    else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1123
    {
1124
        int i;
1125
        int xDstInSrc;
1126
        if      (flags&SWS_BICUBIC) filterSize= 4;
1127
        else if (flags&SWS_X      ) filterSize= 4;
1128
        else                        filterSize= 2; // SWS_BILINEAR / SWS_AREA
1129
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1130

    
1131
        xDstInSrc= xInc/2 - 0x8000;
1132
        for (i=0; i<dstW; i++)
1133
        {
1134
            int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1135
            int j;
1136

    
1137
            (*filterPos)[i]= xx;
1138
                //bilinear upscale / linear interpolate / area averaging
1139
                for (j=0; j<filterSize; j++)
1140
                {
1141
                    int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1142
                    if (coeff<0) coeff=0;
1143
                    filter[i*filterSize + j]= coeff;
1144
                    xx++;
1145
                }
1146
            xDstInSrc+= xInc;
1147
        }
1148
    }
1149
    else
1150
    {
1151
        int xDstInSrc;
1152
        int sizeFactor;
1153

    
1154
        if      (flags&SWS_BICUBIC)      sizeFactor=  4;
1155
        else if (flags&SWS_X)            sizeFactor=  8;
1156
        else if (flags&SWS_AREA)         sizeFactor=  1; //downscale only, for upscale it is bilinear
1157
        else if (flags&SWS_GAUSS)        sizeFactor=  8;   // infinite ;)
1158
        else if (flags&SWS_LANCZOS)      sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1159
        else if (flags&SWS_SINC)         sizeFactor= 20; // infinite ;)
1160
        else if (flags&SWS_SPLINE)       sizeFactor= 20;  // infinite ;)
1161
        else if (flags&SWS_BILINEAR)     sizeFactor=  2;
1162
        else {
1163
            sizeFactor= 0; //GCC warning killer
1164
            assert(0);
1165
        }
1166

    
1167
        if (xInc <= 1<<16)      filterSize= 1 + sizeFactor; // upscale
1168
        else                    filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1169

    
1170
        if (filterSize > srcW-2) filterSize=srcW-2;
1171

    
1172
        filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1173

    
1174
        xDstInSrc= xInc - 0x10000;
1175
        for (i=0; i<dstW; i++)
1176
        {
1177
            int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1178
            int j;
1179
            (*filterPos)[i]= xx;
1180
            for (j=0; j<filterSize; j++)
1181
            {
1182
                int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1183
                double floatd;
1184
                int64_t coeff;
1185

    
1186
                if (xInc > 1<<16)
1187
                    d= d*dstW/srcW;
1188
                floatd= d * (1.0/(1<<30));
1189

    
1190
                if (flags & SWS_BICUBIC)
1191
                {
1192
                    int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1<<24);
1193
                    int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1194
                    int64_t dd = ( d*d)>>30;
1195
                    int64_t ddd= (dd*d)>>30;
1196

    
1197
                    if      (d < 1LL<<30)
1198
                        coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1199
                    else if (d < 1LL<<31)
1200
                        coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1201
                    else
1202
                        coeff=0.0;
1203
                    coeff *= fone>>(30+24);
1204
                }
1205
/*                else if (flags & SWS_X)
1206
                {
1207
                    double p= param ? param*0.01 : 0.3;
1208
                    coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1209
                    coeff*= pow(2.0, - p*d*d);
1210
                }*/
1211
                else if (flags & SWS_X)
1212
                {
1213
                    double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1214
                    double c;
1215

    
1216
                    if (floatd<1.0)
1217
                        c = cos(floatd*PI);
1218
                    else
1219
                        c=-1.0;
1220
                    if (c<0.0)      c= -pow(-c, A);
1221
                    else            c=  pow( c, A);
1222
                    coeff= (c*0.5 + 0.5)*fone;
1223
                }
1224
                else if (flags & SWS_AREA)
1225
                {
1226
                    int64_t d2= d - (1<<29);
1227
                    if      (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1228
                    else if (d2*xInc <  (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1229
                    else coeff=0.0;
1230
                    coeff *= fone>>(30+16);
1231
                }
1232
                else if (flags & SWS_GAUSS)
1233
                {
1234
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1235
                    coeff = (pow(2.0, - p*floatd*floatd))*fone;
1236
                }
1237
                else if (flags & SWS_SINC)
1238
                {
1239
                    coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1240
                }
1241
                else if (flags & SWS_LANCZOS)
1242
                {
1243
                    double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1244
                    coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1245
                    if (floatd>p) coeff=0;
1246
                }
1247
                else if (flags & SWS_BILINEAR)
1248
                {
1249
                    coeff= (1<<30) - d;
1250
                    if (coeff<0) coeff=0;
1251
                    coeff *= fone >> 30;
1252
                }
1253
                else if (flags & SWS_SPLINE)
1254
                {
1255
                    double p=-2.196152422706632;
1256
                    coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1257
                }
1258
                else {
1259
                    coeff= 0.0; //GCC warning killer
1260
                    assert(0);
1261
                }
1262

    
1263
                filter[i*filterSize + j]= coeff;
1264
                xx++;
1265
            }
1266
            xDstInSrc+= 2*xInc;
1267
        }
1268
    }
1269

    
1270
    /* apply src & dst Filter to filter -> filter2
1271
       av_free(filter);
1272
    */
1273
    assert(filterSize>0);
1274
    filter2Size= filterSize;
1275
    if (srcFilter) filter2Size+= srcFilter->length - 1;
1276
    if (dstFilter) filter2Size+= dstFilter->length - 1;
1277
    assert(filter2Size>0);
1278
    filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1279

    
1280
    for (i=0; i<dstW; i++)
1281
    {
1282
        int j, k;
1283

    
1284
        if(srcFilter){
1285
            for (k=0; k<srcFilter->length; k++){
1286
                for (j=0; j<filterSize; j++)
1287
                    filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1288
            }
1289
        }else{
1290
            for (j=0; j<filterSize; j++)
1291
                filter2[i*filter2Size + j]= filter[i*filterSize + j];
1292
        }
1293
        //FIXME dstFilter
1294

    
1295
        (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1296
    }
1297
    av_freep(&filter);
1298

    
1299
    /* try to reduce the filter-size (step1 find size and shift left) */
1300
    // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1301
    minFilterSize= 0;
1302
    for (i=dstW-1; i>=0; i--)
1303
    {
1304
        int min= filter2Size;
1305
        int j;
1306
        int64_t cutOff=0.0;
1307

    
1308
        /* get rid off near zero elements on the left by shifting left */
1309
        for (j=0; j<filter2Size; j++)
1310
        {
1311
            int k;
1312
            cutOff += FFABS(filter2[i*filter2Size]);
1313

    
1314
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1315

    
1316
            /* preserve monotonicity because the core can't handle the filter otherwise */
1317
            if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1318

    
1319
            // move filter coefficients left
1320
            for (k=1; k<filter2Size; k++)
1321
                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1322
            filter2[i*filter2Size + k - 1]= 0;
1323
            (*filterPos)[i]++;
1324
        }
1325

    
1326
        cutOff=0;
1327
        /* count near zeros on the right */
1328
        for (j=filter2Size-1; j>0; j--)
1329
        {
1330
            cutOff += FFABS(filter2[i*filter2Size + j]);
1331

    
1332
            if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1333
            min--;
1334
        }
1335

    
1336
        if (min>minFilterSize) minFilterSize= min;
1337
    }
1338

    
1339
    if (flags & SWS_CPU_CAPS_ALTIVEC) {
1340
        // we can handle the special case 4,
1341
        // so we don't want to go to the full 8
1342
        if (minFilterSize < 5)
1343
            filterAlign = 4;
1344

    
1345
        // We really don't want to waste our time
1346
        // doing useless computation, so fall back on
1347
        // the scalar C code for very small filters.
1348
        // Vectorizing is worth it only if you have a
1349
        // decent-sized vector.
1350
        if (minFilterSize < 3)
1351
            filterAlign = 1;
1352
    }
1353

    
1354
    if (flags & SWS_CPU_CAPS_MMX) {
1355
        // special case for unscaled vertical filtering
1356
        if (minFilterSize == 1 && filterAlign == 2)
1357
            filterAlign= 1;
1358
    }
1359

    
1360
    assert(minFilterSize > 0);
1361
    filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1362
    assert(filterSize > 0);
1363
    filter= av_malloc(filterSize*dstW*sizeof(*filter));
1364
    if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1365
        goto error;
1366
    *outFilterSize= filterSize;
1367

    
1368
    if (flags&SWS_PRINT_INFO)
1369
        av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1370
    /* try to reduce the filter-size (step2 reduce it) */
1371
    for (i=0; i<dstW; i++)
1372
    {
1373
        int j;
1374

    
1375
        for (j=0; j<filterSize; j++)
1376
        {
1377
            if (j>=filter2Size) filter[i*filterSize + j]= 0;
1378
            else               filter[i*filterSize + j]= filter2[i*filter2Size + j];
1379
            if((flags & SWS_BITEXACT) && j>=minFilterSize)
1380
                filter[i*filterSize + j]= 0;
1381
        }
1382
    }
1383

    
1384

    
1385
    //FIXME try to align filterPos if possible
1386

    
1387
    //fix borders
1388
    for (i=0; i<dstW; i++)
1389
    {
1390
        int j;
1391
        if ((*filterPos)[i] < 0)
1392
        {
1393
            // move filter coefficients left to compensate for filterPos
1394
            for (j=1; j<filterSize; j++)
1395
            {
1396
                int left= FFMAX(j + (*filterPos)[i], 0);
1397
                filter[i*filterSize + left] += filter[i*filterSize + j];
1398
                filter[i*filterSize + j]=0;
1399
            }
1400
            (*filterPos)[i]= 0;
1401
        }
1402

    
1403
        if ((*filterPos)[i] + filterSize > srcW)
1404
        {
1405
            int shift= (*filterPos)[i] + filterSize - srcW;
1406
            // move filter coefficients right to compensate for filterPos
1407
            for (j=filterSize-2; j>=0; j--)
1408
            {
1409
                int right= FFMIN(j + shift, filterSize-1);
1410
                filter[i*filterSize +right] += filter[i*filterSize +j];
1411
                filter[i*filterSize +j]=0;
1412
            }
1413
            (*filterPos)[i]= srcW - filterSize;
1414
        }
1415
    }
1416

    
1417
    // Note the +1 is for the MMX scaler which reads over the end
1418
    /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1419
    *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1420

    
1421
    /* normalize & store in outFilter */
1422
    for (i=0; i<dstW; i++)
1423
    {
1424
        int j;
1425
        int64_t error=0;
1426
        int64_t sum=0;
1427

    
1428
        for (j=0; j<filterSize; j++)
1429
        {
1430
            sum+= filter[i*filterSize + j];
1431
        }
1432
        sum= (sum + one/2)/ one;
1433
        for (j=0; j<*outFilterSize; j++)
1434
        {
1435
            int64_t v= filter[i*filterSize + j] + error;
1436
            int intV= ROUNDED_DIV(v, sum);
1437
            (*outFilter)[i*(*outFilterSize) + j]= intV;
1438
            error= v - intV*sum;
1439
        }
1440
    }
1441

    
1442
    (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1443
    for (i=0; i<*outFilterSize; i++)
1444
    {
1445
        int j= dstW*(*outFilterSize);
1446
        (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1447
    }
1448

    
1449
    ret=0;
1450
error:
1451
    av_free(filter);
1452
    av_free(filter2);
1453
    return ret;
1454
}
1455

    
1456
#ifdef COMPILE_MMX2
1457
static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1458
{
1459
    uint8_t *fragmentA;
1460
    long imm8OfPShufW1A;
1461
    long imm8OfPShufW2A;
1462
    long fragmentLengthA;
1463
    uint8_t *fragmentB;
1464
    long imm8OfPShufW1B;
1465
    long imm8OfPShufW2B;
1466
    long fragmentLengthB;
1467
    int fragmentPos;
1468

    
1469
    int xpos, i;
1470

    
1471
    // create an optimized horizontal scaling routine
1472

    
1473
    //code fragment
1474

    
1475
    __asm__ volatile(
1476
        "jmp                         9f                 \n\t"
1477
    // Begin
1478
        "0:                                             \n\t"
1479
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1480
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1481
        "movd   1(%%"REG_c", %%"REG_S"), %%mm1          \n\t"
1482
        "punpcklbw                %%mm7, %%mm1          \n\t"
1483
        "punpcklbw                %%mm7, %%mm0          \n\t"
1484
        "pshufw                   $0xFF, %%mm1, %%mm1   \n\t"
1485
        "1:                                             \n\t"
1486
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1487
        "2:                                             \n\t"
1488
        "psubw                    %%mm1, %%mm0          \n\t"
1489
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1490
        "pmullw                   %%mm3, %%mm0          \n\t"
1491
        "psllw                       $7, %%mm1          \n\t"
1492
        "paddw                    %%mm1, %%mm0          \n\t"
1493

    
1494
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1495

    
1496
        "add                         $8, %%"REG_a"      \n\t"
1497
    // End
1498
        "9:                                             \n\t"
1499
//        "int $3                                         \n\t"
1500
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1501
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1502
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1503
        "dec                         %1                 \n\t"
1504
        "dec                         %2                 \n\t"
1505
        "sub                         %0, %1             \n\t"
1506
        "sub                         %0, %2             \n\t"
1507
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1508
        "sub                         %0, %3             \n\t"
1509

    
1510

    
1511
        :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1512
        "=r" (fragmentLengthA)
1513
    );
1514

    
1515
    __asm__ volatile(
1516
        "jmp                         9f                 \n\t"
1517
    // Begin
1518
        "0:                                             \n\t"
1519
        "movq    (%%"REG_d", %%"REG_a"), %%mm3          \n\t"
1520
        "movd    (%%"REG_c", %%"REG_S"), %%mm0          \n\t"
1521
        "punpcklbw                %%mm7, %%mm0          \n\t"
1522
        "pshufw                   $0xFF, %%mm0, %%mm1   \n\t"
1523
        "1:                                             \n\t"
1524
        "pshufw                   $0xFF, %%mm0, %%mm0   \n\t"
1525
        "2:                                             \n\t"
1526
        "psubw                    %%mm1, %%mm0          \n\t"
1527
        "movl   8(%%"REG_b", %%"REG_a"), %%esi          \n\t"
1528
        "pmullw                   %%mm3, %%mm0          \n\t"
1529
        "psllw                       $7, %%mm1          \n\t"
1530
        "paddw                    %%mm1, %%mm0          \n\t"
1531

    
1532
        "movq                     %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1533

    
1534
        "add                         $8, %%"REG_a"      \n\t"
1535
    // End
1536
        "9:                                             \n\t"
1537
//        "int                       $3                   \n\t"
1538
        "lea                 " LOCAL_MANGLE(0b) ", %0   \n\t"
1539
        "lea                 " LOCAL_MANGLE(1b) ", %1   \n\t"
1540
        "lea                 " LOCAL_MANGLE(2b) ", %2   \n\t"
1541
        "dec                         %1                 \n\t"
1542
        "dec                         %2                 \n\t"
1543
        "sub                         %0, %1             \n\t"
1544
        "sub                         %0, %2             \n\t"
1545
        "lea                 " LOCAL_MANGLE(9b) ", %3   \n\t"
1546
        "sub                         %0, %3             \n\t"
1547

    
1548

    
1549
        :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1550
        "=r" (fragmentLengthB)
1551
    );
1552

    
1553
    xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1554
    fragmentPos=0;
1555

    
1556
    for (i=0; i<dstW/numSplits; i++)
1557
    {
1558
        int xx=xpos>>16;
1559

    
1560
        if ((i&3) == 0)
1561
        {
1562
            int a=0;
1563
            int b=((xpos+xInc)>>16) - xx;
1564
            int c=((xpos+xInc*2)>>16) - xx;
1565
            int d=((xpos+xInc*3)>>16) - xx;
1566

    
1567
            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1568
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1569
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1570
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1571
            filterPos[i/2]= xx;
1572

    
1573
            if (d+1<4)
1574
            {
1575
                int maxShift= 3-(d+1);
1576
                int shift=0;
1577

    
1578
                memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1579

    
1580
                funnyCode[fragmentPos + imm8OfPShufW1B]=
1581
                    (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1582
                funnyCode[fragmentPos + imm8OfPShufW2B]=
1583
                    a | (b<<2) | (c<<4) | (d<<6);
1584

    
1585
                if (i+3>=dstW) shift=maxShift; //avoid overread
1586
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1587

    
1588
                if (shift && i>=shift)
1589
                {
1590
                    funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1591
                    funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1592
                    filterPos[i/2]-=shift;
1593
                }
1594

    
1595
                fragmentPos+= fragmentLengthB;
1596
            }
1597
            else
1598
            {
1599
                int maxShift= 3-d;
1600
                int shift=0;
1601

    
1602
                memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
1603

    
1604
                funnyCode[fragmentPos + imm8OfPShufW1A]=
1605
                funnyCode[fragmentPos + imm8OfPShufW2A]=
1606
                    a | (b<<2) | (c<<4) | (d<<6);
1607

    
1608
                if (i+4>=dstW) shift=maxShift; //avoid overread
1609
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
1610

    
1611
                if (shift && i>=shift)
1612
                {
1613
                    funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1614
                    funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1615
                    filterPos[i/2]-=shift;
1616
                }
1617

    
1618
                fragmentPos+= fragmentLengthA;
1619
            }
1620

    
1621
            funnyCode[fragmentPos]= RET;
1622
        }
1623
        xpos+=xInc;
1624
    }
1625
    filterPos[i/2]= xpos>>16; // needed to jump to the next part
1626
}
1627
#endif /* COMPILE_MMX2 */
1628

    
1629
static void globalInit(void){
1630
    // generating tables:
1631
    int i;
1632
    for (i=0; i<768; i++){
1633
        int c= av_clip_uint8(i-256);
1634
        clip_table[i]=c;
1635
    }
1636
}
1637

    
1638
static SwsFunc getSwsFunc(int flags){
1639

    
1640
#if defined(RUNTIME_CPUDETECT) && CONFIG_GPL
1641
#if ARCH_X86
1642
    // ordered per speed fastest first
1643
    if (flags & SWS_CPU_CAPS_MMX2)
1644
        return swScale_MMX2;
1645
    else if (flags & SWS_CPU_CAPS_3DNOW)
1646
        return swScale_3DNow;
1647
    else if (flags & SWS_CPU_CAPS_MMX)
1648
        return swScale_MMX;
1649
    else
1650
        return swScale_C;
1651

    
1652
#else
1653
#if ARCH_PPC
1654
    if (flags & SWS_CPU_CAPS_ALTIVEC)
1655
        return swScale_altivec;
1656
    else
1657
        return swScale_C;
1658
#endif
1659
    return swScale_C;
1660
#endif /* ARCH_X86 */
1661
#else //RUNTIME_CPUDETECT
1662
#if   HAVE_MMX2
1663
    return swScale_MMX2;
1664
#elif HAVE_AMD3DNOW
1665
    return swScale_3DNow;
1666
#elif HAVE_MMX
1667
    return swScale_MMX;
1668
#elif HAVE_ALTIVEC
1669
    return swScale_altivec;
1670
#else
1671
    return swScale_C;
1672
#endif
1673
#endif //!RUNTIME_CPUDETECT
1674
}
1675

    
1676
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1677
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1678
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1679
    /* Copy Y plane */
1680
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1681
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
1682
    else
1683
    {
1684
        int i;
1685
        uint8_t *srcPtr= src[0];
1686
        uint8_t *dstPtr= dst;
1687
        for (i=0; i<srcSliceH; i++)
1688
        {
1689
            memcpy(dstPtr, srcPtr, c->srcW);
1690
            srcPtr+= srcStride[0];
1691
            dstPtr+= dstStride[0];
1692
        }
1693
    }
1694
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1695
    if (c->dstFormat == PIX_FMT_NV12)
1696
        interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1697
    else
1698
        interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1699

    
1700
    return srcSliceH;
1701
}
1702

    
1703
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1704
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1705
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1706

    
1707
    yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1708

    
1709
    return srcSliceH;
1710
}
1711

    
1712
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1713
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1714
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1715

    
1716
    yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1717

    
1718
    return srcSliceH;
1719
}
1720

    
1721
static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1722
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1723
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1724

    
1725
    yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1726

    
1727
    return srcSliceH;
1728
}
1729

    
1730
static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1731
                                int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1732
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1733

    
1734
    yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1735

    
1736
    return srcSliceH;
1737
}
1738

    
1739
static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1740
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
1741
    const enum PixelFormat srcFormat= c->srcFormat;
1742
    const enum PixelFormat dstFormat= c->dstFormat;
1743
    void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1744
                 const uint8_t *palette)=NULL;
1745
    int i;
1746
    uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1747
    uint8_t *srcPtr= src[0];
1748

    
1749
    if (!usePal(srcFormat))
1750
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1751
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1752

    
1753
    switch(dstFormat){
1754
    case PIX_FMT_RGB32  : conv = palette8topacked32; break;
1755
    case PIX_FMT_BGR32  : conv = palette8topacked32; break;
1756
    case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1757
    case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1758
    case PIX_FMT_RGB24  : conv = palette8topacked24; break;
1759
    case PIX_FMT_BGR24  : conv = palette8topacked24; break;
1760
    default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1761
                    sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1762
    }
1763

    
1764

    
1765
    for (i=0; i<srcSliceH; i++) {
1766
        conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1767
        srcPtr+= srcStride[0];
1768
        dstPtr+= dstStride[0];
1769
    }
1770

    
1771
    return srcSliceH;
1772
}
1773

    
1774
/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1775
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1776
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
1777
    const enum PixelFormat srcFormat= c->srcFormat;
1778
    const enum PixelFormat dstFormat= c->dstFormat;
1779
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
1780
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
1781
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1782
    const int dstId= fmt_depth(dstFormat) >> 2;
1783
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1784

    
1785
    /* BGR -> BGR */
1786
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
1787
       || (isRGB(srcFormat) && isRGB(dstFormat))){
1788
        switch(srcId | (dstId<<4)){
1789
        case 0x34: conv= rgb16to15; break;
1790
        case 0x36: conv= rgb24to15; break;
1791
        case 0x38: conv= rgb32to15; break;
1792
        case 0x43: conv= rgb15to16; break;
1793
        case 0x46: conv= rgb24to16; break;
1794
        case 0x48: conv= rgb32to16; break;
1795
        case 0x63: conv= rgb15to24; break;
1796
        case 0x64: conv= rgb16to24; break;
1797
        case 0x68: conv= rgb32to24; break;
1798
        case 0x83: conv= rgb15to32; break;
1799
        case 0x84: conv= rgb16to32; break;
1800
        case 0x86: conv= rgb24to32; break;
1801
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1802
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1803
        }
1804
    }else if (  (isBGR(srcFormat) && isRGB(dstFormat))
1805
             || (isRGB(srcFormat) && isBGR(dstFormat))){
1806
        switch(srcId | (dstId<<4)){
1807
        case 0x33: conv= rgb15tobgr15; break;
1808
        case 0x34: conv= rgb16tobgr15; break;
1809
        case 0x36: conv= rgb24tobgr15; break;
1810
        case 0x38: conv= rgb32tobgr15; break;
1811
        case 0x43: conv= rgb15tobgr16; break;
1812
        case 0x44: conv= rgb16tobgr16; break;
1813
        case 0x46: conv= rgb24tobgr16; break;
1814
        case 0x48: conv= rgb32tobgr16; break;
1815
        case 0x63: conv= rgb15tobgr24; break;
1816
        case 0x64: conv= rgb16tobgr24; break;
1817
        case 0x66: conv= rgb24tobgr24; break;
1818
        case 0x68: conv= rgb32tobgr24; break;
1819
        case 0x83: conv= rgb15tobgr32; break;
1820
        case 0x84: conv= rgb16tobgr32; break;
1821
        case 0x86: conv= rgb24tobgr32; break;
1822
        case 0x88: conv= rgb32tobgr32; break;
1823
        default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1824
                        sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1825
        }
1826
    }else{
1827
        av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1828
               sws_format_name(srcFormat), sws_format_name(dstFormat));
1829
    }
1830

    
1831
    if(conv)
1832
    {
1833
        uint8_t *srcPtr= src[0];
1834
        if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
1835
            srcPtr += ALT32_CORR;
1836

    
1837
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1838
            conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1839
        else
1840
        {
1841
            int i;
1842
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1843

    
1844
            for (i=0; i<srcSliceH; i++)
1845
            {
1846
                conv(srcPtr, dstPtr, c->srcW*srcBpp);
1847
                srcPtr+= srcStride[0];
1848
                dstPtr+= dstStride[0];
1849
            }
1850
        }
1851
    }
1852
    return srcSliceH;
1853
}
1854

    
1855
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1856
                              int srcSliceH, uint8_t* dst[], int dstStride[]){
1857

    
1858
    rgb24toyv12(
1859
        src[0],
1860
        dst[0]+ srcSliceY    *dstStride[0],
1861
        dst[1]+(srcSliceY>>1)*dstStride[1],
1862
        dst[2]+(srcSliceY>>1)*dstStride[2],
1863
        c->srcW, srcSliceH,
1864
        dstStride[0], dstStride[1], srcStride[0]);
1865
    return srcSliceH;
1866
}
1867

    
1868
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1869
                             int srcSliceH, uint8_t* dst[], int dstStride[]){
1870
    int i;
1871

    
1872
    /* copy Y */
1873
    if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1874
        memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1875
    else{
1876
        uint8_t *srcPtr= src[0];
1877
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1878

    
1879
        for (i=0; i<srcSliceH; i++)
1880
        {
1881
            memcpy(dstPtr, srcPtr, c->srcW);
1882
            srcPtr+= srcStride[0];
1883
            dstPtr+= dstStride[0];
1884
        }
1885
    }
1886

    
1887
    if (c->dstFormat==PIX_FMT_YUV420P){
1888
        planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
1889
        planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
1890
    }else{
1891
        planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
1892
        planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
1893
    }
1894
    return srcSliceH;
1895
}
1896

    
1897
/* unscaled copy like stuff (assumes nearly identical formats) */
1898
static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1899
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1900
{
1901
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1902
        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1903
    else
1904
    {
1905
        int i;
1906
        uint8_t *srcPtr= src[0];
1907
        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1908
        int length=0;
1909

    
1910
        /* universal length finder */
1911
        while(length+c->srcW <= FFABS(dstStride[0])
1912
           && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1913
        assert(length!=0);
1914

    
1915
        for (i=0; i<srcSliceH; i++)
1916
        {
1917
            memcpy(dstPtr, srcPtr, length);
1918
            srcPtr+= srcStride[0];
1919
            dstPtr+= dstStride[0];
1920
        }
1921
    }
1922
    return srcSliceH;
1923
}
1924

    
1925
static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1926
                      int srcSliceH, uint8_t* dst[], int dstStride[])
1927
{
1928
    int plane;
1929
    for (plane=0; plane<3; plane++)
1930
    {
1931
        int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1932
        int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1933
        int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1934

    
1935
        if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
1936
        {
1937
            if (!isGray(c->dstFormat))
1938
                memset(dst[plane], 128, dstStride[plane]*height);
1939
        }
1940
        else
1941
        {
1942
            if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1943
                memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1944
            else
1945
            {
1946
                int i;
1947
                uint8_t *srcPtr= src[plane];
1948
                uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1949
                for (i=0; i<height; i++)
1950
                {
1951
                    memcpy(dstPtr, srcPtr, length);
1952
                    srcPtr+= srcStride[plane];
1953
                    dstPtr+= dstStride[plane];
1954
                }
1955
            }
1956
        }
1957
    }
1958
    return srcSliceH;
1959
}
1960

    
1961
static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1962
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
1963

    
1964
    int length= c->srcW;
1965
    int y=      srcSliceY;
1966
    int height= srcSliceH;
1967
    int i, j;
1968
    uint8_t *srcPtr= src[0];
1969
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1970

    
1971
    if (!isGray(c->dstFormat)){
1972
        int height= -((-srcSliceH)>>c->chrDstVSubSample);
1973
        memset(dst[1], 128, dstStride[1]*height);
1974
        memset(dst[2], 128, dstStride[2]*height);
1975
    }
1976
    if (c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;
1977
    for (i=0; i<height; i++)
1978
    {
1979
        for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1980
        srcPtr+= srcStride[0];
1981
        dstPtr+= dstStride[0];
1982
    }
1983
    return srcSliceH;
1984
}
1985

    
1986
static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1987
                        int srcSliceH, uint8_t* dst[], int dstStride[]){
1988

    
1989
    int length= c->srcW;
1990
    int y=      srcSliceY;
1991
    int height= srcSliceH;
1992
    int i, j;
1993
    uint8_t *srcPtr= src[0];
1994
    uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1995
    for (i=0; i<height; i++)
1996
    {
1997
        for (j=0; j<length; j++)
1998
        {
1999
            dstPtr[j<<1] = srcPtr[j];
2000
            dstPtr[(j<<1)+1] = srcPtr[j];
2001
        }
2002
        srcPtr+= srcStride[0];
2003
        dstPtr+= dstStride[0];
2004
    }
2005
    return srcSliceH;
2006
}
2007

    
2008
static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2009
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
2010

    
2011
    int length= c->srcW;
2012
    int y=      srcSliceY;
2013
    int height= srcSliceH;
2014
    int i, j;
2015
    uint16_t *srcPtr= (uint16_t*)src[0];
2016
    uint16_t *dstPtr= (uint16_t*)(dst[0] + dstStride[0]*y/2);
2017
    for (i=0; i<height; i++)
2018
    {
2019
        for (j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);
2020
        srcPtr+= srcStride[0]/2;
2021
        dstPtr+= dstStride[0]/2;
2022
    }
2023
    return srcSliceH;
2024
}
2025

    
2026

    
2027
static void getSubSampleFactors(int *h, int *v, int format){
2028
    switch(format){
2029
    case PIX_FMT_UYVY422:
2030
    case PIX_FMT_YUYV422:
2031
        *h=1;
2032
        *v=0;
2033
        break;
2034
    case PIX_FMT_YUV420P:
2035
    case PIX_FMT_YUVA420P:
2036
    case PIX_FMT_GRAY16BE:
2037
    case PIX_FMT_GRAY16LE:
2038
    case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2039
    case PIX_FMT_NV12:
2040
    case PIX_FMT_NV21:
2041
        *h=1;
2042
        *v=1;
2043
        break;
2044
    case PIX_FMT_YUV440P:
2045
        *h=0;
2046
        *v=1;
2047
        break;
2048
    case PIX_FMT_YUV410P:
2049
        *h=2;
2050
        *v=2;
2051
        break;
2052
    case PIX_FMT_YUV444P:
2053
        *h=0;
2054
        *v=0;
2055
        break;
2056
    case PIX_FMT_YUV422P:
2057
        *h=1;
2058
        *v=0;
2059
        break;
2060
    case PIX_FMT_YUV411P:
2061
        *h=2;
2062
        *v=0;
2063
        break;
2064
    default:
2065
        *h=0;
2066
        *v=0;
2067
        break;
2068
    }
2069
}
2070

    
2071
static uint16_t roundToInt16(int64_t f){
2072
    int r= (f + (1<<15))>>16;
2073
         if (r<-0x7FFF) return 0x8000;
2074
    else if (r> 0x7FFF) return 0x7FFF;
2075
    else                return r;
2076
}
2077

    
2078
/**
2079
 * @param inv_table the yuv2rgb coefficients, normally ff_yuv2rgb_coeffs[x]
2080
 * @param fullRange if 1 then the luma range is 0..255 if 0 it is 16..235
2081
 * @return -1 if not supported
2082
 */
2083
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2084
    int64_t crv =  inv_table[0];
2085
    int64_t cbu =  inv_table[1];
2086
    int64_t cgu = -inv_table[2];
2087
    int64_t cgv = -inv_table[3];
2088
    int64_t cy  = 1<<16;
2089
    int64_t oy  = 0;
2090

    
2091
    memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2092
    memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
2093

    
2094
    c->brightness= brightness;
2095
    c->contrast  = contrast;
2096
    c->saturation= saturation;
2097
    c->srcRange  = srcRange;
2098
    c->dstRange  = dstRange;
2099
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return 0;
2100

    
2101
    c->uOffset=   0x0400040004000400LL;
2102
    c->vOffset=   0x0400040004000400LL;
2103

    
2104
    if (!srcRange){
2105
        cy= (cy*255) / 219;
2106
        oy= 16<<16;
2107
    }else{
2108
        crv= (crv*224) / 255;
2109
        cbu= (cbu*224) / 255;
2110
        cgu= (cgu*224) / 255;
2111
        cgv= (cgv*224) / 255;
2112
    }
2113

    
2114
    cy = (cy *contrast             )>>16;
2115
    crv= (crv*contrast * saturation)>>32;
2116
    cbu= (cbu*contrast * saturation)>>32;
2117
    cgu= (cgu*contrast * saturation)>>32;
2118
    cgv= (cgv*contrast * saturation)>>32;
2119

    
2120
    oy -= 256*brightness;
2121

    
2122
    c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
2123
    c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
2124
    c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2125
    c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2126
    c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2127
    c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
2128

    
2129
    c->yuv2rgb_y_coeff  = (int16_t)roundToInt16(cy <<13);
2130
    c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2131
    c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2132
    c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2133
    c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2134
    c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2135

    
2136
    sws_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2137
    //FIXME factorize
2138

    
2139
#ifdef COMPILE_ALTIVEC
2140
    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2141
        sws_yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
2142
#endif
2143
    return 0;
2144
}
2145

    
2146
/**
2147
 * @return -1 if not supported
2148
 */
2149
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2150
    if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2151

    
2152
    *inv_table = c->srcColorspaceTable;
2153
    *table     = c->dstColorspaceTable;
2154
    *srcRange  = c->srcRange;
2155
    *dstRange  = c->dstRange;
2156
    *brightness= c->brightness;
2157
    *contrast  = c->contrast;
2158
    *saturation= c->saturation;
2159

    
2160
    return 0;
2161
}
2162

    
2163
static int handle_jpeg(enum PixelFormat *format)
2164
{
2165
    switch (*format) {
2166
        case PIX_FMT_YUVJ420P:
2167
            *format = PIX_FMT_YUV420P;
2168
            return 1;
2169
        case PIX_FMT_YUVJ422P:
2170
            *format = PIX_FMT_YUV422P;
2171
            return 1;
2172
        case PIX_FMT_YUVJ444P:
2173
            *format = PIX_FMT_YUV444P;
2174
            return 1;
2175
        case PIX_FMT_YUVJ440P:
2176
            *format = PIX_FMT_YUV440P;
2177
            return 1;
2178
        default:
2179
            return 0;
2180
    }
2181
}
2182

    
2183
SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2184
                           SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){
2185

    
2186
    SwsContext *c;
2187
    int i;
2188
    int usesVFilter, usesHFilter;
2189
    int unscaled, needsDither;
2190
    int srcRange, dstRange;
2191
    SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2192
#if ARCH_X86
2193
    if (flags & SWS_CPU_CAPS_MMX)
2194
        __asm__ volatile("emms\n\t"::: "memory");
2195
#endif
2196

    
2197
#if !defined(RUNTIME_CPUDETECT) || !CONFIG_GPL //ensure that the flags match the compiled variant if cpudetect is off
2198
    flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2199
#if   HAVE_MMX2
2200
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2201
#elif HAVE_AMD3DNOW
2202
    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2203
#elif HAVE_MMX
2204
    flags |= SWS_CPU_CAPS_MMX;
2205
#elif HAVE_ALTIVEC
2206
    flags |= SWS_CPU_CAPS_ALTIVEC;
2207
#elif ARCH_BFIN
2208
    flags |= SWS_CPU_CAPS_BFIN;
2209
#endif
2210
#endif /* RUNTIME_CPUDETECT */
2211
    if (clip_table[512] != 255) globalInit();
2212
    if (!rgb15to16) sws_rgb2rgb_init(flags);
2213

    
2214
    unscaled = (srcW == dstW && srcH == dstH);
2215
    needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2216
        && (fmt_depth(dstFormat))<24
2217
        && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2218

    
2219
    srcRange = handle_jpeg(&srcFormat);
2220
    dstRange = handle_jpeg(&dstFormat);
2221

    
2222
    if (!isSupportedIn(srcFormat))
2223
    {
2224
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2225
        return NULL;
2226
    }
2227
    if (!isSupportedOut(dstFormat))
2228
    {
2229
        av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2230
        return NULL;
2231
    }
2232

    
2233
    i= flags & ( SWS_POINT
2234
                |SWS_AREA
2235
                |SWS_BILINEAR
2236
                |SWS_FAST_BILINEAR
2237
                |SWS_BICUBIC
2238
                |SWS_X
2239
                |SWS_GAUSS
2240
                |SWS_LANCZOS
2241
                |SWS_SINC
2242
                |SWS_SPLINE
2243
                |SWS_BICUBLIN);
2244
    if(!i || (i & (i-1)))
2245
    {
2246
        av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2247
        return NULL;
2248
    }
2249

    
2250
    /* sanity check */
2251
    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
2252
    {
2253
        av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2254
               srcW, srcH, dstW, dstH);
2255
        return NULL;
2256
    }
2257
    if(srcW > VOFW || dstW > VOFW){
2258
        av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2259
        return NULL;
2260
    }
2261

    
2262
    if (!dstFilter) dstFilter= &dummyFilter;
2263
    if (!srcFilter) srcFilter= &dummyFilter;
2264

    
2265
    c= av_mallocz(sizeof(SwsContext));
2266

    
2267
    c->av_class = &sws_context_class;
2268
    c->srcW= srcW;
2269
    c->srcH= srcH;
2270
    c->dstW= dstW;
2271
    c->dstH= dstH;
2272
    c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2273
    c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2274
    c->flags= flags;
2275
    c->dstFormat= dstFormat;
2276
    c->srcFormat= srcFormat;
2277
    c->vRounder= 4* 0x0001000100010001ULL;
2278

    
2279
    usesHFilter= usesVFilter= 0;
2280
    if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2281
    if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2282
    if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2283
    if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2284
    if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2285
    if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2286
    if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2287
    if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2288

    
2289
    getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2290
    getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2291

    
2292
    // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2293
    if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2294

    
2295
    // drop some chroma lines if the user wants it
2296
    c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2297
    c->chrSrcVSubSample+= c->vChrDrop;
2298

    
2299
    // drop every other pixel for chroma calculation unless user wants full chroma
2300
    if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2301
      && srcFormat!=PIX_FMT_RGB8      && srcFormat!=PIX_FMT_BGR8
2302
      && srcFormat!=PIX_FMT_RGB4      && srcFormat!=PIX_FMT_BGR4
2303
      && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2304
      && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2305
        c->chrSrcHSubSample=1;
2306

    
2307
    if (param){
2308
        c->param[0] = param[0];
2309
        c->param[1] = param[1];
2310
    }else{
2311
        c->param[0] =
2312
        c->param[1] = SWS_PARAM_DEFAULT;
2313
    }
2314

    
2315
    c->chrIntHSubSample= c->chrDstHSubSample;
2316
    c->chrIntVSubSample= c->chrSrcVSubSample;
2317

    
2318
    // Note the -((-x)>>y) is so that we always round toward +inf.
2319
    c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2320
    c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2321
    c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2322
    c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2323

    
2324
    sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2325

    
2326
    /* unscaled special cases */
2327
    if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2328
    {
2329
        /* yv12_to_nv12 */
2330
        if (srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2331
        {
2332
            c->swScale= PlanarToNV12Wrapper;
2333
        }
2334
        /* yuv2bgr */
2335
        if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat))
2336
            && !(flags & SWS_ACCURATE_RND) && !(dstH&1))
2337
        {
2338
            c->swScale= sws_yuv2rgb_get_func_ptr(c);
2339
        }
2340

    
2341
        if (srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P && !(flags & SWS_BITEXACT))
2342
        {
2343
            c->swScale= yvu9toyv12Wrapper;
2344
        }
2345

    
2346
        /* bgr24toYV12 */
2347
        if (srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P && !(flags & SWS_ACCURATE_RND))
2348
            c->swScale= bgr24toyv12Wrapper;
2349

    
2350
        /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2351
        if (  (isBGR(srcFormat) || isRGB(srcFormat))
2352
           && (isBGR(dstFormat) || isRGB(dstFormat))
2353
           && srcFormat != PIX_FMT_BGR8      && dstFormat != PIX_FMT_BGR8
2354
           && srcFormat != PIX_FMT_RGB8      && dstFormat != PIX_FMT_RGB8
2355
           && srcFormat != PIX_FMT_BGR4      && dstFormat != PIX_FMT_BGR4
2356
           && srcFormat != PIX_FMT_RGB4      && dstFormat != PIX_FMT_RGB4
2357
           && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2358
           && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2359
           && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2360
           && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2361
                                             && dstFormat != PIX_FMT_RGB32_1
2362
                                             && dstFormat != PIX_FMT_BGR32_1
2363
           && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2364
             c->swScale= rgb2rgbWrapper;
2365

    
2366
        if ((usePal(srcFormat) && (
2367
                 dstFormat == PIX_FMT_RGB32   ||
2368
                 dstFormat == PIX_FMT_RGB32_1 ||
2369
                 dstFormat == PIX_FMT_RGB24   ||
2370
                 dstFormat == PIX_FMT_BGR32   ||
2371
                 dstFormat == PIX_FMT_BGR32_1 ||
2372
                 dstFormat == PIX_FMT_BGR24)))
2373
             c->swScale= pal2rgbWrapper;
2374

    
2375
        if (srcFormat == PIX_FMT_YUV422P)
2376
        {
2377
            if (dstFormat == PIX_FMT_YUYV422)
2378
                c->swScale= YUV422PToYuy2Wrapper;
2379
            else if (dstFormat == PIX_FMT_UYVY422)
2380
                c->swScale= YUV422PToUyvyWrapper;
2381
        }
2382

    
2383
        /* LQ converters if -sws 0 or -sws 4*/
2384
        if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2385
            /* yv12_to_yuy2 */
2386
            if (srcFormat == PIX_FMT_YUV420P)
2387
            {
2388
                if (dstFormat == PIX_FMT_YUYV422)
2389
                    c->swScale= PlanarToYuy2Wrapper;
2390
                else if (dstFormat == PIX_FMT_UYVY422)
2391
                    c->swScale= PlanarToUyvyWrapper;
2392
            }
2393
        }
2394

    
2395
#ifdef COMPILE_ALTIVEC
2396
        if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2397
            !(c->flags & SWS_BITEXACT) &&
2398
            srcFormat == PIX_FMT_YUV420P) {
2399
          // unscaled YV12 -> packed YUV, we want speed
2400
          if (dstFormat == PIX_FMT_YUYV422)
2401
              c->swScale= yv12toyuy2_unscaled_altivec;
2402
          else if (dstFormat == PIX_FMT_UYVY422)
2403
              c->swScale= yv12touyvy_unscaled_altivec;
2404
        }
2405
#endif
2406

    
2407
        /* simple copy */
2408
        if (  srcFormat == dstFormat
2409
            || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2410
            || (isPlanarYUV(dstFormat) && isGray(srcFormat)))
2411
        {
2412
            if (isPacked(c->srcFormat))
2413
                c->swScale= packedCopy;
2414
            else /* Planar YUV or gray */
2415
                c->swScale= planarCopy;
2416
        }
2417

    
2418
        /* gray16{le,be} conversions */
2419
        if (isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))
2420
        {
2421
            c->swScale= gray16togray;
2422
        }
2423
        if ((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))
2424
        {
2425
            c->swScale= graytogray16;
2426
        }
2427
        if (srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
2428
        {
2429
            c->swScale= gray16swap;
2430
        }
2431

    
2432
#if ARCH_BFIN
2433
        if (flags & SWS_CPU_CAPS_BFIN)
2434
            ff_bfin_get_unscaled_swscale (c);
2435
#endif
2436

    
2437
        if (c->swScale){
2438
            if (flags&SWS_PRINT_INFO)
2439
                av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2440
                                sws_format_name(srcFormat), sws_format_name(dstFormat));
2441
            return c;
2442
        }
2443
    }
2444

    
2445
    if (flags & SWS_CPU_CAPS_MMX2)
2446
    {
2447
        c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2448
        if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2449
        {
2450
            if (flags&SWS_PRINT_INFO)
2451
                av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2452
        }
2453
        if (usesHFilter) c->canMMX2BeUsed=0;
2454
    }
2455
    else
2456
        c->canMMX2BeUsed=0;
2457

    
2458
    c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2459
    c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2460

    
2461
    // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2462
    // but only for the FAST_BILINEAR mode otherwise do correct scaling
2463
    // n-2 is the last chrominance sample available
2464
    // this is not perfect, but no one should notice the difference, the more correct variant
2465
    // would be like the vertical one, but that would require some special code for the
2466
    // first and last pixel
2467
    if (flags&SWS_FAST_BILINEAR)
2468
    {
2469
        if (c->canMMX2BeUsed)
2470
        {
2471
            c->lumXInc+= 20;
2472
            c->chrXInc+= 20;
2473
        }
2474
        //we don't use the x86 asm scaler if MMX is available
2475
        else if (flags & SWS_CPU_CAPS_MMX)
2476
        {
2477
            c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2478
            c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2479
        }
2480
    }
2481

    
2482
    /* precalculate horizontal scaler filter coefficients */
2483
    {
2484
        const int filterAlign=
2485
            (flags & SWS_CPU_CAPS_MMX) ? 4 :
2486
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2487
            1;
2488

    
2489
        initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2490
                   srcW      ,       dstW, filterAlign, 1<<14,
2491
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2492
                   srcFilter->lumH, dstFilter->lumH, c->param);
2493
        initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2494
                   c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2495
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2496
                   srcFilter->chrH, dstFilter->chrH, c->param);
2497

    
2498
#define MAX_FUNNY_CODE_SIZE 10000
2499
#if defined(COMPILE_MMX2)
2500
// can't downscale !!!
2501
        if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2502
        {
2503
#ifdef MAP_ANONYMOUS
2504
            c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2505
            c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2506
#else
2507
            c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2508
            c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2509
#endif
2510

    
2511
            c->lumMmx2Filter   = av_malloc((dstW        /8+8)*sizeof(int16_t));
2512
            c->chrMmx2Filter   = av_malloc((c->chrDstW  /4+8)*sizeof(int16_t));
2513
            c->lumMmx2FilterPos= av_malloc((dstW      /2/8+8)*sizeof(int32_t));
2514
            c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2515

    
2516
            initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2517
            initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2518
        }
2519
#endif /* defined(COMPILE_MMX2) */
2520
    } // initialize horizontal stuff
2521

    
2522

    
2523

    
2524
    /* precalculate vertical scaler filter coefficients */
2525
    {
2526
        const int filterAlign=
2527
            (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2528
            (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2529
            1;
2530

    
2531
        initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2532
                   srcH      ,        dstH, filterAlign, (1<<12),
2533
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2534
                   srcFilter->lumV, dstFilter->lumV, c->param);
2535
        initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2536
                   c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2537
                   (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2538
                   srcFilter->chrV, dstFilter->chrV, c->param);
2539

    
2540
#if HAVE_ALTIVEC
2541
        c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2542
        c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2543

    
2544
        for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2545
            int j;
2546
            short *p = (short *)&c->vYCoeffsBank[i];
2547
            for (j=0;j<8;j++)
2548
                p[j] = c->vLumFilter[i];
2549
        }
2550

    
2551
        for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2552
            int j;
2553
            short *p = (short *)&c->vCCoeffsBank[i];
2554
            for (j=0;j<8;j++)
2555
                p[j] = c->vChrFilter[i];
2556
        }
2557
#endif
2558
    }
2559

    
2560
    // calculate buffer sizes so that they won't run out while handling these damn slices
2561
    c->vLumBufSize= c->vLumFilterSize;
2562
    c->vChrBufSize= c->vChrFilterSize;
2563
    for (i=0; i<dstH; i++)
2564
    {
2565
        int chrI= i*c->chrDstH / dstH;
2566
        int nextSlice= FFMAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2567
                           ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2568

    
2569
        nextSlice>>= c->chrSrcVSubSample;
2570
        nextSlice<<= c->chrSrcVSubSample;
2571
        if (c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2572
            c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2573
        if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2574
            c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2575
    }
2576

    
2577
    // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2578
    c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2579
    c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2580
    //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)
2581
    /* align at 16 bytes for AltiVec */
2582
    for (i=0; i<c->vLumBufSize; i++)
2583
        c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2584
    for (i=0; i<c->vChrBufSize; i++)
2585
        c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2586

    
2587
    //try to avoid drawing green stuff between the right end and the stride end
2588
    for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2589

    
2590
    assert(2*VOFW == VOF);
2591

    
2592
    assert(c->chrDstH <= dstH);
2593

    
2594
    if (flags&SWS_PRINT_INFO)
2595
    {
2596
#ifdef DITHER1XBPP
2597
        const char *dither= " dithered";
2598
#else
2599
        const char *dither= "";
2600
#endif
2601
        if (flags&SWS_FAST_BILINEAR)
2602
            av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2603
        else if (flags&SWS_BILINEAR)
2604
            av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2605
        else if (flags&SWS_BICUBIC)
2606
            av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2607
        else if (flags&SWS_X)
2608
            av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2609
        else if (flags&SWS_POINT)
2610
            av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2611
        else if (flags&SWS_AREA)
2612
            av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2613
        else if (flags&SWS_BICUBLIN)
2614
            av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2615
        else if (flags&SWS_GAUSS)
2616
            av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2617
        else if (flags&SWS_SINC)
2618
            av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2619
        else if (flags&SWS_LANCZOS)
2620
            av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2621
        else if (flags&SWS_SPLINE)
2622
            av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2623
        else
2624
            av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2625

    
2626
        if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2627
            av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2628
                   sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2629
        else
2630
            av_log(c, AV_LOG_INFO, "from %s to %s ",
2631
                   sws_format_name(srcFormat), sws_format_name(dstFormat));
2632

    
2633
        if (flags & SWS_CPU_CAPS_MMX2)
2634
            av_log(c, AV_LOG_INFO, "using MMX2\n");
2635
        else if (flags & SWS_CPU_CAPS_3DNOW)
2636
            av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2637
        else if (flags & SWS_CPU_CAPS_MMX)
2638
            av_log(c, AV_LOG_INFO, "using MMX\n");
2639
        else if (flags & SWS_CPU_CAPS_ALTIVEC)
2640
            av_log(c, AV_LOG_INFO, "using AltiVec\n");
2641
        else
2642
            av_log(c, AV_LOG_INFO, "using C\n");
2643
    }
2644

    
2645
    if (flags & SWS_PRINT_INFO)
2646
    {
2647
        if (flags & SWS_CPU_CAPS_MMX)
2648
        {
2649
            if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2650
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2651
            else
2652
            {
2653
                if (c->hLumFilterSize==4)
2654
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2655
                else if (c->hLumFilterSize==8)
2656
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2657
                else
2658
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2659

    
2660
                if (c->hChrFilterSize==4)
2661
                    av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2662
                else if (c->hChrFilterSize==8)
2663
                    av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2664
                else
2665
                    av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2666
            }
2667
        }
2668
        else
2669
        {
2670
#if ARCH_X86
2671
            av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
2672
#else
2673
            if (flags & SWS_FAST_BILINEAR)
2674
                av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
2675
            else
2676
                av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
2677
#endif
2678
        }
2679
        if (isPlanarYUV(dstFormat))
2680
        {
2681
            if (c->vLumFilterSize==1)
2682
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2683
            else
2684
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2685
        }
2686
        else
2687
        {
2688
            if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
2689
                av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2690
                       "      2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2691
            else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
2692
                av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2693
            else
2694
                av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2695
        }
2696

    
2697
        if (dstFormat==PIX_FMT_BGR24)
2698
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
2699
                   (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2700
        else if (dstFormat==PIX_FMT_RGB32)
2701
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2702
        else if (dstFormat==PIX_FMT_BGR565)
2703
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2704
        else if (dstFormat==PIX_FMT_BGR555)
2705
            av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2706

    
2707
        av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2708
    }
2709
    if (flags & SWS_PRINT_INFO)
2710
    {
2711
        av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2712
               c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2713
        av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2714
               c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2715
    }
2716

    
2717
    c->swScale= getSwsFunc(flags);
2718
    return c;
2719
}
2720

    
2721
/**
2722
 * swscale wrapper, so we don't need to export the SwsContext.
2723
 * Assumes planar YUV to be in YUV order instead of YVU.
2724
 */
2725
int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2726
              int srcSliceH, uint8_t* dst[], int dstStride[]){
2727
    int i;
2728
    uint8_t* src2[4]= {src[0], src[1], src[2]};
2729

    
2730
    if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2731
        av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2732
        return 0;
2733
    }
2734
    if (c->sliceDir == 0) {
2735
        if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2736
    }
2737

    
2738
    if (usePal(c->srcFormat)){
2739
        for (i=0; i<256; i++){
2740
            int p, r, g, b,y,u,v;
2741
            if(c->srcFormat == PIX_FMT_PAL8){
2742
                p=((uint32_t*)(src[1]))[i];
2743
                r= (p>>16)&0xFF;
2744
                g= (p>> 8)&0xFF;
2745
                b=  p     &0xFF;
2746
            }else if(c->srcFormat == PIX_FMT_RGB8){
2747
                r= (i>>5    )*36;
2748
                g= ((i>>2)&7)*36;
2749
                b= (i&3     )*85;
2750
            }else if(c->srcFormat == PIX_FMT_BGR8){
2751
                b= (i>>6    )*85;
2752
                g= ((i>>3)&7)*36;
2753
                r= (i&7     )*36;
2754
            }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
2755
                r= (i>>3    )*255;
2756
                g= ((i>>1)&3)*85;
2757
                b= (i&1     )*255;
2758
            }else {
2759
                assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2760
                b= (i>>3    )*255;
2761
                g= ((i>>1)&3)*85;
2762
                r= (i&1     )*255;
2763
            }
2764
            y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2765
            u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2766
            v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2767
            c->pal_yuv[i]= y + (u<<8) + (v<<16);
2768

    
2769

    
2770
            switch(c->dstFormat) {
2771
            case PIX_FMT_BGR32:
2772
#ifndef WORDS_BIGENDIAN
2773
            case PIX_FMT_RGB24:
2774
#endif
2775
                c->pal_rgb[i]=  r + (g<<8) + (b<<16);
2776
                break;
2777
            case PIX_FMT_BGR32_1:
2778
#ifdef  WORDS_BIGENDIAN
2779
            case PIX_FMT_BGR24:
2780
#endif
2781
                c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2782
                break;
2783
            case PIX_FMT_RGB32_1:
2784
#ifdef  WORDS_BIGENDIAN
2785
            case PIX_FMT_RGB24:
2786
#endif
2787
                c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2788
                break;
2789
            case PIX_FMT_RGB32:
2790
#ifndef WORDS_BIGENDIAN
2791
            case PIX_FMT_BGR24:
2792
#endif
2793
            default:
2794
                c->pal_rgb[i]=  b + (g<<8) + (r<<16);
2795
            }
2796
        }
2797
    }
2798

    
2799
    // copy strides, so they can safely be modified
2800
    if (c->sliceDir == 1) {
2801
        // slices go from top to bottom
2802
        int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};
2803
        int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};
2804
        return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
2805
    } else {
2806
        // slices go from bottom to top => we flip the image internally
2807
        uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],
2808
                           dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],
2809
                           dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};
2810
        int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};
2811
        int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};
2812

    
2813
        src2[0] += (srcSliceH-1)*srcStride[0];
2814
        if (!usePal(c->srcFormat))
2815
            src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2816
        src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2817

    
2818
        return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2819
    }
2820
}
2821

    
2822
#if LIBSWSCALE_VERSION_MAJOR < 1
2823
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2824
                      int srcSliceH, uint8_t* dst[], int dstStride[]){
2825
    return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2826
}
2827
#endif
2828

    
2829
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
2830
                                float lumaSharpen, float chromaSharpen,
2831
                                float chromaHShift, float chromaVShift,
2832
                                int verbose)
2833
{
2834
    SwsFilter *filter= av_malloc(sizeof(SwsFilter));
2835

    
2836
    if (lumaGBlur!=0.0){
2837
        filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
2838
        filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
2839
    }else{
2840
        filter->lumH= sws_getIdentityVec();
2841
        filter->lumV= sws_getIdentityVec();
2842
    }
2843

    
2844
    if (chromaGBlur!=0.0){
2845
        filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
2846
        filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
2847
    }else{
2848
        filter->chrH= sws_getIdentityVec();
2849
        filter->chrV= sws_getIdentityVec();
2850
    }
2851

    
2852
    if (chromaSharpen!=0.0){
2853
        SwsVector *id= sws_getIdentityVec();
2854
        sws_scaleVec(filter->chrH, -chromaSharpen);
2855
        sws_scaleVec(filter->chrV, -chromaSharpen);
2856
        sws_addVec(filter->chrH, id);
2857
        sws_addVec(filter->chrV, id);
2858
        sws_freeVec(id);
2859
    }
2860

    
2861
    if (lumaSharpen!=0.0){
2862
        SwsVector *id= sws_getIdentityVec();
2863
        sws_scaleVec(filter->lumH, -lumaSharpen);
2864
        sws_scaleVec(filter->lumV, -lumaSharpen);
2865
        sws_addVec(filter->lumH, id);
2866
        sws_addVec(filter->lumV, id);
2867
        sws_freeVec(id);
2868
    }
2869

    
2870
    if (chromaHShift != 0.0)
2871
        sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
2872

    
2873
    if (chromaVShift != 0.0)
2874
        sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
2875

    
2876
    sws_normalizeVec(filter->chrH, 1.0);
2877
    sws_normalizeVec(filter->chrV, 1.0);
2878
    sws_normalizeVec(filter->lumH, 1.0);
2879
    sws_normalizeVec(filter->lumV, 1.0);
2880

    
2881
    if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
2882
    if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
2883

    
2884
    return filter;
2885
}
2886

    
2887
SwsVector *sws_getGaussianVec(double variance, double quality){
2888
    const int length= (int)(variance*quality + 0.5) | 1;
2889
    int i;
2890
    double *coeff= av_malloc(length*sizeof(double));
2891
    double middle= (length-1)*0.5;
2892
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2893

    
2894
    vec->coeff= coeff;
2895
    vec->length= length;
2896

    
2897
    for (i=0; i<length; i++)
2898
    {
2899
        double dist= i-middle;
2900
        coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
2901
    }
2902

    
2903
    sws_normalizeVec(vec, 1.0);
2904

    
2905
    return vec;
2906
}
2907

    
2908
SwsVector *sws_getConstVec(double c, int length){
2909
    int i;
2910
    double *coeff= av_malloc(length*sizeof(double));
2911
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2912

    
2913
    vec->coeff= coeff;
2914
    vec->length= length;
2915

    
2916
    for (i=0; i<length; i++)
2917
        coeff[i]= c;
2918

    
2919
    return vec;
2920
}
2921

    
2922

    
2923
SwsVector *sws_getIdentityVec(void){
2924
    return sws_getConstVec(1.0, 1);
2925
}
2926

    
2927
double sws_dcVec(SwsVector *a){
2928
    int i;
2929
    double sum=0;
2930

    
2931
    for (i=0; i<a->length; i++)
2932
        sum+= a->coeff[i];
2933

    
2934
    return sum;
2935
}
2936

    
2937
void sws_scaleVec(SwsVector *a, double scalar){
2938
    int i;
2939

    
2940
    for (i=0; i<a->length; i++)
2941
        a->coeff[i]*= scalar;
2942
}
2943

    
2944
void sws_normalizeVec(SwsVector *a, double height){
2945
    sws_scaleVec(a, height/sws_dcVec(a));
2946
}
2947

    
2948
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
2949
    int length= a->length + b->length - 1;
2950
    double *coeff= av_malloc(length*sizeof(double));
2951
    int i, j;
2952
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2953

    
2954
    vec->coeff= coeff;
2955
    vec->length= length;
2956

    
2957
    for (i=0; i<length; i++) coeff[i]= 0.0;
2958

    
2959
    for (i=0; i<a->length; i++)
2960
    {
2961
        for (j=0; j<b->length; j++)
2962
        {
2963
            coeff[i+j]+= a->coeff[i]*b->coeff[j];
2964
        }
2965
    }
2966

    
2967
    return vec;
2968
}
2969

    
2970
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
2971
    int length= FFMAX(a->length, b->length);
2972
    double *coeff= av_malloc(length*sizeof(double));
2973
    int i;
2974
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2975

    
2976
    vec->coeff= coeff;
2977
    vec->length= length;
2978

    
2979
    for (i=0; i<length; i++) coeff[i]= 0.0;
2980

    
2981
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2982
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
2983

    
2984
    return vec;
2985
}
2986

    
2987
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
2988
    int length= FFMAX(a->length, b->length);
2989
    double *coeff= av_malloc(length*sizeof(double));
2990
    int i;
2991
    SwsVector *vec= av_malloc(sizeof(SwsVector));
2992

    
2993
    vec->coeff= coeff;
2994
    vec->length= length;
2995

    
2996
    for (i=0; i<length; i++) coeff[i]= 0.0;
2997

    
2998
    for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2999
    for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3000

    
3001
    return vec;
3002
}
3003

    
3004
/* shift left / or right if "shift" is negative */
3005
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
3006
    int length= a->length + FFABS(shift)*2;
3007
    double *coeff= av_malloc(length*sizeof(double));
3008
    int i;
3009
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3010

    
3011
    vec->coeff= coeff;
3012
    vec->length= length;
3013

    
3014
    for (i=0; i<length; i++) coeff[i]= 0.0;
3015

    
3016
    for (i=0; i<a->length; i++)
3017
    {
3018
        coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3019
    }
3020

    
3021
    return vec;
3022
}
3023

    
3024
void sws_shiftVec(SwsVector *a, int shift){
3025
    SwsVector *shifted= sws_getShiftedVec(a, shift);
3026
    av_free(a->coeff);
3027
    a->coeff= shifted->coeff;
3028
    a->length= shifted->length;
3029
    av_free(shifted);
3030
}
3031

    
3032
void sws_addVec(SwsVector *a, SwsVector *b){
3033
    SwsVector *sum= sws_sumVec(a, b);
3034
    av_free(a->coeff);
3035
    a->coeff= sum->coeff;
3036
    a->length= sum->length;
3037
    av_free(sum);
3038
}
3039

    
3040
void sws_subVec(SwsVector *a, SwsVector *b){
3041
    SwsVector *diff= sws_diffVec(a, b);
3042
    av_free(a->coeff);
3043
    a->coeff= diff->coeff;
3044
    a->length= diff->length;
3045
    av_free(diff);
3046
}
3047

    
3048
void sws_convVec(SwsVector *a, SwsVector *b){
3049
    SwsVector *conv= sws_getConvVec(a, b);
3050
    av_free(a->coeff);
3051
    a->coeff= conv->coeff;
3052
    a->length= conv->length;
3053
    av_free(conv);
3054
}
3055

    
3056
SwsVector *sws_cloneVec(SwsVector *a){
3057
    double *coeff= av_malloc(a->length*sizeof(double));
3058
    int i;
3059
    SwsVector *vec= av_malloc(sizeof(SwsVector));
3060

    
3061
    vec->coeff= coeff;
3062
    vec->length= a->length;
3063

    
3064
    for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
3065

    
3066
    return vec;
3067
}
3068

    
3069
void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){
3070
    int i;
3071
    double max=0;
3072
    double min=0;
3073
    double range;
3074

    
3075
    for (i=0; i<a->length; i++)
3076
        if (a->coeff[i]>max) max= a->coeff[i];
3077

    
3078
    for (i=0; i<a->length; i++)
3079
        if (a->coeff[i]<min) min= a->coeff[i];
3080

    
3081
    range= max - min;
3082

    
3083
    for (i=0; i<a->length; i++)
3084
    {
3085
        int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3086
        av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3087
        for (;x>0; x--) av_log(log_ctx, log_level, " ");
3088
        av_log(log_ctx, log_level, "|\n");
3089
    }
3090
}
3091

    
3092
#if LIBSWSCALE_VERSION_MAJOR < 1
3093
void sws_printVec(SwsVector *a){
3094
    sws_printVec2(a, NULL, AV_LOG_DEBUG);
3095
}
3096
#endif
3097

    
3098
void sws_freeVec(SwsVector *a){
3099
    if (!a) return;
3100
    av_freep(&a->coeff);
3101
    a->length=0;
3102
    av_free(a);
3103
}
3104

    
3105
void sws_freeFilter(SwsFilter *filter){
3106
    if (!filter) return;
3107

    
3108
    if (filter->lumH) sws_freeVec(filter->lumH);
3109
    if (filter->lumV) sws_freeVec(filter->lumV);
3110
    if (filter->chrH) sws_freeVec(filter->chrH);
3111
    if (filter->chrV) sws_freeVec(filter->chrV);
3112
    av_free(filter);
3113
}
3114

    
3115

    
3116
void sws_freeContext(SwsContext *c){
3117
    int i;
3118
    if (!c) return;
3119

    
3120
    if (c->lumPixBuf)
3121
    {
3122
        for (i=0; i<c->vLumBufSize; i++)
3123
            av_freep(&c->lumPixBuf[i]);
3124
        av_freep(&c->lumPixBuf);
3125
    }
3126

    
3127
    if (c->chrPixBuf)
3128
    {
3129
        for (i=0; i<c->vChrBufSize; i++)
3130
            av_freep(&c->chrPixBuf[i]);
3131
        av_freep(&c->chrPixBuf);
3132
    }
3133

    
3134
    av_freep(&c->vLumFilter);
3135
    av_freep(&c->vChrFilter);
3136
    av_freep(&c->hLumFilter);
3137
    av_freep(&c->hChrFilter);
3138
#if HAVE_ALTIVEC
3139
    av_freep(&c->vYCoeffsBank);
3140
    av_freep(&c->vCCoeffsBank);
3141
#endif
3142

    
3143
    av_freep(&c->vLumFilterPos);
3144
    av_freep(&c->vChrFilterPos);
3145
    av_freep(&c->hLumFilterPos);
3146
    av_freep(&c->hChrFilterPos);
3147

    
3148
#if ARCH_X86 && CONFIG_GPL
3149
#ifdef MAP_ANONYMOUS
3150
    if (c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);
3151
    if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
3152
#else
3153
    av_free(c->funnyYCode);
3154
    av_free(c->funnyUVCode);
3155
#endif
3156
    c->funnyYCode=NULL;
3157
    c->funnyUVCode=NULL;
3158
#endif /* ARCH_X86 && CONFIG_GPL */
3159

    
3160
    av_freep(&c->lumMmx2Filter);
3161
    av_freep(&c->chrMmx2Filter);
3162
    av_freep(&c->lumMmx2FilterPos);
3163
    av_freep(&c->chrMmx2FilterPos);
3164
    av_freep(&c->yuvTable);
3165

    
3166
    av_free(c);
3167
}
3168

    
3169
struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3170
                                        int srcW, int srcH, enum PixelFormat srcFormat,
3171
                                        int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3172
                                        SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)
3173
{
3174
    static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3175

    
3176
    if (!param)
3177
        param = default_param;
3178

    
3179
    if (context) {
3180
        if (context->srcW != srcW || context->srcH != srcH ||
3181
            context->srcFormat != srcFormat ||
3182
            context->dstW != dstW || context->dstH != dstH ||
3183
            context->dstFormat != dstFormat || context->flags != flags ||
3184
            context->param[0] != param[0] || context->param[1] != param[1])
3185
        {
3186
            sws_freeContext(context);
3187
            context = NULL;
3188
        }
3189
    }
3190
    if (!context) {
3191
        return sws_getContext(srcW, srcH, srcFormat,
3192
                              dstW, dstH, dstFormat, flags,
3193
                              srcFilter, dstFilter, param);
3194
    }
3195
    return context;
3196
}
3197