Statistics
| Branch: | Revision:

ffmpeg / postproc / swscale.c @ 38d5c282

History | View | Annotate | Download (73.1 KB)

1
/*
2
    Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3

4
    This program is free software; you can redistribute it and/or modify
5
    it under the terms of the GNU General Public License as published by
6
    the Free Software Foundation; either version 2 of the License, or
7
    (at your option) any later version.
8

9
    This program is distributed in the hope that it will be useful,
10
    but WITHOUT ANY WARRANTY; without even the implied warranty of
11
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12
    GNU General Public License for more details.
13

14
    You should have received a copy of the GNU General Public License
15
    along with this program; if not, write to the Free Software
16
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
17
*/
18

    
19
/*
20
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09
21
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
22
  {BGR,RGB}{1,4,8,15,16} support dithering
23
  
24
  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
25
  YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
26
  x -> x
27
  YUV9 -> YV12
28
  YUV9/YV12 -> Y800
29
  Y800 -> YUV9/YV12
30
  BGR24 -> BGR32 & RGB24 -> RGB32
31
  BGR32 -> BGR24 & RGB32 -> RGB24
32
  BGR15 -> BGR16
33
*/
34

    
35
/* 
36
tested special converters (most are tested actually but i didnt write it down ...)
37
 YV12 -> BGR16
38
 YV12 -> YV12
39
 BGR15 -> BGR16
40
 BGR16 -> BGR16
41
 YVU9 -> YV12
42

43
untested special converters
44
  YV12/I420 -> BGR15/BGR24/BGR32 (its the yuv2rgb stuff, so it should be ok)
45
  YV12/I420 -> YV12/I420
46
  YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
47
  BGR24 -> BGR32 & RGB24 -> RGB32
48
  BGR32 -> BGR24 & RGB32 -> RGB24
49
  BGR24 -> YV12
50
*/
51

    
52
#include <inttypes.h>
53
#include <string.h>
54
#include <math.h>
55
#include <stdio.h>
56
#include "../config.h"
57
#include "../mangle.h"
58
#include <assert.h>
59
#ifdef HAVE_MALLOC_H
60
#include <malloc.h>
61
#else
62
#include <stdlib.h>
63
#endif
64
#ifdef HAVE_SYS_MMAN_H
65
#include <sys/mman.h>
66
#endif
67
#include "swscale.h"
68
#include "swscale_internal.h"
69
#include "../cpudetect.h"
70
#include "../bswap.h"
71
#include "../libvo/img_format.h"
72
#include "rgb2rgb.h"
73
#include "../libvo/fastmemcpy.h"
74

    
75
#undef MOVNTQ
76
#undef PAVGB
77

    
78
//#undef HAVE_MMX2
79
//#define HAVE_3DNOW
80
//#undef HAVE_MMX
81
//#undef ARCH_X86
82
//#define WORDS_BIGENDIAN
83
#define DITHER1XBPP
84

    
85
#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
86

    
87
#define RET 0xC3 //near return opcode for X86
88

    
89
#ifdef MP_DEBUG
90
#define ASSERT(x) assert(x);
91
#else
92
#define ASSERT(x) ;
93
#endif
94

    
95
#ifdef M_PI
96
#define PI M_PI
97
#else
98
#define PI 3.14159265358979323846
99
#endif
100

    
101
//FIXME replace this with something faster
102
#define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_YVU9 \
103
                        || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P)
104
#define isYUV(x)       ((x)==IMGFMT_UYVY || (x)==IMGFMT_YUY2 || isPlanarYUV(x))
105
#define isGray(x)      ((x)==IMGFMT_Y800)
106
#define isRGB(x)       (((x)&IMGFMT_RGB_MASK)==IMGFMT_RGB)
107
#define isBGR(x)       (((x)&IMGFMT_BGR_MASK)==IMGFMT_BGR)
108
#define isSupportedIn(x)  ((x)==IMGFMT_YV12 || (x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY\
109
                        || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\
110
                        || (x)==IMGFMT_RGB32|| (x)==IMGFMT_RGB24\
111
                        || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9\
112
                        || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P)
113
#define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY\
114
                        || (x)==IMGFMT_444P || (x)==IMGFMT_422P || (x)==IMGFMT_411P\
115
                        || isRGB(x) || isBGR(x)\
116
                        || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9)
117
#define isPacked(x)    ((x)==IMGFMT_YUY2 || (x)==IMGFMT_UYVY ||isRGB(x) || isBGR(x))
118

    
119
#define RGB2YUV_SHIFT 16
120
#define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
121
#define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
122
#define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
123
#define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
124
#define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
125
#define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
126
#define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
127
#define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
128
#define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
129

    
130
extern const int32_t Inverse_Table_6_9[8][4];
131

    
132
/*
133
NOTES
134
Special versions: fast Y 1:1 scaling (no interpolation in y direction)
135

136
TODO
137
more intelligent missalignment avoidance for the horizontal scaler
138
write special vertical cubic upscale version
139
Optimize C code (yv12 / minmax)
140
add support for packed pixel yuv input & output
141
add support for Y8 output
142
optimize bgr24 & bgr32
143
add BGR4 output support
144
write special BGR->BGR scaler
145
*/
146

    
147
#define ABS(a) ((a) > 0 ? (a) : (-(a)))
148
#define MIN(a,b) ((a) > (b) ? (b) : (a))
149
#define MAX(a,b) ((a) < (b) ? (b) : (a))
150

    
151
#if defined(ARCH_X86) || defined(ARCH_X86_64)
152
static uint64_t attribute_used __attribute__((aligned(8))) bF8=       0xF8F8F8F8F8F8F8F8LL;
153
static uint64_t attribute_used __attribute__((aligned(8))) bFC=       0xFCFCFCFCFCFCFCFCLL;
154
static uint64_t __attribute__((aligned(8))) w10=       0x0010001000100010LL;
155
static uint64_t attribute_used __attribute__((aligned(8))) w02=       0x0002000200020002LL;
156
static uint64_t attribute_used __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL;
157
static uint64_t attribute_used __attribute__((aligned(8))) bm00000111=0x0000000000FFFFFFLL;
158
static uint64_t attribute_used __attribute__((aligned(8))) bm11111000=0xFFFFFFFFFF000000LL;
159
static uint64_t attribute_used __attribute__((aligned(8))) bm01010101=0x00FF00FF00FF00FFLL;
160

    
161
static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
162
static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
163
static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
164
static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
165

    
166
static uint64_t __attribute__((aligned(8))) dither4[2]={
167
        0x0103010301030103LL,
168
        0x0200020002000200LL,};
169

    
170
static uint64_t __attribute__((aligned(8))) dither8[2]={
171
        0x0602060206020602LL,
172
        0x0004000400040004LL,};
173

    
174
static uint64_t __attribute__((aligned(8))) b16Mask=   0x001F001F001F001FLL;
175
static uint64_t attribute_used __attribute__((aligned(8))) g16Mask=   0x07E007E007E007E0LL;
176
static uint64_t attribute_used __attribute__((aligned(8))) r16Mask=   0xF800F800F800F800LL;
177
static uint64_t __attribute__((aligned(8))) b15Mask=   0x001F001F001F001FLL;
178
static uint64_t attribute_used __attribute__((aligned(8))) g15Mask=   0x03E003E003E003E0LL;
179
static uint64_t attribute_used __attribute__((aligned(8))) r15Mask=   0x7C007C007C007C00LL;
180

    
181
static uint64_t attribute_used __attribute__((aligned(8))) M24A=   0x00FF0000FF0000FFLL;
182
static uint64_t attribute_used __attribute__((aligned(8))) M24B=   0xFF0000FF0000FF00LL;
183
static uint64_t attribute_used __attribute__((aligned(8))) M24C=   0x0000FF0000FF0000LL;
184

    
185
#ifdef FAST_BGR2YV12
186
static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
187
static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
188
static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
189
#else
190
static const uint64_t bgr2YCoeff  attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
191
static const uint64_t bgr2UCoeff  attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
192
static const uint64_t bgr2VCoeff  attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
193
#endif
194
static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
195
static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
196
static const uint64_t w1111       attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
197
#endif
198

    
199
// clipping helper table for C implementations:
200
static unsigned char clip_table[768];
201

    
202
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
203
                  
204
extern const uint8_t dither_2x2_4[2][8];
205
extern const uint8_t dither_2x2_8[2][8];
206
extern const uint8_t dither_8x8_32[8][8];
207
extern const uint8_t dither_8x8_73[8][8];
208
extern const uint8_t dither_8x8_220[8][8];
209

    
210
#if defined(ARCH_X86) || defined(ARCH_X86_64)
211
void in_asm_used_var_warning_killer()
212
{
213
 volatile int i= bF8+bFC+w10+
214
 bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+
215
 M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101;
216
 if(i) i=0;
217
}
218
#endif
219

    
220
static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
221
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
222
                                    uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
223
{
224
        //FIXME Optimize (just quickly writen not opti..)
225
        int i;
226
        for(i=0; i<dstW; i++)
227
        {
228
                int val=1<<18;
229
                int j;
230
                for(j=0; j<lumFilterSize; j++)
231
                        val += lumSrc[j][i] * lumFilter[j];
232

    
233
                dest[i]= MIN(MAX(val>>19, 0), 255);
234
        }
235

    
236
        if(uDest != NULL)
237
                for(i=0; i<chrDstW; i++)
238
                {
239
                        int u=1<<18;
240
                        int v=1<<18;
241
                        int j;
242
                        for(j=0; j<chrFilterSize; j++)
243
                        {
244
                                u += chrSrc[j][i] * chrFilter[j];
245
                                v += chrSrc[j][i + 2048] * chrFilter[j];
246
                        }
247

    
248
                        uDest[i]= MIN(MAX(u>>19, 0), 255);
249
                        vDest[i]= MIN(MAX(v>>19, 0), 255);
250
                }
251
}
252

    
253

    
254
#define YSCALE_YUV_2_PACKEDX_C(type) \
255
                for(i=0; i<(dstW>>1); i++){\
256
                        int j;\
257
                        int Y1=1<<18;\
258
                        int Y2=1<<18;\
259
                        int U=1<<18;\
260
                        int V=1<<18;\
261
                        type *r, *b, *g;\
262
                        const int i2= 2*i;\
263
                        \
264
                        for(j=0; j<lumFilterSize; j++)\
265
                        {\
266
                                Y1 += lumSrc[j][i2] * lumFilter[j];\
267
                                Y2 += lumSrc[j][i2+1] * lumFilter[j];\
268
                        }\
269
                        for(j=0; j<chrFilterSize; j++)\
270
                        {\
271
                                U += chrSrc[j][i] * chrFilter[j];\
272
                                V += chrSrc[j][i+2048] * chrFilter[j];\
273
                        }\
274
                        Y1>>=19;\
275
                        Y2>>=19;\
276
                        U >>=19;\
277
                        V >>=19;\
278
                        if((Y1|Y2|U|V)&256)\
279
                        {\
280
                                if(Y1>255)   Y1=255;\
281
                                else if(Y1<0)Y1=0;\
282
                                if(Y2>255)   Y2=255;\
283
                                else if(Y2<0)Y2=0;\
284
                                if(U>255)    U=255;\
285
                                else if(U<0) U=0;\
286
                                if(V>255)    V=255;\
287
                                else if(V<0) V=0;\
288
                        }
289
                        
290
#define YSCALE_YUV_2_RGBX_C(type) \
291
                        YSCALE_YUV_2_PACKEDX_C(type)\
292
                        r = c->table_rV[V];\
293
                        g = c->table_gU[U] + c->table_gV[V];\
294
                        b = c->table_bU[U];\
295

    
296
#define YSCALE_YUV_2_PACKED2_C \
297
                for(i=0; i<(dstW>>1); i++){\
298
                        const int i2= 2*i;\
299
                        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;\
300
                        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;\
301
                        int U= (uvbuf0[i     ]*uvalpha1+uvbuf1[i     ]*uvalpha)>>19;\
302
                        int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\
303

    
304
#define YSCALE_YUV_2_RGB2_C(type) \
305
                        YSCALE_YUV_2_PACKED2_C\
306
                        type *r, *b, *g;\
307
                        r = c->table_rV[V];\
308
                        g = c->table_gU[U] + c->table_gV[V];\
309
                        b = c->table_bU[U];\
310

    
311
#define YSCALE_YUV_2_PACKED1_C \
312
                for(i=0; i<(dstW>>1); i++){\
313
                        const int i2= 2*i;\
314
                        int Y1= buf0[i2  ]>>7;\
315
                        int Y2= buf0[i2+1]>>7;\
316
                        int U= (uvbuf1[i     ])>>7;\
317
                        int V= (uvbuf1[i+2048])>>7;\
318

    
319
#define YSCALE_YUV_2_RGB1_C(type) \
320
                        YSCALE_YUV_2_PACKED1_C\
321
                        type *r, *b, *g;\
322
                        r = c->table_rV[V];\
323
                        g = c->table_gU[U] + c->table_gV[V];\
324
                        b = c->table_bU[U];\
325

    
326
#define YSCALE_YUV_2_PACKED1B_C \
327
                for(i=0; i<(dstW>>1); i++){\
328
                        const int i2= 2*i;\
329
                        int Y1= buf0[i2  ]>>7;\
330
                        int Y2= buf0[i2+1]>>7;\
331
                        int U= (uvbuf0[i     ] + uvbuf1[i     ])>>8;\
332
                        int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\
333

    
334
#define YSCALE_YUV_2_RGB1B_C(type) \
335
                        YSCALE_YUV_2_PACKED1B_C\
336
                        type *r, *b, *g;\
337
                        r = c->table_rV[V];\
338
                        g = c->table_gU[U] + c->table_gV[V];\
339
                        b = c->table_bU[U];\
340

    
341
#define YSCALE_YUV_2_ANYRGB_C(func, func2)\
342
        switch(c->dstFormat)\
343
        {\
344
        case IMGFMT_BGR32:\
345
        case IMGFMT_RGB32:\
346
                func(uint32_t)\
347
                        ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
348
                        ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
349
                }                \
350
                break;\
351
        case IMGFMT_RGB24:\
352
                func(uint8_t)\
353
                        ((uint8_t*)dest)[0]= r[Y1];\
354
                        ((uint8_t*)dest)[1]= g[Y1];\
355
                        ((uint8_t*)dest)[2]= b[Y1];\
356
                        ((uint8_t*)dest)[3]= r[Y2];\
357
                        ((uint8_t*)dest)[4]= g[Y2];\
358
                        ((uint8_t*)dest)[5]= b[Y2];\
359
                        dest+=6;\
360
                }\
361
                break;\
362
        case IMGFMT_BGR24:\
363
                func(uint8_t)\
364
                        ((uint8_t*)dest)[0]= b[Y1];\
365
                        ((uint8_t*)dest)[1]= g[Y1];\
366
                        ((uint8_t*)dest)[2]= r[Y1];\
367
                        ((uint8_t*)dest)[3]= b[Y2];\
368
                        ((uint8_t*)dest)[4]= g[Y2];\
369
                        ((uint8_t*)dest)[5]= r[Y2];\
370
                        dest+=6;\
371
                }\
372
                break;\
373
        case IMGFMT_RGB16:\
374
        case IMGFMT_BGR16:\
375
                {\
376
                        const int dr1= dither_2x2_8[y&1    ][0];\
377
                        const int dg1= dither_2x2_4[y&1    ][0];\
378
                        const int db1= dither_2x2_8[(y&1)^1][0];\
379
                        const int dr2= dither_2x2_8[y&1    ][1];\
380
                        const int dg2= dither_2x2_4[y&1    ][1];\
381
                        const int db2= dither_2x2_8[(y&1)^1][1];\
382
                        func(uint16_t)\
383
                                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
384
                                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
385
                        }\
386
                }\
387
                break;\
388
        case IMGFMT_RGB15:\
389
        case IMGFMT_BGR15:\
390
                {\
391
                        const int dr1= dither_2x2_8[y&1    ][0];\
392
                        const int dg1= dither_2x2_8[y&1    ][1];\
393
                        const int db1= dither_2x2_8[(y&1)^1][0];\
394
                        const int dr2= dither_2x2_8[y&1    ][1];\
395
                        const int dg2= dither_2x2_8[y&1    ][0];\
396
                        const int db2= dither_2x2_8[(y&1)^1][1];\
397
                        func(uint16_t)\
398
                                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
399
                                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
400
                        }\
401
                }\
402
                break;\
403
        case IMGFMT_RGB8:\
404
        case IMGFMT_BGR8:\
405
                {\
406
                        const uint8_t * const d64= dither_8x8_73[y&7];\
407
                        const uint8_t * const d32= dither_8x8_32[y&7];\
408
                        func(uint8_t)\
409
                                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
410
                                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
411
                        }\
412
                }\
413
                break;\
414
        case IMGFMT_RGB4:\
415
        case IMGFMT_BGR4:\
416
                {\
417
                        const uint8_t * const d64= dither_8x8_73 [y&7];\
418
                        const uint8_t * const d128=dither_8x8_220[y&7];\
419
                        func(uint8_t)\
420
                                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
421
                                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
422
                        }\
423
                }\
424
                break;\
425
        case IMGFMT_RG4B:\
426
        case IMGFMT_BG4B:\
427
                {\
428
                        const uint8_t * const d64= dither_8x8_73 [y&7];\
429
                        const uint8_t * const d128=dither_8x8_220[y&7];\
430
                        func(uint8_t)\
431
                                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
432
                                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
433
                        }\
434
                }\
435
                break;\
436
        case IMGFMT_RGB1:\
437
        case IMGFMT_BGR1:\
438
                {\
439
                        const uint8_t * const d128=dither_8x8_220[y&7];\
440
                        uint8_t *g= c->table_gU[128] + c->table_gV[128];\
441
                        for(i=0; i<dstW-7; i+=8){\
442
                                int acc;\
443
                                acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
444
                                acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
445
                                acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
446
                                acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
447
                                acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
448
                                acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
449
                                acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
450
                                acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
451
                                ((uint8_t*)dest)[0]= acc;\
452
                                dest++;\
453
                        }\
454
\
455
/*\
456
((uint8_t*)dest)-= dstW>>4;\
457
{\
458
                        int acc=0;\
459
                        int left=0;\
460
                        static int top[1024];\
461
                        static int last_new[1024][1024];\
462
                        static int last_in3[1024][1024];\
463
                        static int drift[1024][1024];\
464
                        int topLeft=0;\
465
                        int shift=0;\
466
                        int count=0;\
467
                        const uint8_t * const d128=dither_8x8_220[y&7];\
468
                        int error_new=0;\
469
                        int error_in3=0;\
470
                        int f=0;\
471
                        \
472
                        for(i=dstW>>1; i<dstW; i++){\
473
                                int in= ((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19);\
474
                                int in2 = (76309 * (in - 16) + 32768) >> 16;\
475
                                int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\
476
                                int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\
477
                                        + (last_new[y][i] - in3)*f/256;\
478
                                int new= old> 128 ? 255 : 0;\
479
\
480
                                error_new+= ABS(last_new[y][i] - new);\
481
                                error_in3+= ABS(last_in3[y][i] - in3);\
482
                                f= error_new - error_in3*4;\
483
                                if(f<0) f=0;\
484
                                if(f>256) f=256;\
485
\
486
                                topLeft= top[i];\
487
                                left= top[i]= old - new;\
488
                                last_new[y][i]= new;\
489
                                last_in3[y][i]= in3;\
490
\
491
                                acc+= acc + (new&1);\
492
                                if((i&7)==6){\
493
                                        ((uint8_t*)dest)[0]= acc;\
494
                                        ((uint8_t*)dest)++;\
495
                                }\
496
                        }\
497
}\
498
*/\
499
                }\
500
                break;\
501
        case IMGFMT_YUY2:\
502
                func2\
503
                        ((uint8_t*)dest)[2*i2+0]= Y1;\
504
                        ((uint8_t*)dest)[2*i2+1]= U;\
505
                        ((uint8_t*)dest)[2*i2+2]= Y2;\
506
                        ((uint8_t*)dest)[2*i2+3]= V;\
507
                }                \
508
                break;\
509
        case IMGFMT_UYVY:\
510
                func2\
511
                        ((uint8_t*)dest)[2*i2+0]= U;\
512
                        ((uint8_t*)dest)[2*i2+1]= Y1;\
513
                        ((uint8_t*)dest)[2*i2+2]= V;\
514
                        ((uint8_t*)dest)[2*i2+3]= Y2;\
515
                }                \
516
                break;\
517
        }\
518

    
519

    
520
static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
521
                                    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
522
                                    uint8_t *dest, int dstW, int y)
523
{
524
        int i;
525
        switch(c->dstFormat)
526
        {
527
        case IMGFMT_RGB32:
528
        case IMGFMT_BGR32:
529
                YSCALE_YUV_2_RGBX_C(uint32_t)
530
                        ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];
531
                        ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];
532
                }
533
                break;
534
        case IMGFMT_RGB24:
535
                YSCALE_YUV_2_RGBX_C(uint8_t)
536
                        ((uint8_t*)dest)[0]= r[Y1];
537
                        ((uint8_t*)dest)[1]= g[Y1];
538
                        ((uint8_t*)dest)[2]= b[Y1];
539
                        ((uint8_t*)dest)[3]= r[Y2];
540
                        ((uint8_t*)dest)[4]= g[Y2];
541
                        ((uint8_t*)dest)[5]= b[Y2];
542
                        dest+=6;
543
                }
544
                break;
545
        case IMGFMT_BGR24:
546
                YSCALE_YUV_2_RGBX_C(uint8_t)
547
                        ((uint8_t*)dest)[0]= b[Y1];
548
                        ((uint8_t*)dest)[1]= g[Y1];
549
                        ((uint8_t*)dest)[2]= r[Y1];
550
                        ((uint8_t*)dest)[3]= b[Y2];
551
                        ((uint8_t*)dest)[4]= g[Y2];
552
                        ((uint8_t*)dest)[5]= r[Y2];
553
                        dest+=6;
554
                }
555
                break;
556
        case IMGFMT_RGB16:
557
        case IMGFMT_BGR16:
558
                {
559
                        const int dr1= dither_2x2_8[y&1    ][0];
560
                        const int dg1= dither_2x2_4[y&1    ][0];
561
                        const int db1= dither_2x2_8[(y&1)^1][0];
562
                        const int dr2= dither_2x2_8[y&1    ][1];
563
                        const int dg2= dither_2x2_4[y&1    ][1];
564
                        const int db2= dither_2x2_8[(y&1)^1][1];
565
                        YSCALE_YUV_2_RGBX_C(uint16_t)
566
                                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
567
                                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
568
                        }
569
                }
570
                break;
571
        case IMGFMT_RGB15:
572
        case IMGFMT_BGR15:
573
                {
574
                        const int dr1= dither_2x2_8[y&1    ][0];
575
                        const int dg1= dither_2x2_8[y&1    ][1];
576
                        const int db1= dither_2x2_8[(y&1)^1][0];
577
                        const int dr2= dither_2x2_8[y&1    ][1];
578
                        const int dg2= dither_2x2_8[y&1    ][0];
579
                        const int db2= dither_2x2_8[(y&1)^1][1];
580
                        YSCALE_YUV_2_RGBX_C(uint16_t)
581
                                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
582
                                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
583
                        }
584
                }
585
                break;
586
        case IMGFMT_RGB8:
587
        case IMGFMT_BGR8:
588
                {
589
                        const uint8_t * const d64= dither_8x8_73[y&7];
590
                        const uint8_t * const d32= dither_8x8_32[y&7];
591
                        YSCALE_YUV_2_RGBX_C(uint8_t)
592
                                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];
593
                                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];
594
                        }
595
                }
596
                break;
597
        case IMGFMT_RGB4:
598
        case IMGFMT_BGR4:
599
                {
600
                        const uint8_t * const d64= dither_8x8_73 [y&7];
601
                        const uint8_t * const d128=dither_8x8_220[y&7];
602
                        YSCALE_YUV_2_RGBX_C(uint8_t)
603
                                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]
604
                                                  +((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);
605
                        }
606
                }
607
                break;
608
        case IMGFMT_RG4B:
609
        case IMGFMT_BG4B:
610
                {
611
                        const uint8_t * const d64= dither_8x8_73 [y&7];
612
                        const uint8_t * const d128=dither_8x8_220[y&7];
613
                        YSCALE_YUV_2_RGBX_C(uint8_t)
614
                                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];
615
                                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];
616
                        }
617
                }
618
                break;
619
        case IMGFMT_RGB1:
620
        case IMGFMT_BGR1:
621
                {
622
                        const uint8_t * const d128=dither_8x8_220[y&7];
623
                        uint8_t *g= c->table_gU[128] + c->table_gV[128];
624
                        int acc=0;
625
                        for(i=0; i<dstW-1; i+=2){
626
                                int j;
627
                                int Y1=1<<18;
628
                                int Y2=1<<18;
629

    
630
                                for(j=0; j<lumFilterSize; j++)
631
                                {
632
                                        Y1 += lumSrc[j][i] * lumFilter[j];
633
                                        Y2 += lumSrc[j][i+1] * lumFilter[j];
634
                                }
635
                                Y1>>=19;
636
                                Y2>>=19;
637
                                if((Y1|Y2)&256)
638
                                {
639
                                        if(Y1>255)   Y1=255;
640
                                        else if(Y1<0)Y1=0;
641
                                        if(Y2>255)   Y2=255;
642
                                        else if(Y2<0)Y2=0;
643
                                }
644
                                acc+= acc + g[Y1+d128[(i+0)&7]];
645
                                acc+= acc + g[Y2+d128[(i+1)&7]];
646
                                if((i&7)==6){
647
                                        ((uint8_t*)dest)[0]= acc;
648
                                        dest++;
649
                                }
650
                        }
651
                }
652
                break;
653
        case IMGFMT_YUY2:
654
                YSCALE_YUV_2_PACKEDX_C(void)
655
                        ((uint8_t*)dest)[2*i2+0]= Y1;
656
                        ((uint8_t*)dest)[2*i2+1]= U;
657
                        ((uint8_t*)dest)[2*i2+2]= Y2;
658
                        ((uint8_t*)dest)[2*i2+3]= V;
659
                }
660
                break;
661
        case IMGFMT_UYVY:
662
                YSCALE_YUV_2_PACKEDX_C(void)
663
                        ((uint8_t*)dest)[2*i2+0]= U;
664
                        ((uint8_t*)dest)[2*i2+1]= Y1;
665
                        ((uint8_t*)dest)[2*i2+2]= V;
666
                        ((uint8_t*)dest)[2*i2+3]= Y2;
667
                }
668
                break;
669
        }
670
}
671

    
672

    
673
//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
674
//Plain C versions
675
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT)
676
#define COMPILE_C
677
#endif
678

    
679
#ifdef ARCH_POWERPC
680
#ifdef HAVE_ALTIVEC
681
#define COMPILE_ALTIVEC
682
#endif //HAVE_ALTIVEC
683
#endif //ARCH_POWERPC
684

    
685
#if defined(ARCH_X86) || defined(ARCH_X86_64)
686

    
687
#if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
688
#define COMPILE_MMX
689
#endif
690

    
691
#if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)
692
#define COMPILE_MMX2
693
#endif
694

    
695
#if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
696
#define COMPILE_3DNOW
697
#endif
698
#endif //ARCH_X86 || ARCH_X86_64
699

    
700
#undef HAVE_MMX
701
#undef HAVE_MMX2
702
#undef HAVE_3DNOW
703

    
704
#ifdef COMPILE_C
705
#undef HAVE_MMX
706
#undef HAVE_MMX2
707
#undef HAVE_3DNOW
708
#undef HAVE_ALTIVEC
709
#define RENAME(a) a ## _C
710
#include "swscale_template.c"
711
#endif
712

    
713
#ifdef ARCH_POWERPC
714
#ifdef COMPILE_ALTIVEC
715
#undef RENAME
716
#define HAVE_ALTIVEC
717
#define RENAME(a) a ## _altivec
718
#include "swscale_template.c"
719
#endif
720
#endif //ARCH_POWERPC
721

    
722
#if defined(ARCH_X86) || defined(ARCH_X86_64)
723

    
724
//X86 versions
725
/*
726
#undef RENAME
727
#undef HAVE_MMX
728
#undef HAVE_MMX2
729
#undef HAVE_3DNOW
730
#define ARCH_X86
731
#define RENAME(a) a ## _X86
732
#include "swscale_template.c"
733
*/
734
//MMX versions
735
#ifdef COMPILE_MMX
736
#undef RENAME
737
#define HAVE_MMX
738
#undef HAVE_MMX2
739
#undef HAVE_3DNOW
740
#define RENAME(a) a ## _MMX
741
#include "swscale_template.c"
742
#endif
743

    
744
//MMX2 versions
745
#ifdef COMPILE_MMX2
746
#undef RENAME
747
#define HAVE_MMX
748
#define HAVE_MMX2
749
#undef HAVE_3DNOW
750
#define RENAME(a) a ## _MMX2
751
#include "swscale_template.c"
752
#endif
753

    
754
//3DNOW versions
755
#ifdef COMPILE_3DNOW
756
#undef RENAME
757
#define HAVE_MMX
758
#undef HAVE_MMX2
759
#define HAVE_3DNOW
760
#define RENAME(a) a ## _3DNow
761
#include "swscale_template.c"
762
#endif
763

    
764
#endif //ARCH_X86 || ARCH_X86_64
765

    
766
// minor note: the HAVE_xyz is messed up after that line so don't use it
767

    
768
static double getSplineCoeff(double a, double b, double c, double d, double dist)
769
{
770
//        printf("%f %f %f %f %f\n", a,b,c,d,dist);
771
        if(dist<=1.0)         return ((d*dist + c)*dist + b)*dist +a;
772
        else                return getSplineCoeff(        0.0, 
773
                                                 b+ 2.0*c + 3.0*d,
774
                                                        c + 3.0*d,
775
                                                -b- 3.0*c - 6.0*d,
776
                                                dist-1.0);
777
}
778

    
779
static inline void initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
780
                              int srcW, int dstW, int filterAlign, int one, int flags,
781
                              SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
782
{
783
        int i;
784
        int filterSize;
785
        int filter2Size;
786
        int minFilterSize;
787
        double *filter=NULL;
788
        double *filter2=NULL;
789
#if defined(ARCH_X86) || defined(ARCH_X86_64)
790
        if(flags & SWS_CPU_CAPS_MMX)
791
                asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
792
#endif
793

    
794
        // Note the +1 is for the MMXscaler which reads over the end
795
        *filterPos = (int16_t*)memalign(8, (dstW+1)*sizeof(int16_t));
796

    
797
        if(ABS(xInc - 0x10000) <10) // unscaled
798
        {
799
                int i;
800
                filterSize= 1;
801
                filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
802
                for(i=0; i<dstW*filterSize; i++) filter[i]=0;
803

    
804
                for(i=0; i<dstW; i++)
805
                {
806
                        filter[i*filterSize]=1;
807
                        (*filterPos)[i]=i;
808
                }
809

    
810
        }
811
        else if(flags&SWS_POINT) // lame looking point sampling mode
812
        {
813
                int i;
814
                int xDstInSrc;
815
                filterSize= 1;
816
                filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
817
                
818
                xDstInSrc= xInc/2 - 0x8000;
819
                for(i=0; i<dstW; i++)
820
                {
821
                        int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
822

    
823
                        (*filterPos)[i]= xx;
824
                        filter[i]= 1.0;
825
                        xDstInSrc+= xInc;
826
                }
827
        }
828
        else if((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
829
        {
830
                int i;
831
                int xDstInSrc;
832
                if     (flags&SWS_BICUBIC) filterSize= 4;
833
                else if(flags&SWS_X      ) filterSize= 4;
834
                else                           filterSize= 2; // SWS_BILINEAR / SWS_AREA 
835
                filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
836

    
837
                xDstInSrc= xInc/2 - 0x8000;
838
                for(i=0; i<dstW; i++)
839
                {
840
                        int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
841
                        int j;
842

    
843
                        (*filterPos)[i]= xx;
844
                                //Bilinear upscale / linear interpolate / Area averaging
845
                                for(j=0; j<filterSize; j++)
846
                                {
847
                                        double d= ABS((xx<<16) - xDstInSrc)/(double)(1<<16);
848
                                        double coeff= 1.0 - d;
849
                                        if(coeff<0) coeff=0;
850
                                        filter[i*filterSize + j]= coeff;
851
                                        xx++;
852
                                }
853
                        xDstInSrc+= xInc;
854
                }
855
        }
856
        else
857
        {
858
                double xDstInSrc;
859
                double sizeFactor, filterSizeInSrc;
860
                const double xInc1= (double)xInc / (double)(1<<16);
861

    
862
                if     (flags&SWS_BICUBIC)        sizeFactor= 4.0;
863
                else if(flags&SWS_X)                sizeFactor= 8.0;
864
                else if(flags&SWS_AREA)                sizeFactor= 1.0; //downscale only, for upscale it is bilinear
865
                else if(flags&SWS_GAUSS)        sizeFactor= 8.0;   // infinite ;)
866
                else if(flags&SWS_LANCZOS)        sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;
867
                else if(flags&SWS_SINC)                sizeFactor= 20.0; // infinite ;)
868
                else if(flags&SWS_SPLINE)        sizeFactor= 20.0;  // infinite ;)
869
                else if(flags&SWS_BILINEAR)        sizeFactor= 2.0;
870
                else {
871
                        sizeFactor= 0.0; //GCC warning killer
872
                        ASSERT(0)
873
                }
874
                
875
                if(xInc1 <= 1.0)        filterSizeInSrc= sizeFactor; // upscale
876
                else                        filterSizeInSrc= sizeFactor*srcW / (double)dstW;
877

    
878
                filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
879
                if(filterSize > srcW-2) filterSize=srcW-2;
880

    
881
                filter= (double*)memalign(16, dstW*sizeof(double)*filterSize);
882

    
883
                xDstInSrc= xInc1 / 2.0 - 0.5;
884
                for(i=0; i<dstW; i++)
885
                {
886
                        int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
887
                        int j;
888
                        (*filterPos)[i]= xx;
889
                        for(j=0; j<filterSize; j++)
890
                        {
891
                                double d= ABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
892
                                double coeff;
893
                                if(flags & SWS_BICUBIC)
894
                                {
895
                                        double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
896
                                        double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;
897

    
898
                                        if(d<1.0) 
899
                                                coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
900
                                        else if(d<2.0)
901
                                                coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
902
                                        else
903
                                                coeff=0.0;
904
                                }
905
/*                                else if(flags & SWS_X)
906
                                {
907
                                        double p= param ? param*0.01 : 0.3;
908
                                        coeff = d ? sin(d*PI)/(d*PI) : 1.0;
909
                                        coeff*= pow(2.0, - p*d*d);
910
                                }*/
911
                                else if(flags & SWS_X)
912
                                {
913
                                        double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
914
                                        
915
                                        if(d<1.0)
916
                                                coeff = cos(d*PI);
917
                                        else
918
                                                coeff=-1.0;
919
                                        if(coeff<0.0)         coeff= -pow(-coeff, A);
920
                                        else                coeff=  pow( coeff, A);
921
                                        coeff= coeff*0.5 + 0.5;
922
                                }
923
                                else if(flags & SWS_AREA)
924
                                {
925
                                        double srcPixelSize= 1.0/xInc1;
926
                                        if(d + srcPixelSize/2 < 0.5) coeff= 1.0;
927
                                        else if(d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;
928
                                        else coeff=0.0;
929
                                }
930
                                else if(flags & SWS_GAUSS)
931
                                {
932
                                        double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
933
                                        coeff = pow(2.0, - p*d*d);
934
                                }
935
                                else if(flags & SWS_SINC)
936
                                {
937
                                        coeff = d ? sin(d*PI)/(d*PI) : 1.0;
938
                                }
939
                                else if(flags & SWS_LANCZOS)
940
                                {
941
                                        double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; 
942
                                        coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
943
                                        if(d>p) coeff=0;
944
                                }
945
                                else if(flags & SWS_BILINEAR)
946
                                {
947
                                        coeff= 1.0 - d;
948
                                        if(coeff<0) coeff=0;
949
                                }
950
                                else if(flags & SWS_SPLINE)
951
                                {
952
                                        double p=-2.196152422706632;
953
                                        coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);
954
                                }
955
                                else {
956
                                        coeff= 0.0; //GCC warning killer
957
                                        ASSERT(0)
958
                                }
959

    
960
                                filter[i*filterSize + j]= coeff;
961
                                xx++;
962
                        }
963
                        xDstInSrc+= xInc1;
964
                }
965
        }
966

    
967
        /* apply src & dst Filter to filter -> filter2
968
           free(filter);
969
        */
970
        ASSERT(filterSize>0)
971
        filter2Size= filterSize;
972
        if(srcFilter) filter2Size+= srcFilter->length - 1;
973
        if(dstFilter) filter2Size+= dstFilter->length - 1;
974
        ASSERT(filter2Size>0)
975
        filter2= (double*)memalign(8, filter2Size*dstW*sizeof(double));
976

    
977
        for(i=0; i<dstW; i++)
978
        {
979
                int j;
980
                SwsVector scaleFilter;
981
                SwsVector *outVec;
982

    
983
                scaleFilter.coeff= filter + i*filterSize;
984
                scaleFilter.length= filterSize;
985

    
986
                if(srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);
987
                else              outVec= &scaleFilter;
988

    
989
                ASSERT(outVec->length == filter2Size)
990
                //FIXME dstFilter
991

    
992
                for(j=0; j<outVec->length; j++)
993
                {
994
                        filter2[i*filter2Size + j]= outVec->coeff[j];
995
                }
996

    
997
                (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
998

    
999
                if(outVec != &scaleFilter) sws_freeVec(outVec);
1000
        }
1001
        free(filter); filter=NULL;
1002

    
1003
        /* try to reduce the filter-size (step1 find size and shift left) */
1004
        // Assume its near normalized (*0.5 or *2.0 is ok but * 0.001 is not)
1005
        minFilterSize= 0;
1006
        for(i=dstW-1; i>=0; i--)
1007
        {
1008
                int min= filter2Size;
1009
                int j;
1010
                double cutOff=0.0;
1011

    
1012
                /* get rid off near zero elements on the left by shifting left */
1013
                for(j=0; j<filter2Size; j++)
1014
                {
1015
                        int k;
1016
                        cutOff += ABS(filter2[i*filter2Size]);
1017

    
1018
                        if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;
1019

    
1020
                        /* preserve Monotonicity because the core can't handle the filter otherwise */
1021
                        if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1022

    
1023
                        // Move filter coeffs left
1024
                        for(k=1; k<filter2Size; k++)
1025
                                filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1026
                        filter2[i*filter2Size + k - 1]= 0.0;
1027
                        (*filterPos)[i]++;
1028
                }
1029

    
1030
                cutOff=0.0;
1031
                /* count near zeros on the right */
1032
                for(j=filter2Size-1; j>0; j--)
1033
                {
1034
                        cutOff += ABS(filter2[i*filter2Size + j]);
1035

    
1036
                        if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;
1037
                        min--;
1038
                }
1039

    
1040
                if(min>minFilterSize) minFilterSize= min;
1041
        }
1042

    
1043
        if (flags & SWS_CPU_CAPS_ALTIVEC) {
1044
          // we can handle the special case 4,
1045
          // so we don't want to go to the full 8
1046
          if (minFilterSize < 5)
1047
            filterAlign = 4;
1048

    
1049
          // we really don't want to waste our time
1050
          // doing useless computation, so fall-back on
1051
          // the scalar C code for very small filter.
1052
          // vectorizing is worth it only if you have
1053
          // decent-sized vector.
1054
          if (minFilterSize < 3)
1055
            filterAlign = 1;
1056
        }
1057

    
1058
        ASSERT(minFilterSize > 0)
1059
        filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1060
        ASSERT(filterSize > 0)
1061
        filter= (double*)memalign(8, filterSize*dstW*sizeof(double));
1062
        *outFilterSize= filterSize;
1063

    
1064
        if(flags&SWS_PRINT_INFO)
1065
                MSG_INFO("SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1066
        /* try to reduce the filter-size (step2 reduce it) */
1067
        for(i=0; i<dstW; i++)
1068
        {
1069
                int j;
1070

    
1071
                for(j=0; j<filterSize; j++)
1072
                {
1073
                        if(j>=filter2Size) filter[i*filterSize + j]= 0.0;
1074
                        else                   filter[i*filterSize + j]= filter2[i*filter2Size + j];
1075
                }
1076
        }
1077
        free(filter2); filter2=NULL;
1078
        
1079

    
1080
        //FIXME try to align filterpos if possible
1081

    
1082
        //fix borders
1083
        for(i=0; i<dstW; i++)
1084
        {
1085
                int j;
1086
                if((*filterPos)[i] < 0)
1087
                {
1088
                        // Move filter coeffs left to compensate for filterPos
1089
                        for(j=1; j<filterSize; j++)
1090
                        {
1091
                                int left= MAX(j + (*filterPos)[i], 0);
1092
                                filter[i*filterSize + left] += filter[i*filterSize + j];
1093
                                filter[i*filterSize + j]=0;
1094
                        }
1095
                        (*filterPos)[i]= 0;
1096
                }
1097

    
1098
                if((*filterPos)[i] + filterSize > srcW)
1099
                {
1100
                        int shift= (*filterPos)[i] + filterSize - srcW;
1101
                        // Move filter coeffs right to compensate for filterPos
1102
                        for(j=filterSize-2; j>=0; j--)
1103
                        {
1104
                                int right= MIN(j + shift, filterSize-1);
1105
                                filter[i*filterSize +right] += filter[i*filterSize +j];
1106
                                filter[i*filterSize +j]=0;
1107
                        }
1108
                        (*filterPos)[i]= srcW - filterSize;
1109
                }
1110
        }
1111

    
1112
        // Note the +1 is for the MMXscaler which reads over the end
1113
        *outFilter= (int16_t*)memalign(8, *outFilterSize*(dstW+1)*sizeof(int16_t));
1114
        memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t));
1115

    
1116
        /* Normalize & Store in outFilter */
1117
        for(i=0; i<dstW; i++)
1118
        {
1119
                int j;
1120
                double error=0;
1121
                double sum=0;
1122
                double scale= one;
1123

    
1124
                for(j=0; j<filterSize; j++)
1125
                {
1126
                        sum+= filter[i*filterSize + j];
1127
                }
1128
                scale/= sum;
1129
                for(j=0; j<*outFilterSize; j++)
1130
                {
1131
                        double v= filter[i*filterSize + j]*scale + error;
1132
                        int intV= floor(v + 0.5);
1133
                        (*outFilter)[i*(*outFilterSize) + j]= intV;
1134
                        error = v - intV;
1135
                }
1136
        }
1137
        
1138
        (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1139
        for(i=0; i<*outFilterSize; i++)
1140
        {
1141
                int j= dstW*(*outFilterSize);
1142
                (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1143
        }
1144

    
1145
        free(filter);
1146
}
1147

    
1148
#if defined(ARCH_X86) || defined(ARCH_X86_64)
1149
static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1150
{
1151
        uint8_t *fragmentA;
1152
        long imm8OfPShufW1A;
1153
        long imm8OfPShufW2A;
1154
        long fragmentLengthA;
1155
        uint8_t *fragmentB;
1156
        long imm8OfPShufW1B;
1157
        long imm8OfPShufW2B;
1158
        long fragmentLengthB;
1159
        int fragmentPos;
1160

    
1161
        int xpos, i;
1162

    
1163
        // create an optimized horizontal scaling routine
1164

    
1165
        //code fragment
1166

    
1167
        asm volatile(
1168
                "jmp 9f                                \n\t"
1169
        // Begin
1170
                "0:                                \n\t"
1171
                "movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 
1172
                "movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 
1173
                "movd 1(%%"REG_c", %%"REG_S"), %%mm1\n\t"
1174
                "punpcklbw %%mm7, %%mm1                \n\t"
1175
                "punpcklbw %%mm7, %%mm0                \n\t"
1176
                "pshufw $0xFF, %%mm1, %%mm1        \n\t"
1177
                "1:                                \n\t"
1178
                "pshufw $0xFF, %%mm0, %%mm0        \n\t"
1179
                "2:                                \n\t"
1180
                "psubw %%mm1, %%mm0                \n\t"
1181
                "movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"
1182
                "pmullw %%mm3, %%mm0                \n\t"
1183
                "psllw $7, %%mm1                \n\t"
1184
                "paddw %%mm1, %%mm0                \n\t"
1185

    
1186
                "movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"
1187

    
1188
                "add $8, %%"REG_a"                \n\t"
1189
        // End
1190
                "9:                                \n\t"
1191
//                "int $3\n\t"
1192
                "lea 0b, %0                        \n\t"
1193
                "lea 1b, %1                        \n\t"
1194
                "lea 2b, %2                        \n\t"
1195
                "dec %1                                \n\t"
1196
                "dec %2                                \n\t"
1197
                "sub %0, %1                        \n\t"
1198
                "sub %0, %2                        \n\t"
1199
                "lea 9b, %3                        \n\t"
1200
                "sub %0, %3                        \n\t"
1201

    
1202

    
1203
                :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1204
                "=r" (fragmentLengthA)
1205
        );
1206

    
1207
        asm volatile(
1208
                "jmp 9f                                \n\t"
1209
        // Begin
1210
                "0:                                \n\t"
1211
                "movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 
1212
                "movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 
1213
                "punpcklbw %%mm7, %%mm0                \n\t"
1214
                "pshufw $0xFF, %%mm0, %%mm1        \n\t"
1215
                "1:                                \n\t"
1216
                "pshufw $0xFF, %%mm0, %%mm0        \n\t"
1217
                "2:                                \n\t"
1218
                "psubw %%mm1, %%mm0                \n\t"
1219
                "movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"
1220
                "pmullw %%mm3, %%mm0                \n\t"
1221
                "psllw $7, %%mm1                \n\t"
1222
                "paddw %%mm1, %%mm0                \n\t"
1223

    
1224
                "movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"
1225

    
1226
                "add $8, %%"REG_a"                \n\t"
1227
        // End
1228
                "9:                                \n\t"
1229
//                "int $3\n\t"
1230
                "lea 0b, %0                        \n\t"
1231
                "lea 1b, %1                        \n\t"
1232
                "lea 2b, %2                        \n\t"
1233
                "dec %1                                \n\t"
1234
                "dec %2                                \n\t"
1235
                "sub %0, %1                        \n\t"
1236
                "sub %0, %2                        \n\t"
1237
                "lea 9b, %3                        \n\t"
1238
                "sub %0, %3                        \n\t"
1239

    
1240

    
1241
                :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1242
                "=r" (fragmentLengthB)
1243
        );
1244

    
1245
        xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1246
        fragmentPos=0;
1247
        
1248
        for(i=0; i<dstW/numSplits; i++)
1249
        {
1250
                int xx=xpos>>16;
1251

    
1252
                if((i&3) == 0)
1253
                {
1254
                        int a=0;
1255
                        int b=((xpos+xInc)>>16) - xx;
1256
                        int c=((xpos+xInc*2)>>16) - xx;
1257
                        int d=((xpos+xInc*3)>>16) - xx;
1258

    
1259
                        filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
1260
                        filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
1261
                        filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1262
                        filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1263
                        filterPos[i/2]= xx;
1264

    
1265
                        if(d+1<4)
1266
                        {
1267
                                int maxShift= 3-(d+1);
1268
                                int shift=0;
1269

    
1270
                                memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1271

    
1272
                                funnyCode[fragmentPos + imm8OfPShufW1B]=
1273
                                        (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1274
                                funnyCode[fragmentPos + imm8OfPShufW2B]=
1275
                                        a | (b<<2) | (c<<4) | (d<<6);
1276

    
1277
                                if(i+3>=dstW) shift=maxShift; //avoid overread
1278
                                else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1279

    
1280
                                if(shift && i>=shift)
1281
                                {
1282
                                        funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1283
                                        funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1284
                                        filterPos[i/2]-=shift;
1285
                                }
1286

    
1287
                                fragmentPos+= fragmentLengthB;
1288
                        }
1289
                        else
1290
                        {
1291
                                int maxShift= 3-d;
1292
                                int shift=0;
1293

    
1294
                                memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
1295

    
1296
                                funnyCode[fragmentPos + imm8OfPShufW1A]=
1297
                                funnyCode[fragmentPos + imm8OfPShufW2A]=
1298
                                        a | (b<<2) | (c<<4) | (d<<6);
1299

    
1300
                                if(i+4>=dstW) shift=maxShift; //avoid overread
1301
                                else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
1302

    
1303
                                if(shift && i>=shift)
1304
                                {
1305
                                        funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1306
                                        funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1307
                                        filterPos[i/2]-=shift;
1308
                                }
1309

    
1310
                                fragmentPos+= fragmentLengthA;
1311
                        }
1312

    
1313
                        funnyCode[fragmentPos]= RET;
1314
                }
1315
                xpos+=xInc;
1316
        }
1317
        filterPos[i/2]= xpos>>16; // needed to jump to the next part
1318
}
1319
#endif // ARCH_X86 || ARCH_X86_64
1320

    
1321
static void globalInit(){
1322
    // generating tables:
1323
    int i;
1324
    for(i=0; i<768; i++){
1325
        int c= MIN(MAX(i-256, 0), 255);
1326
        clip_table[i]=c;
1327
    }
1328
}
1329

    
1330
static SwsFunc getSwsFunc(int flags){
1331
    
1332
#ifdef RUNTIME_CPUDETECT
1333
#if defined(ARCH_X86) || defined(ARCH_X86_64)
1334
        // ordered per speed fasterst first
1335
        if(flags & SWS_CPU_CAPS_MMX2)
1336
                return swScale_MMX2;
1337
        else if(flags & SWS_CPU_CAPS_3DNOW)
1338
                return swScale_3DNow;
1339
        else if(flags & SWS_CPU_CAPS_MMX)
1340
                return swScale_MMX;
1341
        else
1342
                return swScale_C;
1343

    
1344
#else
1345
#ifdef ARCH_POWERPC
1346
        if(flags & SWS_CPU_CAPS_ALTIVEC)
1347
          return swScale_altivec;
1348
        else
1349
          return swScale_C;
1350
#endif
1351
        return swScale_C;
1352
#endif
1353
#else //RUNTIME_CPUDETECT
1354
#ifdef HAVE_MMX2
1355
        return swScale_MMX2;
1356
#elif defined (HAVE_3DNOW)
1357
        return swScale_3DNow;
1358
#elif defined (HAVE_MMX)
1359
        return swScale_MMX;
1360
#elif defined (HAVE_ALTIVEC)
1361
        return swScale_altivec;
1362
#else
1363
        return swScale_C;
1364
#endif
1365
#endif //!RUNTIME_CPUDETECT
1366
}
1367

    
1368
static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1369
             int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1370
        uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1371
        /* Copy Y plane */
1372
        if(dstStride[0]==srcStride[0])
1373
                memcpy(dst, src[0], srcSliceH*dstStride[0]);
1374
        else
1375
        {
1376
                int i;
1377
                uint8_t *srcPtr= src[0];
1378
                uint8_t *dstPtr= dst;
1379
                for(i=0; i<srcSliceH; i++)
1380
                {
1381
                        memcpy(dstPtr, srcPtr, srcStride[0]);
1382
                        srcPtr+= srcStride[0];
1383
                        dstPtr+= dstStride[0];
1384
                }
1385
        }
1386
        dst = dstParam[1] + dstStride[1]*srcSliceY;
1387
        interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] );
1388

    
1389
        return srcSliceH;
1390
}
1391

    
1392
static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1393
             int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1394
        uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1395

    
1396
        yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
1397

    
1398
        return srcSliceH;
1399
}
1400

    
1401
static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1402
             int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1403
        uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1404

    
1405
        yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
1406

    
1407
        return srcSliceH;
1408
}
1409

    
1410
/* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */
1411
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1412
                           int srcSliceH, uint8_t* dst[], int dstStride[]){
1413
        const int srcFormat= c->srcFormat;
1414
        const int dstFormat= c->dstFormat;
1415
        const int srcBpp= ((srcFormat&0xFF) + 7)>>3;
1416
        const int dstBpp= ((dstFormat&0xFF) + 7)>>3;
1417
        const int srcId= (srcFormat&0xFF)>>2; // 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 
1418
        const int dstId= (dstFormat&0xFF)>>2;
1419
        void (*conv)(const uint8_t *src, uint8_t *dst, unsigned src_size)=NULL;
1420

    
1421
        /* BGR -> BGR */
1422
        if(   (isBGR(srcFormat) && isBGR(dstFormat))
1423
           || (isRGB(srcFormat) && isRGB(dstFormat))){
1424
                switch(srcId | (dstId<<4)){
1425
                case 0x34: conv= rgb16to15; break;
1426
                case 0x36: conv= rgb24to15; break;
1427
                case 0x38: conv= rgb32to15; break;
1428
                case 0x43: conv= rgb15to16; break;
1429
                case 0x46: conv= rgb24to16; break;
1430
                case 0x48: conv= rgb32to16; break;
1431
                case 0x63: conv= rgb15to24; break;
1432
                case 0x64: conv= rgb16to24; break;
1433
                case 0x68: conv= rgb32to24; break;
1434
                case 0x83: conv= rgb15to32; break;
1435
                case 0x84: conv= rgb16to32; break;
1436
                case 0x86: conv= rgb24to32; break;
1437
                default: MSG_ERR("swScaler: internal error %s -> %s converter\n", 
1438
                                 vo_format_name(srcFormat), vo_format_name(dstFormat)); break;
1439
                }
1440
        }else if(   (isBGR(srcFormat) && isRGB(dstFormat))
1441
                 || (isRGB(srcFormat) && isBGR(dstFormat))){
1442
                switch(srcId | (dstId<<4)){
1443
                case 0x33: conv= rgb15tobgr15; break;
1444
                case 0x34: conv= rgb16tobgr15; break;
1445
                case 0x36: conv= rgb24tobgr15; break;
1446
                case 0x38: conv= rgb32tobgr15; break;
1447
                case 0x43: conv= rgb15tobgr16; break;
1448
                case 0x44: conv= rgb16tobgr16; break;
1449
                case 0x46: conv= rgb24tobgr16; break;
1450
                case 0x48: conv= rgb32tobgr16; break;
1451
                case 0x63: conv= rgb15tobgr24; break;
1452
                case 0x64: conv= rgb16tobgr24; break;
1453
                case 0x66: conv= rgb24tobgr24; break;
1454
                case 0x68: conv= rgb32tobgr24; break;
1455
                case 0x83: conv= rgb15tobgr32; break;
1456
                case 0x84: conv= rgb16tobgr32; break;
1457
                case 0x86: conv= rgb24tobgr32; break;
1458
                case 0x88: conv= rgb32tobgr32; break;
1459
                default: MSG_ERR("swScaler: internal error %s -> %s converter\n", 
1460
                                 vo_format_name(srcFormat), vo_format_name(dstFormat)); break;
1461
                }
1462
        }else{
1463
                MSG_ERR("swScaler: internal error %s -> %s converter\n", 
1464
                         vo_format_name(srcFormat), vo_format_name(dstFormat));
1465
        }
1466

    
1467
        if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)
1468
                conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1469
        else
1470
        {
1471
                int i;
1472
                uint8_t *srcPtr= src[0];
1473
                uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1474

    
1475
                for(i=0; i<srcSliceH; i++)
1476
                {
1477
                        conv(srcPtr, dstPtr, c->srcW*srcBpp);
1478
                        srcPtr+= srcStride[0];
1479
                        dstPtr+= dstStride[0];
1480
                }
1481
        }     
1482
        return srcSliceH;
1483
}
1484

    
1485
static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1486
             int srcSliceH, uint8_t* dst[], int dstStride[]){
1487

    
1488
        rgb24toyv12(
1489
                src[0], 
1490
                dst[0]+ srcSliceY    *dstStride[0], 
1491
                dst[1]+(srcSliceY>>1)*dstStride[1], 
1492
                dst[2]+(srcSliceY>>1)*dstStride[2],
1493
                c->srcW, srcSliceH, 
1494
                dstStride[0], dstStride[1], srcStride[0]);
1495
        return srcSliceH;
1496
}
1497

    
1498
static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1499
             int srcSliceH, uint8_t* dst[], int dstStride[]){
1500
        int i;
1501

    
1502
        /* copy Y */
1503
        if(srcStride[0]==dstStride[0]) 
1504
                memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1505
        else{
1506
                uint8_t *srcPtr= src[0];
1507
                uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1508

    
1509
                for(i=0; i<srcSliceH; i++)
1510
                {
1511
                        memcpy(dstPtr, srcPtr, c->srcW);
1512
                        srcPtr+= srcStride[0];
1513
                        dstPtr+= dstStride[0];
1514
                }
1515
        }
1516

    
1517
        if(c->dstFormat==IMGFMT_YV12){
1518
                planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
1519
                planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
1520
        }else{
1521
                planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
1522
                planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
1523
        }
1524
        return srcSliceH;
1525
}
1526

    
1527
/**
1528
 * bring pointers in YUV order instead of YVU
1529
 */
1530
static inline void sws_orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]){
1531
        if(format == IMGFMT_YV12 || format == IMGFMT_YVU9
1532
           || format == IMGFMT_444P || format == IMGFMT_422P || format == IMGFMT_411P){
1533
                sortedP[0]= p[0];
1534
                sortedP[1]= p[2];
1535
                sortedP[2]= p[1];
1536
                sortedStride[0]= stride[0];
1537
                sortedStride[1]= stride[2];
1538
                sortedStride[2]= stride[1];
1539
        }
1540
        else if(isPacked(format) || isGray(format) || format == IMGFMT_Y8)
1541
        {
1542
                sortedP[0]= p[0];
1543
                sortedP[1]= 
1544
                sortedP[2]= NULL;
1545
                sortedStride[0]= stride[0];
1546
                sortedStride[1]= 
1547
                sortedStride[2]= 0;
1548
        }
1549
        else if(format == IMGFMT_I420 || format == IMGFMT_IYUV)
1550
        {
1551
                sortedP[0]= p[0];
1552
                sortedP[1]= p[1];
1553
                sortedP[2]= p[2];
1554
                sortedStride[0]= stride[0];
1555
                sortedStride[1]= stride[1];
1556
                sortedStride[2]= stride[2];
1557
        }else{
1558
                MSG_ERR("internal error in orderYUV\n");
1559
        }
1560
}
1561

    
1562
/* unscaled copy like stuff (assumes nearly identical formats) */
1563
static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1564
             int srcSliceH, uint8_t* dst[], int dstStride[]){
1565

    
1566
        if(isPacked(c->srcFormat))
1567
        {
1568
                if(dstStride[0]==srcStride[0])
1569
                        memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1570
                else
1571
                {
1572
                        int i;
1573
                        uint8_t *srcPtr= src[0];
1574
                        uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1575
                        int length=0;
1576

    
1577
                        /* universal length finder */
1578
                        while(length+c->srcW <= ABS(dstStride[0]) 
1579
                           && length+c->srcW <= ABS(srcStride[0])) length+= c->srcW;
1580
                        ASSERT(length!=0);
1581

    
1582
                        for(i=0; i<srcSliceH; i++)
1583
                        {
1584
                                memcpy(dstPtr, srcPtr, length);
1585
                                srcPtr+= srcStride[0];
1586
                                dstPtr+= dstStride[0];
1587
                        }
1588
                }
1589
        }
1590
        else 
1591
        { /* Planar YUV or gray */
1592
                int plane;
1593
                for(plane=0; plane<3; plane++)
1594
                {
1595
                        int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);
1596
                        int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1597
                        int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1598

    
1599
                        if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
1600
                        {
1601
                                if(!isGray(c->dstFormat))
1602
                                        memset(dst[plane], 128, dstStride[plane]*height);
1603
                        }
1604
                        else
1605
                        {
1606
                                if(dstStride[plane]==srcStride[plane])
1607
                                        memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1608
                                else
1609
                                {
1610
                                        int i;
1611
                                        uint8_t *srcPtr= src[plane];
1612
                                        uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1613
                                        for(i=0; i<height; i++)
1614
                                        {
1615
                                                memcpy(dstPtr, srcPtr, length);
1616
                                                srcPtr+= srcStride[plane];
1617
                                                dstPtr+= dstStride[plane];
1618
                                        }
1619
                                }
1620
                        }
1621
                }
1622
        }
1623
        return srcSliceH;
1624
}
1625

    
1626
static int remove_dup_fourcc(int fourcc)
1627
{
1628
        switch(fourcc)
1629
        {
1630
            case IMGFMT_I420:
1631
            case IMGFMT_IYUV: return IMGFMT_YV12;
1632
            case IMGFMT_Y8  : return IMGFMT_Y800;
1633
            case IMGFMT_IF09: return IMGFMT_YVU9;
1634
            default: return fourcc;
1635
        }
1636
}
1637

    
1638
static void getSubSampleFactors(int *h, int *v, int format){
1639
        switch(format){
1640
        case IMGFMT_UYVY:
1641
        case IMGFMT_YUY2:
1642
                *h=1;
1643
                *v=0;
1644
                break;
1645
        case IMGFMT_YV12:
1646
        case IMGFMT_Y800: //FIXME remove after different subsamplings are fully implemented
1647
                *h=1;
1648
                *v=1;
1649
                break;
1650
        case IMGFMT_YVU9:
1651
                *h=2;
1652
                *v=2;
1653
                break;
1654
        case IMGFMT_444P:
1655
                *h=0;
1656
                *v=0;
1657
                break;
1658
        case IMGFMT_422P:
1659
                *h=1;
1660
                *v=0;
1661
                break;
1662
        case IMGFMT_411P:
1663
                *h=2;
1664
                *v=0;
1665
                break;
1666
        default:
1667
                *h=0;
1668
                *v=0;
1669
                break;
1670
        }
1671
}
1672

    
1673
static uint16_t roundToInt16(int64_t f){
1674
        int r= (f + (1<<15))>>16;
1675
             if(r<-0x7FFF) return 0x8000;
1676
        else if(r> 0x7FFF) return 0x7FFF;
1677
        else               return r;
1678
}
1679

    
1680
/**
1681
 * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]
1682
 * @param fullRange if 1 then the luma range is 0..255 if 0 its 16..235
1683
 * @return -1 if not supported
1684
 */
1685
int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
1686
        int64_t crv =  inv_table[0];
1687
        int64_t cbu =  inv_table[1];
1688
        int64_t cgu = -inv_table[2];
1689
        int64_t cgv = -inv_table[3];
1690
        int64_t cy  = 1<<16;
1691
        int64_t oy  = 0;
1692

    
1693
        if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
1694
        memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
1695
        memcpy(c->dstColorspaceTable,     table, sizeof(int)*4);
1696

    
1697
        c->brightness= brightness;
1698
        c->contrast  = contrast;
1699
        c->saturation= saturation;
1700
        c->srcRange  = srcRange;
1701
        c->dstRange  = dstRange;
1702

    
1703
        c->uOffset=   0x0400040004000400LL;
1704
        c->vOffset=   0x0400040004000400LL;
1705

    
1706
        if(!srcRange){
1707
                cy= (cy*255) / 219;
1708
                oy= 16<<16;
1709
        }
1710

    
1711
        cy = (cy *contrast             )>>16;
1712
        crv= (crv*contrast * saturation)>>32;
1713
        cbu= (cbu*contrast * saturation)>>32;
1714
        cgu= (cgu*contrast * saturation)>>32;
1715
        cgv= (cgv*contrast * saturation)>>32;
1716

    
1717
        oy -= 256*brightness;
1718

    
1719
        c->yCoeff=    roundToInt16(cy *8192) * 0x0001000100010001ULL;
1720
        c->vrCoeff=   roundToInt16(crv*8192) * 0x0001000100010001ULL;
1721
        c->ubCoeff=   roundToInt16(cbu*8192) * 0x0001000100010001ULL;
1722
        c->vgCoeff=   roundToInt16(cgv*8192) * 0x0001000100010001ULL;
1723
        c->ugCoeff=   roundToInt16(cgu*8192) * 0x0001000100010001ULL;
1724
        c->yOffset=   roundToInt16(oy *   8) * 0x0001000100010001ULL;
1725

    
1726
        yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
1727
        //FIXME factorize
1728

    
1729
#ifdef HAVE_ALTIVEC
1730
        yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
1731
#endif        
1732
        return 0;
1733
}
1734

    
1735
/**
1736
 * @return -1 if not supported
1737
 */
1738
int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
1739
        if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
1740

    
1741
        *inv_table = c->srcColorspaceTable;
1742
        *table     = c->dstColorspaceTable;
1743
        *srcRange  = c->srcRange;
1744
        *dstRange  = c->dstRange;
1745
        *brightness= c->brightness;
1746
        *contrast  = c->contrast;
1747
        *saturation= c->saturation;
1748
        
1749
        return 0;        
1750
}
1751

    
1752
SwsContext *sws_getContext(int srcW, int srcH, int origSrcFormat, int dstW, int dstH, int origDstFormat, int flags,
1753
                         SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){
1754

    
1755
        SwsContext *c;
1756
        int i;
1757
        int usesVFilter, usesHFilter;
1758
        int unscaled, needsDither;
1759
        int srcFormat, dstFormat;
1760
        SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
1761
#if defined(ARCH_X86) || defined(ARCH_X86_64)
1762
        if(flags & SWS_CPU_CAPS_MMX)
1763
                asm volatile("emms\n\t"::: "memory");
1764
#endif
1765

    
1766
#ifndef RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
1767
        flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC);
1768
#ifdef HAVE_MMX2
1769
        flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1770
#elif defined (HAVE_3DNOW)
1771
        flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1772
#elif defined (HAVE_MMX)
1773
        flags |= SWS_CPU_CAPS_MMX;
1774
#elif defined (HAVE_ALTIVEC)
1775
        flags |= SWS_CPU_CAPS_ALTIVEC;
1776
#endif
1777
#endif
1778
        if(clip_table[512] != 255) globalInit();
1779
        if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);
1780

    
1781
        /* avoid duplicate Formats, so we don't need to check to much */
1782
        srcFormat = remove_dup_fourcc(origSrcFormat);
1783
        dstFormat = remove_dup_fourcc(origDstFormat);
1784

    
1785
        unscaled = (srcW == dstW && srcH == dstH);
1786
        needsDither= (isBGR(dstFormat) || isRGB(dstFormat)) 
1787
                     && (dstFormat&0xFF)<24
1788
                     && ((dstFormat&0xFF)<(srcFormat&0xFF) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
1789

    
1790
        if(!isSupportedIn(srcFormat)) 
1791
        {
1792
                MSG_ERR("swScaler: %s is not supported as input format\n", vo_format_name(srcFormat));
1793
                return NULL;
1794
        }
1795
        if(!isSupportedOut(dstFormat))
1796
        {
1797
                MSG_ERR("swScaler: %s is not supported as output format\n", vo_format_name(dstFormat));
1798
                return NULL;
1799
        }
1800

    
1801
        /* sanity check */
1802
        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
1803
        {
1804
                 MSG_ERR("swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", 
1805
                        srcW, srcH, dstW, dstH);
1806
                return NULL;
1807
        }
1808

    
1809
        if(!dstFilter) dstFilter= &dummyFilter;
1810
        if(!srcFilter) srcFilter= &dummyFilter;
1811

    
1812
        c= memalign(64, sizeof(SwsContext));
1813
        memset(c, 0, sizeof(SwsContext));
1814

    
1815
        c->srcW= srcW;
1816
        c->srcH= srcH;
1817
        c->dstW= dstW;
1818
        c->dstH= dstH;
1819
        c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
1820
        c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
1821
        c->flags= flags;
1822
        c->dstFormat= dstFormat;
1823
        c->srcFormat= srcFormat;
1824
        c->origDstFormat= origDstFormat;
1825
        c->origSrcFormat= origSrcFormat;
1826
        c->vRounder= 4* 0x0001000100010001ULL;
1827

    
1828
        usesHFilter= usesVFilter= 0;
1829
        if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;
1830
        if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;
1831
        if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;
1832
        if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;
1833
        if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;
1834
        if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;
1835
        if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;
1836
        if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;
1837

    
1838
        getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
1839
        getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
1840

    
1841
        // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
1842
        if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
1843

    
1844
        // drop some chroma lines if the user wants it
1845
        c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
1846
        c->chrSrcVSubSample+= c->vChrDrop;
1847

    
1848
        // drop every 2. pixel for chroma calculation unless user wants full chroma
1849
        if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)) 
1850
                c->chrSrcHSubSample=1;
1851

    
1852
        if(param){
1853
                c->param[0] = param[0];
1854
                c->param[1] = param[1];
1855
        }else{
1856
                c->param[0] =
1857
                c->param[1] = SWS_PARAM_DEFAULT;
1858
        }
1859

    
1860
        c->chrIntHSubSample= c->chrDstHSubSample;
1861
        c->chrIntVSubSample= c->chrSrcVSubSample;
1862

    
1863
        // note the -((-x)>>y) is so that we allways round toward +inf
1864
        c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
1865
        c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
1866
        c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
1867
        c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
1868

    
1869
        sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], 0, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, 0, 0, 1<<16, 1<<16); 
1870

    
1871
        /* unscaled special Cases */
1872
        if(unscaled && !usesHFilter && !usesVFilter)
1873
        {
1874
                /* yv12_to_nv12 */
1875
                if(srcFormat == IMGFMT_YV12 && dstFormat == IMGFMT_NV12)
1876
                {
1877
                        c->swScale= PlanarToNV12Wrapper;
1878
                }
1879
                /* yuv2bgr */
1880
                if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
1881
                {
1882
                        c->swScale= yuv2rgb_get_func_ptr(c);
1883
                }
1884
                
1885
                if( srcFormat==IMGFMT_YVU9 && dstFormat==IMGFMT_YV12 )
1886
                {
1887
                        c->swScale= yvu9toyv12Wrapper;
1888
                }
1889

    
1890
                /* bgr24toYV12 */
1891
                if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12)
1892
                        c->swScale= bgr24toyv12Wrapper;
1893
                
1894
                /* rgb/bgr -> rgb/bgr (no dither needed forms) */
1895
                if(   (isBGR(srcFormat) || isRGB(srcFormat))
1896
                   && (isBGR(dstFormat) || isRGB(dstFormat)) 
1897
                   && !needsDither)
1898
                        c->swScale= rgb2rgbWrapper;
1899

    
1900
                /* LQ converters if -sws 0 or -sws 4*/
1901
                if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
1902
                        /* rgb/bgr -> rgb/bgr (dither needed forms) */
1903
                        if(  (isBGR(srcFormat) || isRGB(srcFormat))
1904
                          && (isBGR(dstFormat) || isRGB(dstFormat)) 
1905
                          && needsDither)
1906
                                c->swScale= rgb2rgbWrapper;
1907

    
1908
                        /* yv12_to_yuy2 */
1909
                        if(srcFormat == IMGFMT_YV12 && 
1910
                            (dstFormat == IMGFMT_YUY2 || dstFormat == IMGFMT_UYVY))
1911
                        {
1912
                                if (dstFormat == IMGFMT_YUY2)
1913
                                    c->swScale= PlanarToYuy2Wrapper;
1914
                                else
1915
                                    c->swScale= PlanarToUyvyWrapper;
1916
                        }
1917
                }
1918

    
1919
#ifdef HAVE_ALTIVEC
1920
                if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1921
                    ((srcFormat == IMGFMT_YV12 && 
1922
                      (dstFormat == IMGFMT_YUY2 || dstFormat == IMGFMT_UYVY)))) {
1923
                  // unscaled YV12 -> packed YUV, we want speed
1924
                  if (dstFormat == IMGFMT_YUY2)
1925
                    c->swScale= yv12toyuy2_unscaled_altivec;
1926
                  else
1927
                    c->swScale= yv12touyvy_unscaled_altivec;
1928
                }
1929
#endif
1930

    
1931
                /* simple copy */
1932
                if(   srcFormat == dstFormat
1933
                   || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1934
                   || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1935
                  )
1936
                {
1937
                        c->swScale= simpleCopy;
1938
                }
1939

    
1940
                if(c->swScale){
1941
                        if(flags&SWS_PRINT_INFO)
1942
                                MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", 
1943
                                        vo_format_name(srcFormat), vo_format_name(dstFormat));
1944
                        return c;
1945
                }
1946
        }
1947

    
1948
        if(flags & SWS_CPU_CAPS_MMX2)
1949
        {
1950
                c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
1951
                if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
1952
                {
1953
                        if(flags&SWS_PRINT_INFO)
1954
                                MSG_INFO("SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");
1955
                }
1956
                if(usesHFilter) c->canMMX2BeUsed=0;
1957
        }
1958
        else
1959
                c->canMMX2BeUsed=0;
1960

    
1961
        c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
1962
        c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
1963

    
1964
        // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
1965
        // but only for the FAST_BILINEAR mode otherwise do correct scaling
1966
        // n-2 is the last chrominance sample available
1967
        // this is not perfect, but noone shuld notice the difference, the more correct variant
1968
        // would be like the vertical one, but that would require some special code for the
1969
        // first and last pixel
1970
        if(flags&SWS_FAST_BILINEAR)
1971
        {
1972
                if(c->canMMX2BeUsed)
1973
                {
1974
                        c->lumXInc+= 20;
1975
                        c->chrXInc+= 20;
1976
                }
1977
                //we don't use the x86asm scaler if mmx is available
1978
                else if(flags & SWS_CPU_CAPS_MMX)
1979
                {
1980
                        c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
1981
                        c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
1982
                }
1983
        }
1984

    
1985
        /* precalculate horizontal scaler filter coefficients */
1986
        {
1987
                const int filterAlign=
1988
                  (flags & SWS_CPU_CAPS_MMX) ? 4 :
1989
                  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
1990
                  1;
1991

    
1992
                initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
1993
                                 srcW      ,       dstW, filterAlign, 1<<14,
1994
                                 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
1995
                                 srcFilter->lumH, dstFilter->lumH, c->param);
1996
                initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
1997
                                 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
1998
                                 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
1999
                                 srcFilter->chrH, dstFilter->chrH, c->param);
2000

    
2001
#if defined(ARCH_X86) || defined(ARCH_X86_64)
2002
// can't downscale !!!
2003
                if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2004
                {
2005
#define MAX_FUNNY_CODE_SIZE 10000
2006
#ifdef HAVE_SYS_MMAN_H
2007
                        c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2008
                        c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2009
#else
2010
                        c->funnyYCode = (uint8_t*)memalign(32, MAX_FUNNY_CODE_SIZE);
2011
                        c->funnyUVCode = (uint8_t*)memalign(32, MAX_FUNNY_CODE_SIZE);
2012
#endif
2013

    
2014
                        c->lumMmx2Filter   = (int16_t*)memalign(8, (dstW        /8+8)*sizeof(int16_t));
2015
                        c->chrMmx2Filter   = (int16_t*)memalign(8, (c->chrDstW  /4+8)*sizeof(int16_t));
2016
                        c->lumMmx2FilterPos= (int32_t*)memalign(8, (dstW      /2/8+8)*sizeof(int32_t));
2017
                        c->chrMmx2FilterPos= (int32_t*)memalign(8, (c->chrDstW/2/4+8)*sizeof(int32_t));
2018

    
2019
                        initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2020
                        initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2021
                }
2022
#endif
2023
        } // Init Horizontal stuff
2024

    
2025

    
2026

    
2027
        /* precalculate vertical scaler filter coefficients */
2028
        {
2029
                const int filterAlign=
2030
                  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2031
                  1;
2032

    
2033
                initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2034
                                srcH      ,        dstH, filterAlign, (1<<12)-4,
2035
                                (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
2036
                                srcFilter->lumV, dstFilter->lumV, c->param);
2037
                initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2038
                                c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,
2039
                                (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2040
                                srcFilter->chrV, dstFilter->chrV, c->param);
2041
        }
2042

    
2043
        // Calculate Buffer Sizes so that they won't run out while handling these damn slices
2044
        c->vLumBufSize= c->vLumFilterSize;
2045
        c->vChrBufSize= c->vChrFilterSize;
2046
        for(i=0; i<dstH; i++)
2047
        {
2048
                int chrI= i*c->chrDstH / dstH;
2049
                int nextSlice= MAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
2050
                                 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2051

    
2052
                nextSlice>>= c->chrSrcVSubSample;
2053
                nextSlice<<= c->chrSrcVSubSample;
2054
                if(c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
2055
                        c->vLumBufSize= nextSlice - c->vLumFilterPos[i   ];
2056
                if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2057
                        c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2058
        }
2059

    
2060
        // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2061
        c->lumPixBuf= (int16_t**)memalign(4, c->vLumBufSize*2*sizeof(int16_t*));
2062
        c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*));
2063
        //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)
2064
        for(i=0; i<c->vLumBufSize; i++)
2065
                c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= (uint16_t*)memalign(8, 4000);
2066
        for(i=0; i<c->vChrBufSize; i++)
2067
                c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= (uint16_t*)memalign(8, 8000);
2068

    
2069
        //try to avoid drawing green stuff between the right end and the stride end
2070
        for(i=0; i<c->vLumBufSize; i++) memset(c->lumPixBuf[i], 0, 4000);
2071
        for(i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000);
2072

    
2073
        ASSERT(c->chrDstH <= dstH)
2074

    
2075
        if(flags&SWS_PRINT_INFO)
2076
        {
2077
#ifdef DITHER1XBPP
2078
                char *dither= " dithered";
2079
#else
2080
                char *dither= "";
2081
#endif
2082
                if(flags&SWS_FAST_BILINEAR)
2083
                        MSG_INFO("\nSwScaler: FAST_BILINEAR scaler, ");
2084
                else if(flags&SWS_BILINEAR)
2085
                        MSG_INFO("\nSwScaler: BILINEAR scaler, ");
2086
                else if(flags&SWS_BICUBIC)
2087
                        MSG_INFO("\nSwScaler: BICUBIC scaler, ");
2088
                else if(flags&SWS_X)
2089
                        MSG_INFO("\nSwScaler: Experimental scaler, ");
2090
                else if(flags&SWS_POINT)
2091
                        MSG_INFO("\nSwScaler: Nearest Neighbor / POINT scaler, ");
2092
                else if(flags&SWS_AREA)
2093
                        MSG_INFO("\nSwScaler: Area Averageing scaler, ");
2094
                else if(flags&SWS_BICUBLIN)
2095
                        MSG_INFO("\nSwScaler: luma BICUBIC / chroma BILINEAR scaler, ");
2096
                else if(flags&SWS_GAUSS)
2097
                        MSG_INFO("\nSwScaler: Gaussian scaler, ");
2098
                else if(flags&SWS_SINC)
2099
                        MSG_INFO("\nSwScaler: Sinc scaler, ");
2100
                else if(flags&SWS_LANCZOS)
2101
                        MSG_INFO("\nSwScaler: Lanczos scaler, ");
2102
                else if(flags&SWS_SPLINE)
2103
                        MSG_INFO("\nSwScaler: Bicubic spline scaler, ");
2104
                else
2105
                        MSG_INFO("\nSwScaler: ehh flags invalid?! ");
2106

    
2107
                if(dstFormat==IMGFMT_BGR15 || dstFormat==IMGFMT_BGR16)
2108
                        MSG_INFO("from %s to%s %s ", 
2109
                                vo_format_name(srcFormat), dither, vo_format_name(dstFormat));
2110
                else
2111
                        MSG_INFO("from %s to %s ", 
2112
                                vo_format_name(srcFormat), vo_format_name(dstFormat));
2113

    
2114
                if(flags & SWS_CPU_CAPS_MMX2)
2115
                        MSG_INFO("using MMX2\n");
2116
                else if(flags & SWS_CPU_CAPS_3DNOW)
2117
                        MSG_INFO("using 3DNOW\n");
2118
                else if(flags & SWS_CPU_CAPS_MMX)
2119
                        MSG_INFO("using MMX\n");
2120
                else if(flags & SWS_CPU_CAPS_ALTIVEC)
2121
                        MSG_INFO("using AltiVec\n");
2122
                else 
2123
                        MSG_INFO("using C\n");
2124
        }
2125

    
2126
        if(flags & SWS_PRINT_INFO)
2127
        {
2128
                if(flags & SWS_CPU_CAPS_MMX)
2129
                {
2130
                        if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2131
                                MSG_V("SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2132
                        else
2133
                        {
2134
                                if(c->hLumFilterSize==4)
2135
                                        MSG_V("SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n");
2136
                                else if(c->hLumFilterSize==8)
2137
                                        MSG_V("SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n");
2138
                                else
2139
                                        MSG_V("SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n");
2140

    
2141
                                if(c->hChrFilterSize==4)
2142
                                        MSG_V("SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n");
2143
                                else if(c->hChrFilterSize==8)
2144
                                        MSG_V("SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n");
2145
                                else
2146
                                        MSG_V("SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n");
2147
                        }
2148
                }
2149
                else
2150
                {
2151
#if defined(ARCH_X86) || defined(ARCH_X86_64)
2152
                        MSG_V("SwScaler: using X86-Asm scaler for horizontal scaling\n");
2153
#else
2154
                        if(flags & SWS_FAST_BILINEAR)
2155
                                MSG_V("SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n");
2156
                        else
2157
                                MSG_V("SwScaler: using C scaler for horizontal scaling\n");
2158
#endif
2159
                }
2160
                if(isPlanarYUV(dstFormat))
2161
                {
2162
                        if(c->vLumFilterSize==1)
2163
                                MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2164
                        else
2165
                                MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2166
                }
2167
                else
2168
                {
2169
                        if(c->vLumFilterSize==1 && c->vChrFilterSize==2)
2170
                                MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2171
                                       "SwScaler:       2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2172
                        else if(c->vLumFilterSize==2 && c->vChrFilterSize==2)
2173
                                MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2174
                        else
2175
                                MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2176
                }
2177

    
2178
                if(dstFormat==IMGFMT_BGR24)
2179
                        MSG_V("SwScaler: using %s YV12->BGR24 Converter\n",
2180
                                (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2181
                else if(dstFormat==IMGFMT_BGR32)
2182
                        MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2183
                else if(dstFormat==IMGFMT_BGR16)
2184
                        MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2185
                else if(dstFormat==IMGFMT_BGR15)
2186
                        MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2187

    
2188
                MSG_V("SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2189
        }
2190
        if(flags & SWS_PRINT_INFO)
2191
        {
2192
                MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2193
                        c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2194
                MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2195
                        c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2196
        }
2197

    
2198
        c->swScale= getSwsFunc(flags);
2199
        return c;
2200
}
2201

    
2202
/**
2203
 * swscale warper, so we don't need to export the SwsContext.
2204
 * assumes planar YUV to be in YUV order instead of YVU
2205
 */
2206
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2207
                           int srcSliceH, uint8_t* dst[], int dstStride[]){
2208
        //copy strides, so they can safely be modified
2209
        int srcStride2[3]= {srcStride[0], srcStride[1], srcStride[2]};
2210
        int dstStride2[3]= {dstStride[0], dstStride[1], dstStride[2]};
2211
        return c->swScale(c, src, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
2212
}
2213

    
2214
/**
2215
 * swscale warper, so we don't need to export the SwsContext
2216
 */
2217
int sws_scale(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY,
2218
                           int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){
2219
        int srcStride[3];
2220
        int dstStride[3];
2221
        uint8_t *src[3];
2222
        uint8_t *dst[3];
2223
        sws_orderYUV(c->origSrcFormat, src, srcStride, srcParam, srcStrideParam);
2224
        sws_orderYUV(c->origDstFormat, dst, dstStride, dstParam, dstStrideParam);
2225
//printf("sws: slice %d %d\n", srcSliceY, srcSliceH);
2226

    
2227
        return c->swScale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2228
}
2229

    
2230
SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur, 
2231
                                float lumaSharpen, float chromaSharpen,
2232
                                float chromaHShift, float chromaVShift,
2233
                                int verbose)
2234
{
2235
        SwsFilter *filter= malloc(sizeof(SwsFilter));
2236

    
2237
        if(lumaGBlur!=0.0){
2238
                filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
2239
                filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
2240
        }else{
2241
                filter->lumH= sws_getIdentityVec();
2242
                filter->lumV= sws_getIdentityVec();
2243
        }
2244

    
2245
        if(chromaGBlur!=0.0){
2246
                filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
2247
                filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
2248
        }else{
2249
                filter->chrH= sws_getIdentityVec();
2250
                filter->chrV= sws_getIdentityVec();
2251
        }
2252

    
2253
        if(chromaSharpen!=0.0){
2254
                SwsVector *g= sws_getConstVec(-1.0, 3);
2255
                SwsVector *id= sws_getConstVec(10.0/chromaSharpen, 1);
2256
                g->coeff[1]=2.0;
2257
                sws_addVec(id, g);
2258
                sws_convVec(filter->chrH, id);
2259
                sws_convVec(filter->chrV, id);
2260
                sws_freeVec(g);
2261
                sws_freeVec(id);
2262
        }
2263

    
2264
        if(lumaSharpen!=0.0){
2265
                SwsVector *g= sws_getConstVec(-1.0, 3);
2266
                SwsVector *id= sws_getConstVec(10.0/lumaSharpen, 1);
2267
                g->coeff[1]=2.0;
2268
                sws_addVec(id, g);
2269
                sws_convVec(filter->lumH, id);
2270
                sws_convVec(filter->lumV, id);
2271
                sws_freeVec(g);
2272
                sws_freeVec(id);
2273
        }
2274

    
2275
        if(chromaHShift != 0.0)
2276
                sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
2277

    
2278
        if(chromaVShift != 0.0)
2279
                sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
2280

    
2281
        sws_normalizeVec(filter->chrH, 1.0);
2282
        sws_normalizeVec(filter->chrV, 1.0);
2283
        sws_normalizeVec(filter->lumH, 1.0);
2284
        sws_normalizeVec(filter->lumV, 1.0);
2285

    
2286
        if(verbose) sws_printVec(filter->chrH);
2287
        if(verbose) sws_printVec(filter->lumH);
2288

    
2289
        return filter;
2290
}
2291

    
2292
/**
2293
 * returns a normalized gaussian curve used to filter stuff
2294
 * quality=3 is high quality, lowwer is lowwer quality
2295
 */
2296
SwsVector *sws_getGaussianVec(double variance, double quality){
2297
        const int length= (int)(variance*quality + 0.5) | 1;
2298
        int i;
2299
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2300
        double middle= (length-1)*0.5;
2301
        SwsVector *vec= malloc(sizeof(SwsVector));
2302

    
2303
        vec->coeff= coeff;
2304
        vec->length= length;
2305

    
2306
        for(i=0; i<length; i++)
2307
        {
2308
                double dist= i-middle;
2309
                coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI);
2310
        }
2311

    
2312
        sws_normalizeVec(vec, 1.0);
2313

    
2314
        return vec;
2315
}
2316

    
2317
SwsVector *sws_getConstVec(double c, int length){
2318
        int i;
2319
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2320
        SwsVector *vec= malloc(sizeof(SwsVector));
2321

    
2322
        vec->coeff= coeff;
2323
        vec->length= length;
2324

    
2325
        for(i=0; i<length; i++)
2326
                coeff[i]= c;
2327

    
2328
        return vec;
2329
}
2330

    
2331

    
2332
SwsVector *sws_getIdentityVec(void){
2333
        double *coeff= memalign(sizeof(double), sizeof(double));
2334
        SwsVector *vec= malloc(sizeof(SwsVector));
2335
        coeff[0]= 1.0;
2336

    
2337
        vec->coeff= coeff;
2338
        vec->length= 1;
2339

    
2340
        return vec;
2341
}
2342

    
2343
void sws_normalizeVec(SwsVector *a, double height){
2344
        int i;
2345
        double sum=0;
2346
        double inv;
2347

    
2348
        for(i=0; i<a->length; i++)
2349
                sum+= a->coeff[i];
2350

    
2351
        inv= height/sum;
2352

    
2353
        for(i=0; i<a->length; i++)
2354
                a->coeff[i]*= inv;
2355
}
2356

    
2357
void sws_scaleVec(SwsVector *a, double scalar){
2358
        int i;
2359

    
2360
        for(i=0; i<a->length; i++)
2361
                a->coeff[i]*= scalar;
2362
}
2363

    
2364
static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
2365
        int length= a->length + b->length - 1;
2366
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2367
        int i, j;
2368
        SwsVector *vec= malloc(sizeof(SwsVector));
2369

    
2370
        vec->coeff= coeff;
2371
        vec->length= length;
2372

    
2373
        for(i=0; i<length; i++) coeff[i]= 0.0;
2374

    
2375
        for(i=0; i<a->length; i++)
2376
        {
2377
                for(j=0; j<b->length; j++)
2378
                {
2379
                        coeff[i+j]+= a->coeff[i]*b->coeff[j];
2380
                }
2381
        }
2382

    
2383
        return vec;
2384
}
2385

    
2386
static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
2387
        int length= MAX(a->length, b->length);
2388
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2389
        int i;
2390
        SwsVector *vec= malloc(sizeof(SwsVector));
2391

    
2392
        vec->coeff= coeff;
2393
        vec->length= length;
2394

    
2395
        for(i=0; i<length; i++) coeff[i]= 0.0;
2396

    
2397
        for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2398
        for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
2399

    
2400
        return vec;
2401
}
2402

    
2403
static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
2404
        int length= MAX(a->length, b->length);
2405
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2406
        int i;
2407
        SwsVector *vec= malloc(sizeof(SwsVector));
2408

    
2409
        vec->coeff= coeff;
2410
        vec->length= length;
2411

    
2412
        for(i=0; i<length; i++) coeff[i]= 0.0;
2413

    
2414
        for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2415
        for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
2416

    
2417
        return vec;
2418
}
2419

    
2420
/* shift left / or right if "shift" is negative */
2421
static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
2422
        int length= a->length + ABS(shift)*2;
2423
        double *coeff= memalign(sizeof(double), length*sizeof(double));
2424
        int i;
2425
        SwsVector *vec= malloc(sizeof(SwsVector));
2426

    
2427
        vec->coeff= coeff;
2428
        vec->length= length;
2429

    
2430
        for(i=0; i<length; i++) coeff[i]= 0.0;
2431

    
2432
        for(i=0; i<a->length; i++)
2433
        {
2434
                coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
2435
        }
2436

    
2437
        return vec;
2438
}
2439

    
2440
void sws_shiftVec(SwsVector *a, int shift){
2441
        SwsVector *shifted= sws_getShiftedVec(a, shift);
2442
        free(a->coeff);
2443
        a->coeff= shifted->coeff;
2444
        a->length= shifted->length;
2445
        free(shifted);
2446
}
2447

    
2448
void sws_addVec(SwsVector *a, SwsVector *b){
2449
        SwsVector *sum= sws_sumVec(a, b);
2450
        free(a->coeff);
2451
        a->coeff= sum->coeff;
2452
        a->length= sum->length;
2453
        free(sum);
2454
}
2455

    
2456
void sws_subVec(SwsVector *a, SwsVector *b){
2457
        SwsVector *diff= sws_diffVec(a, b);
2458
        free(a->coeff);
2459
        a->coeff= diff->coeff;
2460
        a->length= diff->length;
2461
        free(diff);
2462
}
2463

    
2464
void sws_convVec(SwsVector *a, SwsVector *b){
2465
        SwsVector *conv= sws_getConvVec(a, b);
2466
        free(a->coeff);  
2467
        a->coeff= conv->coeff;
2468
        a->length= conv->length;
2469
        free(conv);
2470
}
2471

    
2472
SwsVector *sws_cloneVec(SwsVector *a){
2473
        double *coeff= memalign(sizeof(double), a->length*sizeof(double));
2474
        int i;
2475
        SwsVector *vec= malloc(sizeof(SwsVector));
2476

    
2477
        vec->coeff= coeff;
2478
        vec->length= a->length;
2479

    
2480
        for(i=0; i<a->length; i++) coeff[i]= a->coeff[i];
2481

    
2482
        return vec;
2483
}
2484

    
2485
void sws_printVec(SwsVector *a){
2486
        int i;
2487
        double max=0;
2488
        double min=0;
2489
        double range;
2490

    
2491
        for(i=0; i<a->length; i++)
2492
                if(a->coeff[i]>max) max= a->coeff[i];
2493

    
2494
        for(i=0; i<a->length; i++)
2495
                if(a->coeff[i]<min) min= a->coeff[i];
2496

    
2497
        range= max - min;
2498

    
2499
        for(i=0; i<a->length; i++)
2500
        {
2501
                int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
2502
                MSG_DBG2("%1.3f ", a->coeff[i]);
2503
                for(;x>0; x--) MSG_DBG2(" ");
2504
                MSG_DBG2("|\n");
2505
        }
2506
}
2507

    
2508
void sws_freeVec(SwsVector *a){
2509
        if(!a) return;
2510
        if(a->coeff) free(a->coeff);
2511
        a->coeff=NULL;
2512
        a->length=0;
2513
        free(a);
2514
}
2515

    
2516
void sws_freeFilter(SwsFilter *filter){
2517
        if(!filter) return;
2518

    
2519
        if(filter->lumH) sws_freeVec(filter->lumH);
2520
        if(filter->lumV) sws_freeVec(filter->lumV);
2521
        if(filter->chrH) sws_freeVec(filter->chrH);
2522
        if(filter->chrV) sws_freeVec(filter->chrV);
2523
        free(filter);
2524
}
2525

    
2526

    
2527
void sws_freeContext(SwsContext *c){
2528
        int i;
2529
        if(!c) return;
2530

    
2531
        if(c->lumPixBuf)
2532
        {
2533
                for(i=0; i<c->vLumBufSize; i++)
2534
                {
2535
                        if(c->lumPixBuf[i]) free(c->lumPixBuf[i]);
2536
                        c->lumPixBuf[i]=NULL;
2537
                }
2538
                free(c->lumPixBuf);
2539
                c->lumPixBuf=NULL;
2540
        }
2541

    
2542
        if(c->chrPixBuf)
2543
        {
2544
                for(i=0; i<c->vChrBufSize; i++)
2545
                {
2546
                        if(c->chrPixBuf[i]) free(c->chrPixBuf[i]);
2547
                        c->chrPixBuf[i]=NULL;
2548
                }
2549
                free(c->chrPixBuf);
2550
                c->chrPixBuf=NULL;
2551
        }
2552

    
2553
        if(c->vLumFilter) free(c->vLumFilter);
2554
        c->vLumFilter = NULL;
2555
        if(c->vChrFilter) free(c->vChrFilter);
2556
        c->vChrFilter = NULL;
2557
        if(c->hLumFilter) free(c->hLumFilter);
2558
        c->hLumFilter = NULL;
2559
        if(c->hChrFilter) free(c->hChrFilter);
2560
        c->hChrFilter = NULL;
2561

    
2562
        if(c->vLumFilterPos) free(c->vLumFilterPos);
2563
        c->vLumFilterPos = NULL;
2564
        if(c->vChrFilterPos) free(c->vChrFilterPos);
2565
        c->vChrFilterPos = NULL;
2566
        if(c->hLumFilterPos) free(c->hLumFilterPos);
2567
        c->hLumFilterPos = NULL;
2568
        if(c->hChrFilterPos) free(c->hChrFilterPos);
2569
        c->hChrFilterPos = NULL;
2570

    
2571
#ifdef HAVE_SYS_MMAN_H
2572
        if(c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);
2573
        if(c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
2574
#else
2575
        if(c->funnyYCode) free(c->funnyYCode);
2576
        if(c->funnyUVCode) free(c->funnyUVCode);
2577
#endif
2578
        c->funnyYCode=NULL;
2579
        c->funnyUVCode=NULL;
2580

    
2581
        if(c->lumMmx2Filter) free(c->lumMmx2Filter);
2582
        c->lumMmx2Filter=NULL;
2583
        if(c->chrMmx2Filter) free(c->chrMmx2Filter);
2584
        c->chrMmx2Filter=NULL;
2585
        if(c->lumMmx2FilterPos) free(c->lumMmx2FilterPos);
2586
        c->lumMmx2FilterPos=NULL;
2587
        if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos);
2588
        c->chrMmx2FilterPos=NULL;
2589
        if(c->yuvTable) free(c->yuvTable);
2590
        c->yuvTable=NULL;
2591

    
2592
        free(c);
2593
}
2594