projects / libav.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
check for channels<=0 and print a reasonable error message
[libav.git] / libswscale / swscale.c
CommitLineData
fe8054c0 1/*
d026b45e
DB		 2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
19 *
807e0c66
LA		 20 * the C code (not assembly, mmx, ...) of this file can be used
21 * under the LGPL license too
d026b45e 22 */
783e9cc9 23
28bf81c9 24/*
e28630fc 25 supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09, PAL8
caeaabe7 26 supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
e09d12f4 27 {BGR,RGB}{1,4,8,15,16} support dithering
6a4970ab 28
e09d12f4
MN		 29 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
30 YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
31 x -> x
32 YUV9 -> YV12
33 YUV9/YV12 -> Y800
34 Y800 -> YUV9/YV12
b935781b
MN		 35 BGR24 -> BGR32 & RGB24 -> RGB32
36 BGR32 -> BGR24 & RGB32 -> RGB24
4bb3fa5e 37 BGR15 -> BGR16
b935781b
MN		 38*/
39
6a4970ab 40/*
e09d12f4
MN		 41tested special converters (most are tested actually but i didnt write it down ...)
42 YV12 -> BGR16
b935781b 43 YV12 -> YV12
4bb3fa5e 44 BGR15 -> BGR16
1e1c4fe9 45 BGR16 -> BGR16
e09d12f4 46 YVU9 -> YV12
b935781b
MN		 47
48untested special converters
1e1c4fe9
MN		 49 YV12/I420 -> BGR15/BGR24/BGR32 (its the yuv2rgb stuff, so it should be ok)
50 YV12/I420 -> YV12/I420
51 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
b935781b
MN		 52 BGR24 -> BGR32 & RGB24 -> RGB32
53 BGR32 -> BGR24 & RGB32 -> RGB24
ec22603f 54 BGR24 -> YV12
28bf81c9
MN		 55*/
56
d3f41512 57#include <inttypes.h>
dda87e9f 58#include <string.h>
077ea8a7 59#include <math.h>
c1b0bfb4 60#include <stdio.h>
171d7d78 61#include <unistd.h>
b2d374c9 62#include "config.h"
81b7c056 63#include <assert.h>
38d5c282
AJ		 64#ifdef HAVE_SYS_MMAN_H
65#include <sys/mman.h>
113ef149
RD		 66#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
67#define MAP_ANONYMOUS MAP_ANON
68#endif
38d5c282 69#endif
d604bab9 70#include "swscale.h"
5427e242 71#include "swscale_internal.h"
94c4def2 72#include "x86_cpu.h"
b2d374c9 73#include "bswap.h"
37079906 74#include "rgb2rgb.h"
94c4def2 75#ifdef USE_FASTMEMCPY
b2d374c9 76#include "libvo/fastmemcpy.h"
94c4def2 77#endif
0d9f3d85 78
541c4eb9 79#undef MOVNTQ
7d7f78b5 80#undef PAVGB
d3f41512 81
783e9cc9 82//#undef HAVE_MMX2
7f56a527 83//#define HAVE_3DNOW
d3f41512 84//#undef HAVE_MMX
783e9cc9 85//#undef ARCH_X86
2ba1bff0 86//#define WORDS_BIGENDIAN
d604bab9 87#define DITHER1XBPP
d3f41512 88
ac6a2e45
MN		 89#define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit
90
1e621b18 91#define RET 0xC3 //near return opcode for X86
c1b0bfb4 92
28bf81c9 93#ifdef MP_DEBUG
81b7c056 94#define ASSERT(x) assert(x);
28bf81c9 95#else
c1b0bfb4 96#define ASSERT(x) ;
28bf81c9
MN		 97#endif
98
99#ifdef M_PI
100#define PI M_PI
101#else
102#define PI 3.14159265358979323846
103#endif
c1b0bfb4 104
e9e12f0e
LA		 105#define isSupportedIn(x) ((x)==PIX_FMT_YUV420P || (x)==PIX_FMT_YUYV422 || (x)==PIX_FMT_UYVY422\
106 || (x)==PIX_FMT_RGB32|| (x)==PIX_FMT_BGR24|| (x)==PIX_FMT_BGR565|| (x)==PIX_FMT_BGR555\
a43fb6b3 107 || (x)==PIX_FMT_BGR32|| (x)==PIX_FMT_RGB24|| (x)==PIX_FMT_RGB565|| (x)==PIX_FMT_RGB555\
e9e12f0e 108 || (x)==PIX_FMT_GRAY8 || (x)==PIX_FMT_YUV410P\
4884b9e5 109 || (x)==PIX_FMT_GRAY16BE || (x)==PIX_FMT_GRAY16LE\
e28630fc 110 || (x)==PIX_FMT_YUV444P || (x)==PIX_FMT_YUV422P || (x)==PIX_FMT_YUV411P\
18064f5c
MN		 111 || (x)==PIX_FMT_PAL8 || (x)==PIX_FMT_BGR8 || (x)==PIX_FMT_RGB8\
112 || (x)==PIX_FMT_BGR4_BYTE || (x)==PIX_FMT_RGB4_BYTE)
e9e12f0e
LA		 113#define isSupportedOut(x) ((x)==PIX_FMT_YUV420P || (x)==PIX_FMT_YUYV422 || (x)==PIX_FMT_UYVY422\
114 || (x)==PIX_FMT_YUV444P || (x)==PIX_FMT_YUV422P || (x)==PIX_FMT_YUV411P\
cf7d1c1a 115 || isRGB(x) || isBGR(x)\
e9e12f0e 116 || (x)==PIX_FMT_NV12 || (x)==PIX_FMT_NV21\
4884b9e5 117 || (x)==PIX_FMT_GRAY16BE || (x)==PIX_FMT_GRAY16LE\
e9e12f0e 118 || (x)==PIX_FMT_GRAY8 || (x)==PIX_FMT_YUV410P)
6cbac7ed
KB		 119#define isPacked(x) ((x)==PIX_FMT_PAL8 || (x)==PIX_FMT_YUYV422 ||\
120 (x)==PIX_FMT_UYVY422 || isRGB(x) || isBGR(x))
6ff0ad6b
MN		 121
122#define RGB2YUV_SHIFT 16
1e621b18
MN		 123#define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5))
124#define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5))
125#define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
126#define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5))
127#define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5))
128#define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5))
129#define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5))
130#define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5))
131#define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5))
6c7506de 132
0481412a
MN		 133extern const int32_t Inverse_Table_6_9[8][4];
134
783e9cc9
MN		 135/*
136NOTES
d604bab9 137Special versions: fast Y 1:1 scaling (no interpolation in y direction)
31190492 138
783e9cc9 139TODO
d604bab9 140more intelligent missalignment avoidance for the horizontal scaler
c1b0bfb4
MN		 141write special vertical cubic upscale version
142Optimize C code (yv12 / minmax)
ff7ba856 143add support for packed pixel yuv input & output
6ff0ad6b
MN		 144add support for Y8 output
145optimize bgr24 & bgr32
ff7ba856 146add BGR4 output support
1e621b18 147write special BGR->BGR scaler
783e9cc9 148*/
31190492 149
9bde778e 150#if defined(ARCH_X86) && defined (CONFIG_GPL)
db7a2e0d
MC		 151static uint64_t attribute_used __attribute__((aligned(8))) bF8= 0xF8F8F8F8F8F8F8F8LL;
152static uint64_t attribute_used __attribute__((aligned(8))) bFC= 0xFCFCFCFCFCFCFCFCLL;
d604bab9 153static uint64_t __attribute__((aligned(8))) w10= 0x0010001000100010LL;
db7a2e0d
MC		 154static uint64_t attribute_used __attribute__((aligned(8))) w02= 0x0002000200020002LL;
155static uint64_t attribute_used __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL;
156static uint64_t attribute_used __attribute__((aligned(8))) bm00000111=0x0000000000FFFFFFLL;
157static uint64_t attribute_used __attribute__((aligned(8))) bm11111000=0xFFFFFFFFFF000000LL;
158static uint64_t attribute_used __attribute__((aligned(8))) bm01010101=0x00FF00FF00FF00FFLL;
d604bab9 159
db7a2e0d
MC		 160static volatile uint64_t attribute_used __attribute__((aligned(8))) b5Dither;
161static volatile uint64_t attribute_used __attribute__((aligned(8))) g5Dither;
162static volatile uint64_t attribute_used __attribute__((aligned(8))) g6Dither;
163static volatile uint64_t attribute_used __attribute__((aligned(8))) r5Dither;
d8fa3c54
MN		 164
165static uint64_t __attribute__((aligned(8))) dither4[2]={
166 0x0103010301030103LL,
167 0x0200020002000200LL,};
168
169static uint64_t __attribute__((aligned(8))) dither8[2]={
170 0x0602060206020602LL,
171 0x0004000400040004LL,};
d604bab9
MN		 172
173static uint64_t __attribute__((aligned(8))) b16Mask= 0x001F001F001F001FLL;
db7a2e0d
MC		 174static uint64_t attribute_used __attribute__((aligned(8))) g16Mask= 0x07E007E007E007E0LL;
175static uint64_t attribute_used __attribute__((aligned(8))) r16Mask= 0xF800F800F800F800LL;
d604bab9 176static uint64_t __attribute__((aligned(8))) b15Mask= 0x001F001F001F001FLL;
db7a2e0d
MC		 177static uint64_t attribute_used __attribute__((aligned(8))) g15Mask= 0x03E003E003E003E0LL;
178static uint64_t attribute_used __attribute__((aligned(8))) r15Mask= 0x7C007C007C007C00LL;
d604bab9 179
db7a2e0d
MC		 180static uint64_t attribute_used __attribute__((aligned(8))) M24A= 0x00FF0000FF0000FFLL;
181static uint64_t attribute_used __attribute__((aligned(8))) M24B= 0xFF0000FF0000FF00LL;
182static uint64_t attribute_used __attribute__((aligned(8))) M24C= 0x0000FF0000FF0000LL;
99d2cb72 183
ac6a2e45 184#ifdef FAST_BGR2YV12
db7a2e0d
MC		 185static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000000210041000DULL;
186static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL;
187static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL;
ac6a2e45 188#else
db7a2e0d
MC		 189static const uint64_t bgr2YCoeff attribute_used __attribute__((aligned(8))) = 0x000020E540830C8BULL;
190static const uint64_t bgr2UCoeff attribute_used __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL;
191static const uint64_t bgr2VCoeff attribute_used __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL;
69796008 192#endif /* FAST_BGR2YV12 */
db7a2e0d
MC		 193static const uint64_t bgr2YOffset attribute_used __attribute__((aligned(8))) = 0x1010101010101010ULL;
194static const uint64_t bgr2UVOffset attribute_used __attribute__((aligned(8)))= 0x8080808080808080ULL;
195static const uint64_t w1111 attribute_used __attribute__((aligned(8))) = 0x0001000100010001ULL;
3d6a30d9 196#endif /* defined(ARCH_X86) */
783e9cc9
MN		 197
198// clipping helper table for C implementations:
199static unsigned char clip_table[768];
200
d4e24275 201static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
6a4970ab 202
cf7d1c1a
MN		 203extern const uint8_t dither_2x2_4[2][8];
204extern const uint8_t dither_2x2_8[2][8];
205extern const uint8_t dither_8x8_32[8][8];
206extern const uint8_t dither_8x8_73[8][8];
207extern const uint8_t dither_8x8_220[8][8];
5cebb24b 208
2d529db5
LA		 209static const char * sws_context_to_name(void * ptr) {
210 return "swscaler";
211}
212
213static AVClass sws_context_class = { "SWScaler", sws_context_to_name, NULL };
214
e9e12f0e 215char *sws_format_name(enum PixelFormat format)
94c4def2 216{
e9e12f0e
LA		 217 switch (format) {
218 case PIX_FMT_YUV420P:
219 return "yuv420p";
220 case PIX_FMT_YUYV422:
221 return "yuyv422";
222 case PIX_FMT_RGB24:
223 return "rgb24";
224 case PIX_FMT_BGR24:
225 return "bgr24";
226 case PIX_FMT_YUV422P:
227 return "yuv422p";
228 case PIX_FMT_YUV444P:
229 return "yuv444p";
230 case PIX_FMT_RGB32:
231 return "rgb32";
232 case PIX_FMT_YUV410P:
233 return "yuv410p";
234 case PIX_FMT_YUV411P:
235 return "yuv411p";
236 case PIX_FMT_RGB565:
237 return "rgb565";
238 case PIX_FMT_RGB555:
239 return "rgb555";
4884b9e5
KS		 240 case PIX_FMT_GRAY16BE:
241 return "gray16be";
242 case PIX_FMT_GRAY16LE:
243 return "gray16le";
e9e12f0e
LA		 244 case PIX_FMT_GRAY8:
245 return "gray8";
246 case PIX_FMT_MONOWHITE:
247 return "mono white";
248 case PIX_FMT_MONOBLACK:
249 return "mono black";
250 case PIX_FMT_PAL8:
251 return "Palette";
252 case PIX_FMT_YUVJ420P:
253 return "yuvj420p";
254 case PIX_FMT_YUVJ422P:
255 return "yuvj422p";
256 case PIX_FMT_YUVJ444P:
257 return "yuvj444p";
258 case PIX_FMT_XVMC_MPEG2_MC:
259 return "xvmc_mpeg2_mc";
260 case PIX_FMT_XVMC_MPEG2_IDCT:
261 return "xvmc_mpeg2_idct";
262 case PIX_FMT_UYVY422:
263 return "uyvy422";
264 case PIX_FMT_UYYVYY411:
265 return "uyyvyy411";
266 case PIX_FMT_RGB32_1:
267 return "rgb32x";
268 case PIX_FMT_BGR32_1:
269 return "bgr32x";
270 case PIX_FMT_BGR32:
271 return "bgr32";
272 case PIX_FMT_BGR565:
273 return "bgr565";
274 case PIX_FMT_BGR555:
275 return "bgr555";
276 case PIX_FMT_BGR8:
277 return "bgr8";
278 case PIX_FMT_BGR4:
279 return "bgr4";
280 case PIX_FMT_BGR4_BYTE:
281 return "bgr4 byte";
282 case PIX_FMT_RGB8:
283 return "rgb8";
284 case PIX_FMT_RGB4:
285 return "rgb4";
286 case PIX_FMT_RGB4_BYTE:
287 return "rgb4 byte";
288 case PIX_FMT_NV12:
289 return "nv12";
290 case PIX_FMT_NV21:
291 return "nv21";
292 default:
293 return "Unknown format";
294 }
94c4def2
LA		 295}
296
9bde778e 297#if defined(ARCH_X86) && defined (CONFIG_GPL)
96034638
MN		 298void in_asm_used_var_warning_killer()
299{
20380eb8 300 volatile int i= bF8+bFC+w10+
5ac80202 301 bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+
6ff0ad6b 302 M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101;
96034638
MN		 303 if(i) i=0;
304}
305#endif
d604bab9 306
5859233b 307static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
e3d2500f 308 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
5859233b 309 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW)
e3d2500f
MN		 310{
311 //FIXME Optimize (just quickly writen not opti..)
312 int i;
5859233b 313 for(i=0; i<dstW; i++)
e3d2500f 314 {
379a2036 315 int val=1<<18;
e3d2500f
MN		 316 int j;
317 for(j=0; j<lumFilterSize; j++)
318 val += lumSrc[j][i] * lumFilter[j];
319
adcec46a 320 dest[i]= av_clip_uint8(val>>19);
e3d2500f
MN		 321 }
322
323 if(uDest != NULL)
5859233b 324 for(i=0; i<chrDstW; i++)
e3d2500f 325 {
379a2036
MN		 326 int u=1<<18;
327 int v=1<<18;
e3d2500f 328 int j;
627690b5 329 for(j=0; j<chrFilterSize; j++)
e3d2500f
MN		 330 {
331 u += chrSrc[j][i] * chrFilter[j];
332 v += chrSrc[j][i + 2048] * chrFilter[j];
333 }
334
adcec46a
RD		 335 uDest[i]= av_clip_uint8(u>>19);
336 vDest[i]= av_clip_uint8(v>>19);
e3d2500f
MN		 337 }
338}
339
6118e52e
VS		 340static inline void yuv2nv12XinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
341 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
342 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
343{
344 //FIXME Optimize (just quickly writen not opti..)
345 int i;
346 for(i=0; i<dstW; i++)
347 {
348 int val=1<<18;
349 int j;
350 for(j=0; j<lumFilterSize; j++)
351 val += lumSrc[j][i] * lumFilter[j];
352
adcec46a 353 dest[i]= av_clip_uint8(val>>19);
6118e52e
VS		 354 }
355
356 if(uDest == NULL)
357 return;
358
e9e12f0e 359 if(dstFormat == PIX_FMT_NV12)
6118e52e
VS		 360 for(i=0; i<chrDstW; i++)
361 {
362 int u=1<<18;
363 int v=1<<18;
364 int j;
365 for(j=0; j<chrFilterSize; j++)
366 {
367 u += chrSrc[j][i] * chrFilter[j];
368 v += chrSrc[j][i + 2048] * chrFilter[j];
369 }
370
adcec46a
RD		 371 uDest[2*i]= av_clip_uint8(u>>19);
372 uDest[2*i+1]= av_clip_uint8(v>>19);
6118e52e
VS		 373 }
374 else
375 for(i=0; i<chrDstW; i++)
376 {
377 int u=1<<18;
378 int v=1<<18;
379 int j;
380 for(j=0; j<chrFilterSize; j++)
381 {
382 u += chrSrc[j][i] * chrFilter[j];
383 v += chrSrc[j][i + 2048] * chrFilter[j];
384 }
385
adcec46a
RD		 386 uDest[2*i]= av_clip_uint8(v>>19);
387 uDest[2*i+1]= av_clip_uint8(u>>19);
6118e52e
VS		 388 }
389}
46de8b73 390
25593e29 391#define YSCALE_YUV_2_PACKEDX_C(type) \
cf7d1c1a
MN		 392 for(i=0; i<(dstW>>1); i++){\
393 int j;\
379a2036
MN		 394 int Y1=1<<18;\
395 int Y2=1<<18;\
396 int U=1<<18;\
397 int V=1<<18;\
789237ca 398 type attribute_unused *r, *b, *g;\
cf7d1c1a
MN		 399 const int i2= 2*i;\
400 \
401 for(j=0; j<lumFilterSize; j++)\
402 {\
403 Y1 += lumSrc[j][i2] * lumFilter[j];\
404 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
405 }\
406 for(j=0; j<chrFilterSize; j++)\
407 {\
408 U += chrSrc[j][i] * chrFilter[j];\
409 V += chrSrc[j][i+2048] * chrFilter[j];\
410 }\
411 Y1>>=19;\
412 Y2>>=19;\
413 U >>=19;\
414 V >>=19;\
415 if((Y1|Y2|U|V)&256)\
416 {\
417 if(Y1>255) Y1=255;\
418 else if(Y1<0)Y1=0;\
419 if(Y2>255) Y2=255;\
420 else if(Y2<0)Y2=0;\
421 if(U>255) U=255;\
422 else if(U<0) U=0;\
423 if(V>255) V=255;\
424 else if(V<0) V=0;\
46de8b73 425 }
6a4970ab 426
46de8b73 427#define YSCALE_YUV_2_RGBX_C(type) \
25593e29 428 YSCALE_YUV_2_PACKEDX_C(type)\
628bc5e8
LA		 429 r = (type *)c->table_rV[V];\
430 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
431 b = (type *)c->table_bU[U];\
cf7d1c1a 432
25593e29 433#define YSCALE_YUV_2_PACKED2_C \
cf7d1c1a
MN		 434 for(i=0; i<(dstW>>1); i++){\
435 const int i2= 2*i;\
436 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19;\
437 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;\
438 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19;\
439 int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\
46de8b73
MN		 440
441#define YSCALE_YUV_2_RGB2_C(type) \
25593e29 442 YSCALE_YUV_2_PACKED2_C\
cf7d1c1a 443 type *r, *b, *g;\
628bc5e8
LA		 444 r = (type *)c->table_rV[V];\
445 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
446 b = (type *)c->table_bU[U];\
cf7d1c1a 447
25593e29 448#define YSCALE_YUV_2_PACKED1_C \
cf7d1c1a
MN		 449 for(i=0; i<(dstW>>1); i++){\
450 const int i2= 2*i;\
451 int Y1= buf0[i2 ]>>7;\
452 int Y2= buf0[i2+1]>>7;\
453 int U= (uvbuf1[i ])>>7;\
454 int V= (uvbuf1[i+2048])>>7;\
46de8b73
MN		 455
456#define YSCALE_YUV_2_RGB1_C(type) \
25593e29 457 YSCALE_YUV_2_PACKED1_C\
cf7d1c1a 458 type *r, *b, *g;\
628bc5e8
LA		 459 r = (type *)c->table_rV[V];\
460 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
461 b = (type *)c->table_bU[U];\
cf7d1c1a 462
25593e29 463#define YSCALE_YUV_2_PACKED1B_C \
cf7d1c1a
MN		 464 for(i=0; i<(dstW>>1); i++){\
465 const int i2= 2*i;\
466 int Y1= buf0[i2 ]>>7;\
467 int Y2= buf0[i2+1]>>7;\
468 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
469 int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\
46de8b73
MN		 470
471#define YSCALE_YUV_2_RGB1B_C(type) \
25593e29 472 YSCALE_YUV_2_PACKED1B_C\
cf7d1c1a 473 type *r, *b, *g;\
628bc5e8
LA		 474 r = (type *)c->table_rV[V];\
475 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
476 b = (type *)c->table_bU[U];\
cf7d1c1a 477
46de8b73 478#define YSCALE_YUV_2_ANYRGB_C(func, func2)\
cf7d1c1a
MN		 479 switch(c->dstFormat)\
480 {\
e9e12f0e
LA		 481 case PIX_FMT_RGB32:\
482 case PIX_FMT_BGR32:\
cf7d1c1a
MN		 483 func(uint32_t)\
484 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
485 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
486 } \
487 break;\
e9e12f0e 488 case PIX_FMT_RGB24:\
cf7d1c1a
MN		 489 func(uint8_t)\
490 ((uint8_t*)dest)[0]= r[Y1];\
491 ((uint8_t*)dest)[1]= g[Y1];\
492 ((uint8_t*)dest)[2]= b[Y1];\
493 ((uint8_t*)dest)[3]= r[Y2];\
494 ((uint8_t*)dest)[4]= g[Y2];\
495 ((uint8_t*)dest)[5]= b[Y2];\
ae4cffd9 496 dest+=6;\
cf7d1c1a
MN		 497 }\
498 break;\
e9e12f0e 499 case PIX_FMT_BGR24:\
cf7d1c1a
MN		 500 func(uint8_t)\
501 ((uint8_t*)dest)[0]= b[Y1];\
502 ((uint8_t*)dest)[1]= g[Y1];\
503 ((uint8_t*)dest)[2]= r[Y1];\
504 ((uint8_t*)dest)[3]= b[Y2];\
505 ((uint8_t*)dest)[4]= g[Y2];\
506 ((uint8_t*)dest)[5]= r[Y2];\
ae4cffd9 507 dest+=6;\
cf7d1c1a
MN		 508 }\
509 break;\
e9e12f0e
LA		 510 case PIX_FMT_RGB565:\
511 case PIX_FMT_BGR565:\
cf7d1c1a
MN		 512 {\
513 const int dr1= dither_2x2_8[y&1 ][0];\
514 const int dg1= dither_2x2_4[y&1 ][0];\
515 const int db1= dither_2x2_8[(y&1)^1][0];\
516 const int dr2= dither_2x2_8[y&1 ][1];\
517 const int dg2= dither_2x2_4[y&1 ][1];\
518 const int db2= dither_2x2_8[(y&1)^1][1];\
519 func(uint16_t)\
520 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
521 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
522 }\
523 }\
524 break;\
e9e12f0e
LA		 525 case PIX_FMT_RGB555:\
526 case PIX_FMT_BGR555:\
cf7d1c1a
MN		 527 {\
528 const int dr1= dither_2x2_8[y&1 ][0];\
529 const int dg1= dither_2x2_8[y&1 ][1];\
530 const int db1= dither_2x2_8[(y&1)^1][0];\
531 const int dr2= dither_2x2_8[y&1 ][1];\
532 const int dg2= dither_2x2_8[y&1 ][0];\
533 const int db2= dither_2x2_8[(y&1)^1][1];\
534 func(uint16_t)\
535 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
536 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
537 }\
538 }\
539 break;\
e9e12f0e
LA		 540 case PIX_FMT_RGB8:\
541 case PIX_FMT_BGR8:\
cf7d1c1a
MN		 542 {\
543 const uint8_t * const d64= dither_8x8_73[y&7];\
544 const uint8_t * const d32= dither_8x8_32[y&7];\
545 func(uint8_t)\
546 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
547 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
548 }\
549 }\
550 break;\
e9e12f0e
LA		 551 case PIX_FMT_RGB4:\
552 case PIX_FMT_BGR4:\
cf7d1c1a
MN		 553 {\
554 const uint8_t * const d64= dither_8x8_73 [y&7];\
555 const uint8_t * const d128=dither_8x8_220[y&7];\
556 func(uint8_t)\
799fd467 557 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
f17457ac
MN		 558 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
559 }\
560 }\
561 break;\
e9e12f0e
LA		 562 case PIX_FMT_RGB4_BYTE:\
563 case PIX_FMT_BGR4_BYTE:\
f17457ac
MN		 564 {\
565 const uint8_t * const d64= dither_8x8_73 [y&7];\
566 const uint8_t * const d128=dither_8x8_220[y&7];\
567 func(uint8_t)\
cf7d1c1a
MN		 568 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
569 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
570 }\
571 }\
572 break;\
e9e12f0e 573 case PIX_FMT_MONOBLACK:\
cf7d1c1a
MN		 574 {\
575 const uint8_t * const d128=dither_8x8_220[y&7];\
576 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
577 for(i=0; i<dstW-7; i+=8){\
578 int acc;\
579 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
580 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
581 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
582 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
583 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
584 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
585 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
586 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
587 ((uint8_t*)dest)[0]= acc;\
ae4cffd9 588 dest++;\
cf7d1c1a
MN		 589 }\
590\
591/*\
592((uint8_t*)dest)-= dstW>>4;\
593{\
594 int acc=0;\
595 int left=0;\
596 static int top[1024];\
597 static int last_new[1024][1024];\
598 static int last_in3[1024][1024];\
599 static int drift[1024][1024];\
600 int topLeft=0;\
601 int shift=0;\
602 int count=0;\
603 const uint8_t * const d128=dither_8x8_220[y&7];\
604 int error_new=0;\
605 int error_in3=0;\
606 int f=0;\
607 \
608 for(i=dstW>>1; i<dstW; i++){\
609 int in= ((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19);\
610 int in2 = (76309 * (in - 16) + 32768) >> 16;\
611 int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\
612 int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\
613 + (last_new[y][i] - in3)*f/256;\
614 int new= old> 128 ? 255 : 0;\
615\
05749630
DB		 616 error_new+= FFABS(last_new[y][i] - new);\
617 error_in3+= FFABS(last_in3[y][i] - in3);\
cf7d1c1a
MN		 618 f= error_new - error_in3*4;\
619 if(f<0) f=0;\
620 if(f>256) f=256;\
621\
622 topLeft= top[i];\
623 left= top[i]= old - new;\
624 last_new[y][i]= new;\
625 last_in3[y][i]= in3;\
626\
627 acc+= acc + (new&1);\
628 if((i&7)==6){\
629 ((uint8_t*)dest)[0]= acc;\
630 ((uint8_t*)dest)++;\
631 }\
632 }\
633}\
634*/\
635 }\
636 break;\
e9e12f0e 637 case PIX_FMT_YUYV422:\
46de8b73
MN		 638 func2\
639 ((uint8_t*)dest)[2*i2+0]= Y1;\
640 ((uint8_t*)dest)[2*i2+1]= U;\
641 ((uint8_t*)dest)[2*i2+2]= Y2;\
642 ((uint8_t*)dest)[2*i2+3]= V;\
643 } \
644 break;\
e9e12f0e 645 case PIX_FMT_UYVY422:\
caeaabe7
AB		 646 func2\
647 ((uint8_t*)dest)[2*i2+0]= U;\
648 ((uint8_t*)dest)[2*i2+1]= Y1;\
649 ((uint8_t*)dest)[2*i2+2]= V;\
650 ((uint8_t*)dest)[2*i2+3]= Y2;\
651 } \
652 break;\
cf7d1c1a
MN		 653 }\
654
655
25593e29 656static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
e3d2500f 657 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
cf7d1c1a 658 uint8_t *dest, int dstW, int y)
e3d2500f 659{
cf7d1c1a
MN		 660 int i;
661 switch(c->dstFormat)
e3d2500f 662 {
e9e12f0e
LA		 663 case PIX_FMT_BGR32:
664 case PIX_FMT_RGB32:
cf7d1c1a
MN		 665 YSCALE_YUV_2_RGBX_C(uint32_t)
666 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];
667 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];
e3d2500f 668 }
cf7d1c1a 669 break;
e9e12f0e 670 case PIX_FMT_RGB24:
cf7d1c1a
MN		 671 YSCALE_YUV_2_RGBX_C(uint8_t)
672 ((uint8_t*)dest)[0]= r[Y1];
673 ((uint8_t*)dest)[1]= g[Y1];
674 ((uint8_t*)dest)[2]= b[Y1];
675 ((uint8_t*)dest)[3]= r[Y2];
676 ((uint8_t*)dest)[4]= g[Y2];
677 ((uint8_t*)dest)[5]= b[Y2];
ae4cffd9 678 dest+=6;
cf7d1c1a
MN		 679 }
680 break;
e9e12f0e 681 case PIX_FMT_BGR24:
cf7d1c1a
MN		 682 YSCALE_YUV_2_RGBX_C(uint8_t)
683 ((uint8_t*)dest)[0]= b[Y1];
684 ((uint8_t*)dest)[1]= g[Y1];
685 ((uint8_t*)dest)[2]= r[Y1];
686 ((uint8_t*)dest)[3]= b[Y2];
687 ((uint8_t*)dest)[4]= g[Y2];
688 ((uint8_t*)dest)[5]= r[Y2];
ae4cffd9 689 dest+=6;
cf7d1c1a
MN		 690 }
691 break;
e9e12f0e
LA		 692 case PIX_FMT_RGB565:
693 case PIX_FMT_BGR565:
cf7d1c1a
MN		 694 {
695 const int dr1= dither_2x2_8[y&1 ][0];
696 const int dg1= dither_2x2_4[y&1 ][0];
697 const int db1= dither_2x2_8[(y&1)^1][0];
698 const int dr2= dither_2x2_8[y&1 ][1];
699 const int dg2= dither_2x2_4[y&1 ][1];
700 const int db2= dither_2x2_8[(y&1)^1][1];
701 YSCALE_YUV_2_RGBX_C(uint16_t)
702 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
703 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
e3d2500f 704 }
e3d2500f 705 }
cf7d1c1a 706 break;
e9e12f0e
LA		 707 case PIX_FMT_RGB555:
708 case PIX_FMT_BGR555:
cf7d1c1a
MN		 709 {
710 const int dr1= dither_2x2_8[y&1 ][0];
711 const int dg1= dither_2x2_8[y&1 ][1];
712 const int db1= dither_2x2_8[(y&1)^1][0];
713 const int dr2= dither_2x2_8[y&1 ][1];
714 const int dg2= dither_2x2_8[y&1 ][0];
715 const int db2= dither_2x2_8[(y&1)^1][1];
716 YSCALE_YUV_2_RGBX_C(uint16_t)
717 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
718 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
e3d2500f 719 }
cf7d1c1a
MN		 720 }
721 break;
e9e12f0e
LA		 722 case PIX_FMT_RGB8:
723 case PIX_FMT_BGR8:
cf7d1c1a
MN		 724 {
725 const uint8_t * const d64= dither_8x8_73[y&7];
726 const uint8_t * const d32= dither_8x8_32[y&7];
727 YSCALE_YUV_2_RGBX_C(uint8_t)
728 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];
729 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];
e3d2500f 730 }
e3d2500f 731 }
cf7d1c1a 732 break;
e9e12f0e
LA		 733 case PIX_FMT_RGB4:
734 case PIX_FMT_BGR4:
cf7d1c1a
MN		 735 {
736 const uint8_t * const d64= dither_8x8_73 [y&7];
737 const uint8_t * const d128=dither_8x8_220[y&7];
738 YSCALE_YUV_2_RGBX_C(uint8_t)
799fd467 739 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]
f17457ac
MN		 740 +((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);
741 }
742 }
743 break;
e9e12f0e
LA		 744 case PIX_FMT_RGB4_BYTE:
745 case PIX_FMT_BGR4_BYTE:
f17457ac
MN		 746 {
747 const uint8_t * const d64= dither_8x8_73 [y&7];
748 const uint8_t * const d128=dither_8x8_220[y&7];
749 YSCALE_YUV_2_RGBX_C(uint8_t)
cf7d1c1a
MN		 750 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];
751 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];
e3d2500f 752 }
cf7d1c1a
MN		 753 }
754 break;
e9e12f0e 755 case PIX_FMT_MONOBLACK:
cf7d1c1a
MN		 756 {
757 const uint8_t * const d128=dither_8x8_220[y&7];
758 uint8_t *g= c->table_gU[128] + c->table_gV[128];
759 int acc=0;
760 for(i=0; i<dstW-1; i+=2){
761 int j;
379a2036
MN		 762 int Y1=1<<18;
763 int Y2=1<<18;
cf7d1c1a
MN		 764
765 for(j=0; j<lumFilterSize; j++)
766 {
767 Y1 += lumSrc[j][i] * lumFilter[j];
768 Y2 += lumSrc[j][i+1] * lumFilter[j];
769 }
770 Y1>>=19;
771 Y2>>=19;
772 if((Y1|Y2)&256)
773 {
774 if(Y1>255) Y1=255;
775 else if(Y1<0)Y1=0;
776 if(Y2>255) Y2=255;
777 else if(Y2<0)Y2=0;
778 }
779 acc+= acc + g[Y1+d128[(i+0)&7]];
780 acc+= acc + g[Y2+d128[(i+1)&7]];
781 if((i&7)==6){
782 ((uint8_t*)dest)[0]= acc;
ae4cffd9 783 dest++;
cf7d1c1a 784 }
e3d2500f 785 }
e3d2500f 786 }
cf7d1c1a 787 break;
e9e12f0e 788 case PIX_FMT_YUYV422:
25593e29 789 YSCALE_YUV_2_PACKEDX_C(void)
46de8b73
MN		 790 ((uint8_t*)dest)[2*i2+0]= Y1;
791 ((uint8_t*)dest)[2*i2+1]= U;
792 ((uint8_t*)dest)[2*i2+2]= Y2;
793 ((uint8_t*)dest)[2*i2+3]= V;
794 }
795 break;
e9e12f0e 796 case PIX_FMT_UYVY422:
caeaabe7
AB		 797 YSCALE_YUV_2_PACKEDX_C(void)
798 ((uint8_t*)dest)[2*i2+0]= U;
799 ((uint8_t*)dest)[2*i2+1]= Y1;
800 ((uint8_t*)dest)[2*i2+2]= V;
801 ((uint8_t*)dest)[2*i2+3]= Y2;
802 }
803 break;
e3d2500f
MN		 804 }
805}
806
807
7630f2e0
MN		 808//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
809//Plain C versions
9bde778e 810#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) || !defined(CONFIG_GPL)
726a959a
MN		 811#define COMPILE_C
812#endif
813
a2faa401 814#ifdef ARCH_POWERPC
9bde778e 815#if (defined (HAVE_ALTIVEC) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
a2faa401
RD		 816#define COMPILE_ALTIVEC
817#endif //HAVE_ALTIVEC
818#endif //ARCH_POWERPC
819
3d6a30d9 820#if defined(ARCH_X86)
726a959a 821
9bde778e 822#if ((defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
726a959a
MN		 823#define COMPILE_MMX
824#endif
825
9bde778e 826#if (defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
726a959a
MN		 827#define COMPILE_MMX2
828#endif
829
9bde778e 830#if ((defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)) && defined (CONFIG_GPL)
726a959a
MN		 831#define COMPILE_3DNOW
832#endif
6e1c66bc 833#endif //ARCH_X86 || ARCH_X86_64
726a959a
MN		 834
835#undef HAVE_MMX
836#undef HAVE_MMX2
837#undef HAVE_3DNOW
726a959a
MN		 838
839#ifdef COMPILE_C
7630f2e0
MN		 840#undef HAVE_MMX
841#undef HAVE_MMX2
842#undef HAVE_3DNOW
a2faa401 843#undef HAVE_ALTIVEC
7630f2e0
MN		 844#define RENAME(a) a ## _C
845#include "swscale_template.c"
726a959a 846#endif
397c035e 847
a2faa401
RD		 848#ifdef ARCH_POWERPC
849#ifdef COMPILE_ALTIVEC
850#undef RENAME
851#define HAVE_ALTIVEC
852#define RENAME(a) a ## _altivec
853#include "swscale_template.c"
854#endif
855#endif //ARCH_POWERPC
856
3d6a30d9 857#if defined(ARCH_X86)
397c035e 858
7630f2e0
MN		 859//X86 versions
860/*
861#undef RENAME
862#undef HAVE_MMX
863#undef HAVE_MMX2
864#undef HAVE_3DNOW
865#define ARCH_X86
866#define RENAME(a) a ## _X86
867#include "swscale_template.c"
1faf0867 868*/
7630f2e0 869//MMX versions
726a959a 870#ifdef COMPILE_MMX
7630f2e0
MN		 871#undef RENAME
872#define HAVE_MMX
873#undef HAVE_MMX2
874#undef HAVE_3DNOW
7630f2e0
MN		 875#define RENAME(a) a ## _MMX
876#include "swscale_template.c"
726a959a 877#endif
7630f2e0
MN		 878
879//MMX2 versions
726a959a 880#ifdef COMPILE_MMX2
7630f2e0
MN		 881#undef RENAME
882#define HAVE_MMX
883#define HAVE_MMX2
884#undef HAVE_3DNOW
7630f2e0
MN		 885#define RENAME(a) a ## _MMX2
886#include "swscale_template.c"
726a959a 887#endif
7630f2e0
MN		 888
889//3DNOW versions
726a959a 890#ifdef COMPILE_3DNOW
7630f2e0
MN		 891#undef RENAME
892#define HAVE_MMX
893#undef HAVE_MMX2
894#define HAVE_3DNOW
7630f2e0
MN		 895#define RENAME(a) a ## _3DNow
896#include "swscale_template.c"
726a959a 897#endif
7630f2e0 898
6e1c66bc 899#endif //ARCH_X86 || ARCH_X86_64
7630f2e0 900
77a416e8 901// minor note: the HAVE_xyz is messed up after that line so don't use it
d604bab9 902
a86c461c
MN		 903static double getSplineCoeff(double a, double b, double c, double d, double dist)
904{
905// printf("%f %f %f %f %f\n", a,b,c,d,dist);
906 if(dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
6a4970ab 907 else return getSplineCoeff( 0.0,
a86c461c
MN		 908 b+ 2.0*c + 3.0*d,
909 c + 3.0*d,
910 -b- 3.0*c - 6.0*d,
911 dist-1.0);
912}
6c7506de 913
bca11e75 914static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
c7f822d9 915 int srcW, int dstW, int filterAlign, int one, int flags,
66d1cdb6 916 SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
28bf81c9
MN		 917{
918 int i;
c7f822d9
MN		 919 int filterSize;
920 int filter2Size;
921 int minFilterSize;
922 double *filter=NULL;
923 double *filter2=NULL;
3d6a30d9 924#if defined(ARCH_X86)
516b1f82 925 if(flags & SWS_CPU_CAPS_MMX)
28bf81c9 926 asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
726a959a 927#endif
31190492 928
adeaecb9 929 // Note the +1 is for the MMXscaler which reads over the end
8dfe9094 930 *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
6c7506de 931
05749630 932 if(FFABS(xInc - 0x10000) <10) // unscaled
28bf81c9
MN		 933 {
934 int i;
c7f822d9 935 filterSize= 1;
8dfe9094 936 filter= av_malloc(dstW*sizeof(double)*filterSize);
c7f822d9 937 for(i=0; i<dstW*filterSize; i++) filter[i]=0;
28bf81c9
MN		 938
939 for(i=0; i<dstW; i++)
940 {
c7f822d9
MN		 941 filter[i*filterSize]=1;
942 (*filterPos)[i]=i;
28bf81c9
MN		 943 }
944
945 }
ff7ba856
MN		 946 else if(flags&SWS_POINT) // lame looking point sampling mode
947 {
948 int i;
949 int xDstInSrc;
950 filterSize= 1;
8dfe9094 951 filter= av_malloc(dstW*sizeof(double)*filterSize);
6a4970ab 952
ff7ba856
MN		 953 xDstInSrc= xInc/2 - 0x8000;
954 for(i=0; i<dstW; i++)
955 {
8a01d20c 956 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
ff7ba856
MN		 957
958 (*filterPos)[i]= xx;
959 filter[i]= 1.0;
960 xDstInSrc+= xInc;
961 }
962 }
a86c461c 963 else if((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
28bf81c9
MN		 964 {
965 int i;
966 int xDstInSrc;
c7f822d9
MN		 967 if (flags&SWS_BICUBIC) filterSize= 4;
968 else if(flags&SWS_X ) filterSize= 4;
6a4970ab 969 else filterSize= 2; // SWS_BILINEAR / SWS_AREA
8dfe9094 970 filter= av_malloc(dstW*sizeof(double)*filterSize);
28bf81c9
MN		 971
972 xDstInSrc= xInc/2 - 0x8000;
973 for(i=0; i<dstW; i++)
974 {
8a01d20c 975 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
28bf81c9
MN		 976 int j;
977
c7f822d9 978 (*filterPos)[i]= xx;
d8863d37 979 //Bilinear upscale / linear interpolate / Area averaging
c7f822d9 980 for(j=0; j<filterSize; j++)
28bf81c9 981 {
05749630 982 double d= FFABS((xx<<16) - xDstInSrc)/(double)(1<<16);
28bf81c9
MN		 983 double coeff= 1.0 - d;
984 if(coeff<0) coeff=0;
c7f822d9 985 filter[i*filterSize + j]= coeff;
28bf81c9
MN		 986 xx++;
987 }
28bf81c9
MN		 988 xDstInSrc+= xInc;
989 }
990 }
a86c461c 991 else
28bf81c9 992 {
a86c461c
MN		 993 double xDstInSrc;
994 double sizeFactor, filterSizeInSrc;
995 const double xInc1= (double)xInc / (double)(1<<16);
a86c461c
MN		 996
997 if (flags&SWS_BICUBIC) sizeFactor= 4.0;
998 else if(flags&SWS_X) sizeFactor= 8.0;
999 else if(flags&SWS_AREA) sizeFactor= 1.0; //downscale only, for upscale it is bilinear
1000 else if(flags&SWS_GAUSS) sizeFactor= 8.0; // infinite ;)
66d1cdb6 1001 else if(flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? 2.0*param[0] : 6.0;
93768378 1002 else if(flags&SWS_SINC) sizeFactor= 20.0; // infinite ;)
a86c461c
MN		 1003 else if(flags&SWS_SPLINE) sizeFactor= 20.0; // infinite ;)
1004 else if(flags&SWS_BILINEAR) sizeFactor= 2.0;
93768378
MN		 1005 else {
1006 sizeFactor= 0.0; //GCC warning killer
1007 ASSERT(0)
1008 }
6a4970ab 1009
a86c461c
MN		 1010 if(xInc1 <= 1.0) filterSizeInSrc= sizeFactor; // upscale
1011 else filterSizeInSrc= sizeFactor*srcW / (double)dstW;
81b7c056 1012
a86c461c
MN		 1013 filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
1014 if(filterSize > srcW-2) filterSize=srcW-2;
28bf81c9 1015
8dfe9094 1016 filter= av_malloc(dstW*sizeof(double)*filterSize);
a86c461c
MN		 1017
1018 xDstInSrc= xInc1 / 2.0 - 0.5;
28bf81c9
MN		 1019 for(i=0; i<dstW; i++)
1020 {
a86c461c 1021 int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
28bf81c9 1022 int j;
c7f822d9
MN		 1023 (*filterPos)[i]= xx;
1024 for(j=0; j<filterSize; j++)
28bf81c9 1025 {
05749630 1026 double d= FFABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
28bf81c9 1027 double coeff;
a86c461c 1028 if(flags & SWS_BICUBIC)
28bf81c9 1029 {
66d1cdb6
MN		 1030 double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
1031 double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;
1032
6a4970ab 1033 if(d<1.0)
66d1cdb6 1034 coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
28bf81c9 1035 else if(d<2.0)
66d1cdb6 1036 coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
28bf81c9
MN		 1037 else
1038 coeff=0.0;
1039 }
a86c461c
MN		 1040/* else if(flags & SWS_X)
1041 {
1042 double p= param ? param*0.01 : 0.3;
1043 coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1044 coeff*= pow(2.0, - p*d*d);
1045 }*/
1046 else if(flags & SWS_X)
1047 {
66d1cdb6 1048 double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
6a4970ab 1049
a86c461c
MN		 1050 if(d<1.0)
1051 coeff = cos(d*PI);
1052 else
1053 coeff=-1.0;
1054 if(coeff<0.0) coeff= -pow(-coeff, A);
1055 else coeff= pow( coeff, A);
1056 coeff= coeff*0.5 + 0.5;
1057 }
d8863d37 1058 else if(flags & SWS_AREA)
28bf81c9 1059 {
a86c461c 1060 double srcPixelSize= 1.0/xInc1;
d8863d37
MN		 1061 if(d + srcPixelSize/2 < 0.5) coeff= 1.0;
1062 else if(d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5;
1063 else coeff=0.0;
1064 }
a86c461c
MN		 1065 else if(flags & SWS_GAUSS)
1066 {
66d1cdb6 1067 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
a86c461c
MN		 1068 coeff = pow(2.0, - p*d*d);
1069 }
1070 else if(flags & SWS_SINC)
1071 {
1072 coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1073 }
1074 else if(flags & SWS_LANCZOS)
1075 {
6a4970ab 1076 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
a86c461c
MN		 1077 coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
1078 if(d>p) coeff=0;
1079 }
1080 else if(flags & SWS_BILINEAR)
28bf81c9
MN		 1081 {
1082 coeff= 1.0 - d;
1083 if(coeff<0) coeff=0;
1084 }
a86c461c
MN		 1085 else if(flags & SWS_SPLINE)
1086 {
1087 double p=-2.196152422706632;
1088 coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, d);
1089 }
93768378
MN		 1090 else {
1091 coeff= 0.0; //GCC warning killer
1092 ASSERT(0)
1093 }
a86c461c 1094
c7f822d9 1095 filter[i*filterSize + j]= coeff;
28bf81c9
MN		 1096 xx++;
1097 }
a86c461c 1098 xDstInSrc+= xInc1;
28bf81c9
MN		 1099 }
1100 }
1101
c7f822d9 1102 /* apply src & dst Filter to filter -> filter2
94c4def2 1103 av_free(filter);
c7f822d9 1104 */
81b7c056 1105 ASSERT(filterSize>0)
c7f822d9
MN		 1106 filter2Size= filterSize;
1107 if(srcFilter) filter2Size+= srcFilter->length - 1;
1108 if(dstFilter) filter2Size+= dstFilter->length - 1;
81b7c056 1109 ASSERT(filter2Size>0)
8dfe9094 1110 filter2= av_malloc(filter2Size*dstW*sizeof(double));
c7f822d9
MN		 1111
1112 for(i=0; i<dstW; i++)
1113 {
1114 int j;
1115 SwsVector scaleFilter;
1116 SwsVector *outVec;
1117
1118 scaleFilter.coeff= filter + i*filterSize;
1119 scaleFilter.length= filterSize;
1120
d4e24275 1121 if(srcFilter) outVec= sws_getConvVec(srcFilter, &scaleFilter);
c7f822d9
MN		 1122 else outVec= &scaleFilter;
1123
1124 ASSERT(outVec->length == filter2Size)
1125 //FIXME dstFilter
1126
1127 for(j=0; j<outVec->length; j++)
1128 {
1129 filter2[i*filter2Size + j]= outVec->coeff[j];
1130 }
1131
1132 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1133
d4e24275 1134 if(outVec != &scaleFilter) sws_freeVec(outVec);
c7f822d9 1135 }
94c4def2 1136 av_free(filter); filter=NULL;
c7f822d9
MN		 1137
1138 /* try to reduce the filter-size (step1 find size and shift left) */
1139 // Assume its near normalized (*0.5 or *2.0 is ok but * 0.001 is not)
1140 minFilterSize= 0;
1141 for(i=dstW-1; i>=0; i--)
1142 {
1143 int min= filter2Size;
1144 int j;
1145 double cutOff=0.0;
1146
1147 /* get rid off near zero elements on the left by shifting left */
1148 for(j=0; j<filter2Size; j++)
1149 {
1150 int k;
05749630 1151 cutOff += FFABS(filter2[i*filter2Size]);
c7f822d9
MN		 1152
1153 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;
1154
77a416e8 1155 /* preserve Monotonicity because the core can't handle the filter otherwise */
c7f822d9
MN		 1156 if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1157
1158 // Move filter coeffs left
1159 for(k=1; k<filter2Size; k++)
1160 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1161 filter2[i*filter2Size + k - 1]= 0.0;
1162 (*filterPos)[i]++;
1163 }
1164
1165 cutOff=0.0;
1166 /* count near zeros on the right */
1167 for(j=filter2Size-1; j>0; j--)
1168 {
05749630 1169 cutOff += FFABS(filter2[i*filter2Size + j]);
c7f822d9
MN		 1170
1171 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break;
1172 min--;
1173 }
1174
1175 if(min>minFilterSize) minFilterSize= min;
1176 }
1177
8c266f0c
RD		 1178 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1179 // we can handle the special case 4,
1180 // so we don't want to go to the full 8
1181 if (minFilterSize < 5)
1182 filterAlign = 4;
1183
1184 // we really don't want to waste our time
1185 // doing useless computation, so fall-back on
1186 // the scalar C code for very small filter.
1187 // vectorizing is worth it only if you have
1188 // decent-sized vector.
1189 if (minFilterSize < 3)
1190 filterAlign = 1;
1191 }
1192
bca11e75
MN		 1193 if (flags & SWS_CPU_CAPS_MMX) {
1194 // special case for unscaled vertical filtering
1195 if(minFilterSize == 1 && filterAlign == 2)
1196 filterAlign= 1;
1197 }
1198
81b7c056 1199 ASSERT(minFilterSize > 0)
6c7506de 1200 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
81b7c056 1201 ASSERT(filterSize > 0)
8dfe9094 1202 filter= av_malloc(filterSize*dstW*sizeof(double));
bca11e75
MN		 1203 if(filterSize >= MAX_FILTER_SIZE)
1204 return -1;
6c7506de
MN		 1205 *outFilterSize= filterSize;
1206
4a53a912 1207 if(flags&SWS_PRINT_INFO)
2d529db5 1208 av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
c7f822d9
MN		 1209 /* try to reduce the filter-size (step2 reduce it) */
1210 for(i=0; i<dstW; i++)
1211 {
1212 int j;
1213
6c7506de
MN		 1214 for(j=0; j<filterSize; j++)
1215 {
1216 if(j>=filter2Size) filter[i*filterSize + j]= 0.0;
1217 else filter[i*filterSize + j]= filter2[i*filter2Size + j];
1218 }
c7f822d9 1219 }
94c4def2 1220 av_free(filter2); filter2=NULL;
6a4970ab 1221
c7f822d9
MN		 1222
1223 //FIXME try to align filterpos if possible
1224
28bf81c9
MN		 1225 //fix borders
1226 for(i=0; i<dstW; i++)
1227 {
1228 int j;
c7f822d9 1229 if((*filterPos)[i] < 0)
28bf81c9
MN		 1230 {
1231 // Move filter coeffs left to compensate for filterPos
6c7506de 1232 for(j=1; j<filterSize; j++)
28bf81c9 1233 {
6f209cb5 1234 int left= FFMAX(j + (*filterPos)[i], 0);
6c7506de
MN		 1235 filter[i*filterSize + left] += filter[i*filterSize + j];
1236 filter[i*filterSize + j]=0;
28bf81c9 1237 }
c7f822d9 1238 (*filterPos)[i]= 0;
28bf81c9
MN		 1239 }
1240
6c7506de 1241 if((*filterPos)[i] + filterSize > srcW)
28bf81c9 1242 {
6c7506de 1243 int shift= (*filterPos)[i] + filterSize - srcW;
28bf81c9 1244 // Move filter coeffs right to compensate for filterPos
6c7506de 1245 for(j=filterSize-2; j>=0; j--)
28bf81c9 1246 {
6f209cb5 1247 int right= FFMIN(j + shift, filterSize-1);
6c7506de
MN		 1248 filter[i*filterSize +right] += filter[i*filterSize +j];
1249 filter[i*filterSize +j]=0;
28bf81c9 1250 }
6c7506de 1251 (*filterPos)[i]= srcW - filterSize;
28bf81c9
MN		 1252 }
1253 }
1254
6c7506de 1255 // Note the +1 is for the MMXscaler which reads over the end
5edb653b 1256 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
0dea6433 1257 *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
c7f822d9
MN		 1258
1259 /* Normalize & Store in outFilter */
28bf81c9
MN		 1260 for(i=0; i<dstW; i++)
1261 {
1262 int j;
ff490720 1263 double error=0;
28bf81c9
MN		 1264 double sum=0;
1265 double scale= one;
ff490720 1266
6c7506de 1267 for(j=0; j<filterSize; j++)
28bf81c9 1268 {
6c7506de 1269 sum+= filter[i*filterSize + j];
28bf81c9
MN		 1270 }
1271 scale/= sum;
93768378 1272 for(j=0; j<*outFilterSize; j++)
28bf81c9 1273 {
ff490720
MN		 1274 double v= filter[i*filterSize + j]*scale + error;
1275 int intV= floor(v + 0.5);
1276 (*outFilter)[i*(*outFilterSize) + j]= intV;
1277 error = v - intV;
28bf81c9
MN		 1278 }
1279 }
6a4970ab 1280
adeaecb9
MN		 1281 (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1282 for(i=0; i<*outFilterSize; i++)
1283 {
1284 int j= dstW*(*outFilterSize);
1285 (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1286 }
c7f822d9 1287
94c4def2 1288 av_free(filter);
bca11e75 1289 return 0;
7630f2e0 1290}
31190492 1291
17c613ef 1292#ifdef COMPILE_MMX2
b7dc6f66 1293static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
28bf81c9 1294{
b7dc6f66 1295 uint8_t *fragmentA;
6e1c66bc
AJ		 1296 long imm8OfPShufW1A;
1297 long imm8OfPShufW2A;
1298 long fragmentLengthA;
b7dc6f66 1299 uint8_t *fragmentB;
6e1c66bc
AJ		 1300 long imm8OfPShufW1B;
1301 long imm8OfPShufW2B;
1302 long fragmentLengthB;
b7dc6f66 1303 int fragmentPos;
28bf81c9
MN		 1304
1305 int xpos, i;
1306
1307 // create an optimized horizontal scaling routine
1308
1309 //code fragment
1310
1311 asm volatile(
1312 "jmp 9f \n\t"
1313 // Begin
1314 "0: \n\t"
6a4970ab
DB		 1315 "movq (%%"REG_d", %%"REG_a"), %%mm3\n\t"
1316 "movd (%%"REG_c", %%"REG_S"), %%mm0\n\t"
6e1c66bc 1317 "movd 1(%%"REG_c", %%"REG_S"), %%mm1\n\t"
b7dc6f66
MN		 1318 "punpcklbw %%mm7, %%mm1 \n\t"
1319 "punpcklbw %%mm7, %%mm0 \n\t"
28bf81c9
MN		 1320 "pshufw $0xFF, %%mm1, %%mm1 \n\t"
1321 "1: \n\t"
28bf81c9
MN		 1322 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1323 "2: \n\t"
28bf81c9 1324 "psubw %%mm1, %%mm0 \n\t"
6d606c4f 1325 "movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"
28bf81c9 1326 "pmullw %%mm3, %%mm0 \n\t"
28bf81c9
MN		 1327 "psllw $7, %%mm1 \n\t"
1328 "paddw %%mm1, %%mm0 \n\t"
1329
6e1c66bc 1330 "movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"
28bf81c9 1331
6e1c66bc 1332 "add $8, %%"REG_a" \n\t"
28bf81c9
MN		 1333 // End
1334 "9: \n\t"
1335// "int $3\n\t"
6e1c66bc
AJ		 1336 "lea 0b, %0 \n\t"
1337 "lea 1b, %1 \n\t"
1338 "lea 2b, %2 \n\t"
1339 "dec %1 \n\t"
1340 "dec %2 \n\t"
1341 "sub %0, %1 \n\t"
1342 "sub %0, %2 \n\t"
1343 "lea 9b, %3 \n\t"
1344 "sub %0, %3 \n\t"
b7dc6f66
MN		 1345
1346
1347 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1348 "=r" (fragmentLengthA)
28bf81c9
MN		 1349 );
1350
b7dc6f66
MN		 1351 asm volatile(
1352 "jmp 9f \n\t"
1353 // Begin
1354 "0: \n\t"
6a4970ab
DB		 1355 "movq (%%"REG_d", %%"REG_a"), %%mm3\n\t"
1356 "movd (%%"REG_c", %%"REG_S"), %%mm0\n\t"
b7dc6f66
MN		 1357 "punpcklbw %%mm7, %%mm0 \n\t"
1358 "pshufw $0xFF, %%mm0, %%mm1 \n\t"
1359 "1: \n\t"
1360 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1361 "2: \n\t"
1362 "psubw %%mm1, %%mm0 \n\t"
6d606c4f 1363 "movl 8(%%"REG_b", %%"REG_a"), %%esi\n\t"
b7dc6f66
MN		 1364 "pmullw %%mm3, %%mm0 \n\t"
1365 "psllw $7, %%mm1 \n\t"
1366 "paddw %%mm1, %%mm0 \n\t"
1367
6e1c66bc 1368 "movq %%mm0, (%%"REG_D", %%"REG_a")\n\t"
28bf81c9 1369
6e1c66bc 1370 "add $8, %%"REG_a" \n\t"
b7dc6f66
MN		 1371 // End
1372 "9: \n\t"
1373// "int $3\n\t"
6e1c66bc
AJ		 1374 "lea 0b, %0 \n\t"
1375 "lea 1b, %1 \n\t"
1376 "lea 2b, %2 \n\t"
1377 "dec %1 \n\t"
1378 "dec %2 \n\t"
1379 "sub %0, %1 \n\t"
1380 "sub %0, %2 \n\t"
1381 "lea 9b, %3 \n\t"
1382 "sub %0, %3 \n\t"
b7dc6f66
MN		 1383
1384
1385 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1386 "=r" (fragmentLengthB)
1387 );
1388
1389 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1390 fragmentPos=0;
6a4970ab 1391
b7dc6f66 1392 for(i=0; i<dstW/numSplits; i++)
28bf81c9
MN		 1393 {
1394 int xx=xpos>>16;
1395
1396 if((i&3) == 0)
1397 {
1398 int a=0;
1399 int b=((xpos+xInc)>>16) - xx;
1400 int c=((xpos+xInc*2)>>16) - xx;
1401 int d=((xpos+xInc*3)>>16) - xx;
1402
b7dc6f66
MN		 1403 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9;
1404 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9;
1405 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1406 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1407 filterPos[i/2]= xx;
1408
1409 if(d+1<4)
1410 {
1411 int maxShift= 3-(d+1);
1412 int shift=0;
1413
1414 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1415
1416 funnyCode[fragmentPos + imm8OfPShufW1B]=
1417 (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1418 funnyCode[fragmentPos + imm8OfPShufW2B]=
1419 a | (b<<2) | (c<<4) | (d<<6);
1420
1421 if(i+3>=dstW) shift=maxShift; //avoid overread
1422 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1423
1424 if(shift && i>=shift)
1425 {
1426 funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1427 funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1428 filterPos[i/2]-=shift;
1429 }
1430
1431 fragmentPos+= fragmentLengthB;
1432 }
1433 else
1434 {
1435 int maxShift= 3-d;
1436 int shift=0;
1437
1438 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
28bf81c9 1439
b7dc6f66
MN		 1440 funnyCode[fragmentPos + imm8OfPShufW1A]=
1441 funnyCode[fragmentPos + imm8OfPShufW2A]=
1442 a | (b<<2) | (c<<4) | (d<<6);
28bf81c9 1443
b7dc6f66
MN		 1444 if(i+4>=dstW) shift=maxShift; //avoid overread
1445 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
28bf81c9 1446
b7dc6f66
MN		 1447 if(shift && i>=shift)
1448 {
1449 funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1450 funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1451 filterPos[i/2]-=shift;
1452 }
1453
1454 fragmentPos+= fragmentLengthA;
1455 }
1456
1457 funnyCode[fragmentPos]= RET;
28bf81c9
MN		 1458 }
1459 xpos+=xInc;
1460 }
b7dc6f66 1461 filterPos[i/2]= xpos>>16; // needed to jump to the next part
28bf81c9 1462}
17c613ef 1463#endif /* COMPILE_MMX2 */
28bf81c9 1464
9b2283cc 1465static void globalInit(void){
31190492
A		 1466 // generating tables:
1467 int i;
c1b0bfb4 1468 for(i=0; i<768; i++){
adcec46a 1469 int c= av_clip_uint8(i-256);
c1b0bfb4 1470 clip_table[i]=c;
b18ea156 1471 }
516b1f82 1472}
c1b0bfb4 1473
516b1f82 1474static SwsFunc getSwsFunc(int flags){
6a4970ab 1475
9bde778e 1476#if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)
3d6a30d9 1477#if defined(ARCH_X86)
28bf81c9 1478 // ordered per speed fasterst first
516b1f82
MN		 1479 if(flags & SWS_CPU_CAPS_MMX2)
1480 return swScale_MMX2;
1481 else if(flags & SWS_CPU_CAPS_3DNOW)
1482 return swScale_3DNow;
1483 else if(flags & SWS_CPU_CAPS_MMX)
1484 return swScale_MMX;
28bf81c9 1485 else
516b1f82 1486 return swScale_C;
28bf81c9
MN		 1487
1488#else
a2faa401
RD		 1489#ifdef ARCH_POWERPC
1490 if(flags & SWS_CPU_CAPS_ALTIVEC)
1491 return swScale_altivec;
1492 else
1493 return swScale_C;
1494#endif
516b1f82 1495 return swScale_C;
3d6a30d9 1496#endif /* defined(ARCH_X86) */
28bf81c9
MN		 1497#else //RUNTIME_CPUDETECT
1498#ifdef HAVE_MMX2
516b1f82 1499 return swScale_MMX2;
28bf81c9 1500#elif defined (HAVE_3DNOW)
516b1f82 1501 return swScale_3DNow;
28bf81c9 1502#elif defined (HAVE_MMX)
516b1f82 1503 return swScale_MMX;
a2faa401
RD		 1504#elif defined (HAVE_ALTIVEC)
1505 return swScale_altivec;
28bf81c9 1506#else
516b1f82 1507 return swScale_C;
28bf81c9
MN		 1508#endif
1509#endif //!RUNTIME_CPUDETECT
31190492 1510}
7630f2e0 1511
d4e24275 1512static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
0d9f3d85
A		 1513 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1514 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1515 /* Copy Y plane */
12e11662 1516 if(dstStride[0]==srcStride[0] && srcStride[0] > 0)
0d9f3d85
A		 1517 memcpy(dst, src[0], srcSliceH*dstStride[0]);
1518 else
1519 {
1520 int i;
1521 uint8_t *srcPtr= src[0];
1522 uint8_t *dstPtr= dst;
1523 for(i=0; i<srcSliceH; i++)
1524 {
6118e52e 1525 memcpy(dstPtr, srcPtr, c->srcW);
0d9f3d85
A		 1526 srcPtr+= srcStride[0];
1527 dstPtr+= dstStride[0];
1528 }
1529 }
6118e52e 1530 dst = dstParam[1] + dstStride[1]*srcSliceY/2;
e9e12f0e 1531 if (c->dstFormat == PIX_FMT_NV12)
6118e52e
VS		 1532 interleaveBytes( src[1],src[2],dst,c->srcW/2,srcSliceH/2,srcStride[1],srcStride[2],dstStride[0] );
1533 else
1534 interleaveBytes( src[2],src[1],dst,c->srcW/2,srcSliceH/2,srcStride[2],srcStride[1],dstStride[0] );
fccb9b2b 1535
d4e24275 1536 return srcSliceH;
0d9f3d85
A		 1537}
1538
d4e24275 1539static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
0d9f3d85
A		 1540 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1541 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1542
fccb9b2b
MN		 1543 yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
1544
d4e24275 1545 return srcSliceH;
0d9f3d85
A		 1546}
1547
caeaabe7
AB		 1548static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1549 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1550 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1551
1552 yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );
1553
1554 return srcSliceH;
1555}
1556
e09d12f4 1557/* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */
d4e24275 1558static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
e09d12f4
MN		 1559 int srcSliceH, uint8_t* dst[], int dstStride[]){
1560 const int srcFormat= c->srcFormat;
1561 const int dstFormat= c->dstFormat;
e9e12f0e
LA		 1562 const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
1563 const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
1564 const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1565 const int dstId= fmt_depth(dstFormat) >> 2;
73cbfb1c 1566 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
e09d12f4
MN		 1567
1568 /* BGR -> BGR */
20380eb8
MN		 1569 if( (isBGR(srcFormat) && isBGR(dstFormat))
1570 || (isRGB(srcFormat) && isRGB(dstFormat))){
e09d12f4
MN		 1571 switch(srcId | (dstId<<4)){
1572 case 0x34: conv= rgb16to15; break;
1573 case 0x36: conv= rgb24to15; break;
1574 case 0x38: conv= rgb32to15; break;
1575 case 0x43: conv= rgb15to16; break;
1576 case 0x46: conv= rgb24to16; break;
1577 case 0x48: conv= rgb32to16; break;
1578 case 0x63: conv= rgb15to24; break;
1579 case 0x64: conv= rgb16to24; break;
1580 case 0x68: conv= rgb32to24; break;
1581 case 0x83: conv= rgb15to32; break;
1582 case 0x84: conv= rgb16to32; break;
1583 case 0x86: conv= rgb24to32; break;
6a4970ab 1584 default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
94c4def2 1585 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
b935781b 1586 }
20380eb8
MN		 1587 }else if( (isBGR(srcFormat) && isRGB(dstFormat))
1588 || (isRGB(srcFormat) && isBGR(dstFormat))){
e09d12f4
MN		 1589 switch(srcId | (dstId<<4)){
1590 case 0x33: conv= rgb15tobgr15; break;
1591 case 0x34: conv= rgb16tobgr15; break;
1592 case 0x36: conv= rgb24tobgr15; break;
1593 case 0x38: conv= rgb32tobgr15; break;
1594 case 0x43: conv= rgb15tobgr16; break;
1595 case 0x44: conv= rgb16tobgr16; break;
1596 case 0x46: conv= rgb24tobgr16; break;
1597 case 0x48: conv= rgb32tobgr16; break;
1598 case 0x63: conv= rgb15tobgr24; break;
1599 case 0x64: conv= rgb16tobgr24; break;
1600 case 0x66: conv= rgb24tobgr24; break;
1601 case 0x68: conv= rgb32tobgr24; break;
1602 case 0x83: conv= rgb15tobgr32; break;
1603 case 0x84: conv= rgb16tobgr32; break;
1604 case 0x86: conv= rgb24tobgr32; break;
1605 case 0x88: conv= rgb32tobgr32; break;
6a4970ab 1606 default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
94c4def2 1607 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
0d9f3d85 1608 }
20380eb8 1609 }else{
6a4970ab 1610 av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
94c4def2 1611 sws_format_name(srcFormat), sws_format_name(dstFormat));
e09d12f4 1612 }
20380eb8 1613
e09d12f4
MN		 1614 if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)
1615 conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
0d9f3d85
A		 1616 else
1617 {
1618 int i;
1619 uint8_t *srcPtr= src[0];
1620 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1621
1622 for(i=0; i<srcSliceH; i++)
1623 {
e09d12f4 1624 conv(srcPtr, dstPtr, c->srcW*srcBpp);
0d9f3d85
A		 1625 srcPtr+= srcStride[0];
1626 dstPtr+= dstStride[0];
1627 }
6a4970ab 1628 }
d4e24275 1629 return srcSliceH;
0d9f3d85
A		 1630}
1631
d4e24275 1632static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
ec22603f
MN		 1633 int srcSliceH, uint8_t* dst[], int dstStride[]){
1634
1635 rgb24toyv12(
6a4970ab
DB		 1636 src[0],
1637 dst[0]+ srcSliceY *dstStride[0],
1638 dst[1]+(srcSliceY>>1)*dstStride[1],
ec22603f 1639 dst[2]+(srcSliceY>>1)*dstStride[2],
6a4970ab 1640 c->srcW, srcSliceH,
ec22603f 1641 dstStride[0], dstStride[1], srcStride[0]);
d4e24275 1642 return srcSliceH;
ec22603f
MN		 1643}
1644
d4e24275 1645static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
b241cbf2
MN		 1646 int srcSliceH, uint8_t* dst[], int dstStride[]){
1647 int i;
1648
1649 /* copy Y */
6a4970ab 1650 if(srcStride[0]==dstStride[0] && srcStride[0] > 0)
b241cbf2
MN		 1651 memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1652 else{
1653 uint8_t *srcPtr= src[0];
1654 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1655
1656 for(i=0; i<srcSliceH; i++)
1657 {
1658 memcpy(dstPtr, srcPtr, c->srcW);
1659 srcPtr+= srcStride[0];
1660 dstPtr+= dstStride[0];
1661 }
1662 }
1663
e9e12f0e 1664 if(c->dstFormat==PIX_FMT_YUV420P){
b241cbf2
MN		 1665 planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
1666 planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
1667 }else{
1668 planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
1669 planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
1670 }
d4e24275 1671 return srcSliceH;
b241cbf2
MN		 1672}
1673
b6654a54 1674/* unscaled copy like stuff (assumes nearly identical formats) */
3e499f53
MN		 1675static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1676 int srcSliceH, uint8_t* dst[], int dstStride[]){
b6654a54
MN		 1677
1678 if(isPacked(c->srcFormat))
1679 {
12e11662 1680 if(dstStride[0]==srcStride[0] && srcStride[0] > 0)
b6654a54
MN		 1681 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1682 else
1683 {
1684 int i;
1685 uint8_t *srcPtr= src[0];
1686 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
a861d4d7
MN		 1687 int length=0;
1688
1689 /* universal length finder */
6a4970ab 1690 while(length+c->srcW <= FFABS(dstStride[0])
05749630 1691 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
a861d4d7 1692 ASSERT(length!=0);
b6654a54
MN		 1693
1694 for(i=0; i<srcSliceH; i++)
1695 {
1696 memcpy(dstPtr, srcPtr, length);
1697 srcPtr+= srcStride[0];
1698 dstPtr+= dstStride[0];
1699 }
1700 }
1701 }
6a4970ab 1702 else
44c1035c 1703 { /* Planar YUV or gray */
b6654a54
MN		 1704 int plane;
1705 for(plane=0; plane<3; plane++)
1706 {
e616aa93
MN		 1707 int length= plane==0 ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1708 int y= plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1709 int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
44c1035c
MN		 1710
1711 if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
0d9f3d85 1712 {
44c1035c 1713 if(!isGray(c->dstFormat))
e616aa93 1714 memset(dst[plane], 128, dstStride[plane]*height);
0d9f3d85 1715 }
b6654a54
MN		 1716 else
1717 {
12e11662 1718 if(dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
44c1035c
MN		 1719 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1720 else
b6654a54 1721 {
44c1035c
MN		 1722 int i;
1723 uint8_t *srcPtr= src[plane];
1724 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1725 for(i=0; i<height; i++)
1726 {
1727 memcpy(dstPtr, srcPtr, length);
1728 srcPtr+= srcStride[plane];
1729 dstPtr+= dstStride[plane];
1730 }
b6654a54
MN		 1731 }
1732 }
1733 }
1734 }
d4e24275 1735 return srcSliceH;
37079906 1736}
28bf81c9 1737
4884b9e5
KS		 1738static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1739 int srcSliceH, uint8_t* dst[], int dstStride[]){
1740
1741 int length= c->srcW;
1742 int y= srcSliceY;
1743 int height= srcSliceH;
1744 int i, j;
1745 uint8_t *srcPtr= src[0];
1746 uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1747
1748 if(!isGray(c->dstFormat)){
1749 int height= -((-srcSliceH)>>c->chrDstVSubSample);
1750 memset(dst[1], 128, dstStride[1]*height);
1751 memset(dst[2], 128, dstStride[2]*height);
1752 }
1753 if(c->srcFormat == PIX_FMT_GRAY16LE) srcPtr++;
1754 for(i=0; i<height; i++)
1755 {
1756 for(j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1757 srcPtr+= srcStride[0];
1758 dstPtr+= dstStride[0];
1759 }
1760 return srcSliceH;
1761}
1762
1763static int graytogray16(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1764 int srcSliceH, uint8_t* dst[], int dstStride[]){
1765
1766 int length= c->srcW;
1767 int y= srcSliceY;
1768 int height= srcSliceH;
1769 int i, j;
1770 uint8_t *srcPtr= src[0];
1771 uint8_t *dstPtr= dst[0] + dstStride[0]*y;
1772 for(i=0; i<height; i++)
1773 {
1774 for(j=0; j<length; j++)
1775 {
1776 dstPtr[j<<1] = srcPtr[j];
1777 dstPtr[(j<<1)+1] = srcPtr[j];
1778 }
1779 srcPtr+= srcStride[0];
1780 dstPtr+= dstStride[0];
1781 }
1782 return srcSliceH;
1783}
1784
1785static int gray16swap(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1786 int srcSliceH, uint8_t* dst[], int dstStride[]){
1787
1788 int length= c->srcW;
1789 int y= srcSliceY;
1790 int height= srcSliceH;
1791 int i, j;
1792 uint16_t *srcPtr= src[0];
1793 uint16_t *dstPtr= dst[0] + dstStride[0]*y/2;
1794 for(i=0; i<height; i++)
1795 {
1796 for(j=0; j<length; j++) dstPtr[j] = bswap_16(srcPtr[j]);
1797 srcPtr+= srcStride[0]/2;
1798 dstPtr+= dstStride[0]/2;
1799 }
1800 return srcSliceH;
1801}
1802
1803
c7a810cc
MN		 1804static void getSubSampleFactors(int *h, int *v, int format){
1805 switch(format){
e9e12f0e
LA		 1806 case PIX_FMT_UYVY422:
1807 case PIX_FMT_YUYV422:
c7a810cc
MN		 1808 *h=1;
1809 *v=0;
1810 break;
e9e12f0e 1811 case PIX_FMT_YUV420P:
4884b9e5
KS		 1812 case PIX_FMT_GRAY16BE:
1813 case PIX_FMT_GRAY16LE:
e9e12f0e
LA		 1814 case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
1815 case PIX_FMT_NV12:
1816 case PIX_FMT_NV21:
c7a810cc
MN		 1817 *h=1;
1818 *v=1;
1819 break;
e9e12f0e 1820 case PIX_FMT_YUV410P:
c7a810cc
MN		 1821 *h=2;
1822 *v=2;
1823 break;
e9e12f0e 1824 case PIX_FMT_YUV444P:
d80e2fa2
MN		 1825 *h=0;
1826 *v=0;
1827 break;
e9e12f0e 1828 case PIX_FMT_YUV422P:
d80e2fa2
MN		 1829 *h=1;
1830 *v=0;
1831 break;
e9e12f0e 1832 case PIX_FMT_YUV411P:
d80e2fa2
MN		 1833 *h=2;
1834 *v=0;
1835 break;
c7a810cc
MN		 1836 default:
1837 *h=0;
1838 *v=0;
1839 break;
1840 }
1841}
1842
5427e242
MN		 1843static uint16_t roundToInt16(int64_t f){
1844 int r= (f + (1<<15))>>16;
1845 if(r<-0x7FFF) return 0x8000;
1846 else if(r> 0x7FFF) return 0x7FFF;
1847 else return r;
0481412a
MN		 1848}
1849
1850/**
5427e242 1851 * @param inv_table the yuv2rgb coeffs, normally Inverse_Table_6_9[x]
0481412a 1852 * @param fullRange if 1 then the luma range is 0..255 if 0 its 16..235
5427e242 1853 * @return -1 if not supported
0481412a 1854 */
5427e242
MN		 1855int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
1856 int64_t crv = inv_table[0];
1857 int64_t cbu = inv_table[1];
1858 int64_t cgu = -inv_table[2];
1859 int64_t cgv = -inv_table[3];
1860 int64_t cy = 1<<16;
1861 int64_t oy = 0;
1862
1863 if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
1864 memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
1865 memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
1866
1867 c->brightness= brightness;
1868 c->contrast = contrast;
1869 c->saturation= saturation;
1870 c->srcRange = srcRange;
1871 c->dstRange = dstRange;
0481412a
MN		 1872
1873 c->uOffset= 0x0400040004000400LL;
1874 c->vOffset= 0x0400040004000400LL;
1875
5427e242
MN		 1876 if(!srcRange){
1877 cy= (cy*255) / 219;
1878 oy= 16<<16;
edf97a75
MN		 1879 }else{
1880 crv= (crv*224) / 255;
1881 cbu= (cbu*224) / 255;
1882 cgu= (cgu*224) / 255;
1883 cgv= (cgv*224) / 255;
1884 }
0481412a 1885
5427e242
MN		 1886 cy = (cy *contrast )>>16;
1887 crv= (crv*contrast * saturation)>>32;
1888 cbu= (cbu*contrast * saturation)>>32;
1889 cgu= (cgu*contrast * saturation)>>32;
1890 cgv= (cgv*contrast * saturation)>>32;
0481412a 1891
5427e242 1892 oy -= 256*brightness;
0481412a
MN		 1893
1894 c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL;
1895 c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL;
1896 c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL;
1897 c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL;
1898 c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL;
1899 c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL;
5427e242
MN		 1900
1901 yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
1902 //FIXME factorize
a31de956 1903
6634d0ef
NP		 1904#ifdef COMPILE_ALTIVEC
1905 if (c->flags & SWS_CPU_CAPS_ALTIVEC)
1906 yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
6a4970ab 1907#endif
5427e242
MN		 1908 return 0;
1909}
1910
1911/**
1912 * @return -1 if not supported
1913 */
1914int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
1915 if(isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
1916
1917 *inv_table = c->srcColorspaceTable;
1918 *table = c->dstColorspaceTable;
1919 *srcRange = c->srcRange;
1920 *dstRange = c->dstRange;
1921 *brightness= c->brightness;
1922 *contrast = c->contrast;
1923 *saturation= c->saturation;
6a4970ab
DB		 1924
1925 return 0;
0481412a
MN		 1926}
1927
44cdb423
LA		 1928static int handle_jpeg(int *format)
1929{
1930 switch (*format) {
1931 case PIX_FMT_YUVJ420P:
1932 *format = PIX_FMT_YUV420P;
1933 return 1;
1934 case PIX_FMT_YUVJ422P:
1935 *format = PIX_FMT_YUV422P;
1936 return 1;
1937 case PIX_FMT_YUVJ444P:
1938 *format = PIX_FMT_YUV444P;
1939 return 1;
1940 default:
1941 return 0;
1942 }
1943}
1944
e9e12f0e 1945SwsContext *sws_getContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags,
66d1cdb6 1946 SwsFilter *srcFilter, SwsFilter *dstFilter, double *param){
28bf81c9 1947
28bf81c9
MN		 1948 SwsContext *c;
1949 int i;
ec62c38f 1950 int usesVFilter, usesHFilter;
e09d12f4 1951 int unscaled, needsDither;
44cdb423 1952 int srcRange, dstRange;
c7f822d9 1953 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
3d6a30d9 1954#if defined(ARCH_X86)
516b1f82 1955 if(flags & SWS_CPU_CAPS_MMX)
5cebb24b
MN		 1956 asm volatile("emms\n\t"::: "memory");
1957#endif
516b1f82 1958
9bde778e 1959#if !defined(RUNTIME_CPUDETECT) || !defined (CONFIG_GPL) //ensure that the flags match the compiled variant if cpudetect is off
a2faa401 1960 flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC);
516b1f82
MN		 1961#ifdef HAVE_MMX2
1962 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1963#elif defined (HAVE_3DNOW)
1964 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1965#elif defined (HAVE_MMX)
1966 flags |= SWS_CPU_CAPS_MMX;
a2faa401
RD		 1967#elif defined (HAVE_ALTIVEC)
1968 flags |= SWS_CPU_CAPS_ALTIVEC;
516b1f82 1969#endif
69796008 1970#endif /* RUNTIME_CPUDETECT */
516b1f82 1971 if(clip_table[512] != 255) globalInit();
700490a4 1972 if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);
fccb9b2b 1973
44c1035c 1974 unscaled = (srcW == dstW && srcH == dstH);
6a4970ab 1975 needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
e9e12f0e
LA		 1976 && (fmt_depth(dstFormat))<24
1977 && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
44c1035c 1978
44cdb423
LA		 1979 srcRange = handle_jpeg(&srcFormat);
1980 dstRange = handle_jpeg(&dstFormat);
1981
6a4970ab 1982 if(!isSupportedIn(srcFormat))
b81cf274 1983 {
2d529db5 1984 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input format\n", sws_format_name(srcFormat));
44c1035c
MN		 1985 return NULL;
1986 }
1987 if(!isSupportedOut(dstFormat))
1988 {
2d529db5 1989 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output format\n", sws_format_name(dstFormat));
44c1035c 1990 return NULL;
b81cf274 1991 }
44c1035c 1992
28bf81c9 1993 /* sanity check */
b81cf274
MN		 1994 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
1995 {
6a4970ab 1996 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
b81cf274
MN		 1997 srcW, srcH, dstW, dstH);
1998 return NULL;
1999 }
28bf81c9 2000
c7f822d9
MN		 2001 if(!dstFilter) dstFilter= &dummyFilter;
2002 if(!srcFilter) srcFilter= &dummyFilter;
2003
0dea6433 2004 c= av_mallocz(sizeof(SwsContext));
28bf81c9 2005
2d529db5 2006 c->av_class = &sws_context_class;
28bf81c9
MN		 2007 c->srcW= srcW;
2008 c->srcH= srcH;
2009 c->dstW= dstW;
2010 c->dstH= dstH;
5521b193
MN		 2011 c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2012 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
28bf81c9
MN		 2013 c->flags= flags;
2014 c->dstFormat= dstFormat;
2015 c->srcFormat= srcFormat;
379a2036 2016 c->vRounder= 4* 0x0001000100010001ULL;
28bf81c9 2017
ec62c38f
MN		 2018 usesHFilter= usesVFilter= 0;
2019 if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;
2020 if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;
2021 if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;
2022 if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;
2023 if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;
2024 if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;
2025 if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;
2026 if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;
e616aa93
MN		 2027
2028 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2029 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2030
2031 // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
2032 if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2033
5859233b
MN		 2034 // drop some chroma lines if the user wants it
2035 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2036 c->chrSrcVSubSample+= c->vChrDrop;
e616aa93 2037
5859233b 2038 // drop every 2. pixel for chroma calculation unless user wants full chroma
fa65e2f6
MN		 2039 if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2040 && srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8
2041 && srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4
2042 && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE)
e616aa93
MN		 2043 c->chrSrcHSubSample=1;
2044
66d1cdb6
MN		 2045 if(param){
2046 c->param[0] = param[0];
2047 c->param[1] = param[1];
2048 }else{
2049 c->param[0] =
2050 c->param[1] = SWS_PARAM_DEFAULT;
2051 }
2052
e616aa93
MN		 2053 c->chrIntHSubSample= c->chrDstHSubSample;
2054 c->chrIntVSubSample= c->chrSrcVSubSample;
5427e242 2055
911406f2 2056 // Note the -((-x)>>y) is so that we always round toward +inf.
e616aa93
MN		 2057 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2058 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2059 c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2060 c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
5427e242 2061
6a4970ab 2062 sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
cf7d1c1a 2063
b935781b 2064 /* unscaled special Cases */
ec62c38f 2065 if(unscaled && !usesHFilter && !usesVFilter)
37079906 2066 {
0d9f3d85 2067 /* yv12_to_nv12 */
e9e12f0e 2068 if(srcFormat == PIX_FMT_YUV420P && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
0d9f3d85
A		 2069 {
2070 c->swScale= PlanarToNV12Wrapper;
0d9f3d85 2071 }
9bde778e 2072#ifdef CONFIG_GPL
37079906 2073 /* yuv2bgr */
e9e12f0e 2074 if((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
37079906 2075 {
5427e242 2076 c->swScale= yuv2rgb_get_func_ptr(c);
37079906 2077 }
9bde778e 2078#endif
6a4970ab 2079
e9e12f0e 2080 if( srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P )
b241cbf2
MN		 2081 {
2082 c->swScale= yvu9toyv12Wrapper;
b241cbf2
MN		 2083 }
2084
ec22603f 2085 /* bgr24toYV12 */
e9e12f0e 2086 if(srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)
ec22603f 2087 c->swScale= bgr24toyv12Wrapper;
6a4970ab 2088
e09d12f4
MN		 2089 /* rgb/bgr -> rgb/bgr (no dither needed forms) */
2090 if( (isBGR(srcFormat) || isRGB(srcFormat))
6a4970ab 2091 && (isBGR(dstFormat) || isRGB(dstFormat))
94bd08b0
CEH		 2092 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
2093 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
2094 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
2095 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
2096 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2097 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2098 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
e09d12f4
MN		 2099 && !needsDither)
2100 c->swScale= rgb2rgbWrapper;
2101
2102 /* LQ converters if -sws 0 or -sws 4*/
2103 if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2104 /* rgb/bgr -> rgb/bgr (dither needed forms) */
2105 if( (isBGR(srcFormat) || isRGB(srcFormat))
6a4970ab 2106 && (isBGR(dstFormat) || isRGB(dstFormat))
e09d12f4
MN		 2107 && needsDither)
2108 c->swScale= rgb2rgbWrapper;
2ce486d8
MN		 2109
2110 /* yv12_to_yuy2 */
6a4970ab 2111 if(srcFormat == PIX_FMT_YUV420P &&
e9e12f0e 2112 (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))
2ce486d8 2113 {
e9e12f0e 2114 if (dstFormat == PIX_FMT_YUYV422)
caeaabe7
AB		 2115 c->swScale= PlanarToYuy2Wrapper;
2116 else
2117 c->swScale= PlanarToUyvyWrapper;
2ce486d8 2118 }
e09d12f4 2119 }
ec22603f 2120
6634d0ef 2121#ifdef COMPILE_ALTIVEC
b71cf33c 2122 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
6a4970ab 2123 ((srcFormat == PIX_FMT_YUV420P &&
e9e12f0e 2124 (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {
b71cf33c 2125 // unscaled YV12 -> packed YUV, we want speed
e9e12f0e 2126 if (dstFormat == PIX_FMT_YUYV422)
b71cf33c
RD		 2127 c->swScale= yv12toyuy2_unscaled_altivec;
2128 else
2129 c->swScale= yv12touyvy_unscaled_altivec;
2130 }
2131#endif
2132
20380eb8
MN		 2133 /* simple copy */
2134 if( srcFormat == dstFormat
20380eb8
MN		 2135 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2136 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2137 )
2138 {
2139 c->swScale= simpleCopy;
2140 }
2141
4884b9e5
KS		 2142 /* gray16{le,be} conversions */
2143 if(isGray16(srcFormat) && (isPlanarYUV(dstFormat) || (dstFormat == PIX_FMT_GRAY8)))
2144 {
2145 c->swScale= gray16togray;
2146 }
2147 if((isPlanarYUV(srcFormat) || (srcFormat == PIX_FMT_GRAY8)) && isGray16(dstFormat))
2148 {
2149 c->swScale= graytogray16;
2150 }
2151 if(srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
2152 {
2153 c->swScale= gray16swap;
6a4970ab 2154 }
4884b9e5 2155
e09d12f4 2156 if(c->swScale){
ec22603f 2157 if(flags&SWS_PRINT_INFO)
6a4970ab 2158 av_log(c, AV_LOG_INFO, "SwScaler: using unscaled %s -> %s special converter\n",
94c4def2 2159 sws_format_name(srcFormat), sws_format_name(dstFormat));
ec22603f
MN		 2160 return c;
2161 }
37079906
MN		 2162 }
2163
516b1f82 2164 if(flags & SWS_CPU_CAPS_MMX2)
28bf81c9
MN		 2165 {
2166 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2167 if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2168 {
2169 if(flags&SWS_PRINT_INFO)
2d529db5 2170 av_log(c, AV_LOG_INFO, "SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");
28bf81c9 2171 }
ec62c38f 2172 if(usesHFilter) c->canMMX2BeUsed=0;
28bf81c9
MN		 2173 }
2174 else
2175 c->canMMX2BeUsed=0;
2176
1e621b18
MN		 2177 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2178 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2179
28bf81c9
MN		 2180 // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2181 // but only for the FAST_BILINEAR mode otherwise do correct scaling
2182 // n-2 is the last chrominance sample available
2183 // this is not perfect, but noone shuld notice the difference, the more correct variant
2184 // would be like the vertical one, but that would require some special code for the
2185 // first and last pixel
2186 if(flags&SWS_FAST_BILINEAR)
2187 {
1e621b18
MN		 2188 if(c->canMMX2BeUsed)
2189 {
2190 c->lumXInc+= 20;
2191 c->chrXInc+= 20;
2192 }
77a416e8 2193 //we don't use the x86asm scaler if mmx is available
516b1f82 2194 else if(flags & SWS_CPU_CAPS_MMX)
1e621b18
MN		 2195 {
2196 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2197 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2198 }
28bf81c9
MN		 2199 }
2200
28bf81c9
MN		 2201 /* precalculate horizontal scaler filter coefficients */
2202 {
8c266f0c
RD		 2203 const int filterAlign=
2204 (flags & SWS_CPU_CAPS_MMX) ? 4 :
2205 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2206 1;
28bf81c9 2207
c7f822d9 2208 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
81a571a8
MN		 2209 srcW , dstW, filterAlign, 1<<14,
2210 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
66d1cdb6 2211 srcFilter->lumH, dstFilter->lumH, c->param);
c7f822d9 2212 initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
81a571a8
MN		 2213 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2214 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
66d1cdb6 2215 srcFilter->chrH, dstFilter->chrH, c->param);
28bf81c9 2216
dbdae6ec 2217#define MAX_FUNNY_CODE_SIZE 10000
17c613ef 2218#if defined(COMPILE_MMX2)
77a416e8 2219// can't downscale !!!
28bf81c9
MN		 2220 if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2221 {
113ef149 2222#ifdef MAP_ANONYMOUS
38d5c282
AJ		 2223 c->funnyYCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2224 c->funnyUVCode = (uint8_t*)mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2225#else
8dfe9094
RD		 2226 c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2227 c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
38d5c282
AJ		 2228#endif
2229
8dfe9094
RD		 2230 c->lumMmx2Filter = av_malloc((dstW /8+8)*sizeof(int16_t));
2231 c->chrMmx2Filter = av_malloc((c->chrDstW /4+8)*sizeof(int16_t));
2232 c->lumMmx2FilterPos= av_malloc((dstW /2/8+8)*sizeof(int32_t));
2233 c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
b7dc6f66
MN		 2234
2235 initMMX2HScaler( dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2236 initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
28bf81c9 2237 }
17c613ef 2238#endif /* defined(COMPILE_MMX2) */
28bf81c9
MN		 2239 } // Init Horizontal stuff
2240
2241
2242
2243 /* precalculate vertical scaler filter coefficients */
8c266f0c
RD		 2244 {
2245 const int filterAlign=
bca11e75 2246 (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
8c266f0c
RD		 2247 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2248 1;
2249
2250 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2251 srcH , dstH, filterAlign, (1<<12)-4,
2252 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
66d1cdb6 2253 srcFilter->lumV, dstFilter->lumV, c->param);
8c266f0c
RD		 2254 initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2255 c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,
2256 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
66d1cdb6 2257 srcFilter->chrV, dstFilter->chrV, c->param);
d33d485e
AC		 2258
2259#ifdef HAVE_ALTIVEC
8dfe9094 2260 c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
d8cfa553 2261 c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
d33d485e
AC		 2262
2263 for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2264 int j;
2265 short *p = (short *)&c->vYCoeffsBank[i];
2266 for (j=0;j<8;j++)
2267 p[j] = c->vLumFilter[i];
2268 }
2269
247d2c96 2270 for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
d33d485e
AC		 2271 int j;
2272 short *p = (short *)&c->vCCoeffsBank[i];
2273 for (j=0;j<8;j++)
2274 p[j] = c->vChrFilter[i];
2275 }
2276#endif
8c266f0c 2277 }
28bf81c9 2278
77a416e8 2279 // Calculate Buffer Sizes so that they won't run out while handling these damn slices
28bf81c9
MN		 2280 c->vLumBufSize= c->vLumFilterSize;
2281 c->vChrBufSize= c->vChrFilterSize;
2282 for(i=0; i<dstH; i++)
2283 {
2284 int chrI= i*c->chrDstH / dstH;
6f209cb5 2285 int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
e616aa93 2286 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
c4f1e443
MN		 2287
2288 nextSlice>>= c->chrSrcVSubSample;
2289 nextSlice<<= c->chrSrcVSubSample;
28bf81c9
MN		 2290 if(c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice)
2291 c->vLumBufSize= nextSlice - c->vLumFilterPos[i ];
e616aa93
MN		 2292 if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2293 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
28bf81c9
MN		 2294 }
2295
2296 // allocate pixbufs (we use dynamic allocation because otherwise we would need to
8dfe9094
RD		 2297 c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2298 c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
6c7506de 2299 //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)
5edb653b 2300 /* align at 16 bytes for AltiVec */
28bf81c9 2301 for(i=0; i<c->vLumBufSize; i++)
0dea6433 2302 c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(4000);
28bf81c9 2303 for(i=0; i<c->vChrBufSize; i++)
8dfe9094 2304 c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc(8000);
28bf81c9
MN		 2305
2306 //try to avoid drawing green stuff between the right end and the stride end
28bf81c9
MN		 2307 for(i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000);
2308
2309 ASSERT(c->chrDstH <= dstH)
28bf81c9 2310
28bf81c9
MN		 2311 if(flags&SWS_PRINT_INFO)
2312 {
2313#ifdef DITHER1XBPP
5521b193
MN		 2314 char *dither= " dithered";
2315#else
2316 char *dither= "";
28bf81c9
MN		 2317#endif
2318 if(flags&SWS_FAST_BILINEAR)
2d529db5 2319 av_log(c, AV_LOG_INFO, "SwScaler: FAST_BILINEAR scaler, ");
28bf81c9 2320 else if(flags&SWS_BILINEAR)
2d529db5 2321 av_log(c, AV_LOG_INFO, "SwScaler: BILINEAR scaler, ");
28bf81c9 2322 else if(flags&SWS_BICUBIC)
2d529db5 2323 av_log(c, AV_LOG_INFO, "SwScaler: BICUBIC scaler, ");
1e621b18 2324 else if(flags&SWS_X)
2d529db5 2325 av_log(c, AV_LOG_INFO, "SwScaler: Experimental scaler, ");
ff7ba856 2326 else if(flags&SWS_POINT)
2d529db5 2327 av_log(c, AV_LOG_INFO, "SwScaler: Nearest Neighbor / POINT scaler, ");
d8863d37 2328 else if(flags&SWS_AREA)
2d529db5 2329 av_log(c, AV_LOG_INFO, "SwScaler: Area Averageing scaler, ");
81a571a8 2330 else if(flags&SWS_BICUBLIN)
2d529db5 2331 av_log(c, AV_LOG_INFO, "SwScaler: luma BICUBIC / chroma BILINEAR scaler, ");
a86c461c 2332 else if(flags&SWS_GAUSS)
2d529db5 2333 av_log(c, AV_LOG_INFO, "SwScaler: Gaussian scaler, ");
a86c461c 2334 else if(flags&SWS_SINC)
2d529db5 2335 av_log(c, AV_LOG_INFO, "SwScaler: Sinc scaler, ");
a86c461c 2336 else if(flags&SWS_LANCZOS)
2d529db5 2337 av_log(c, AV_LOG_INFO, "SwScaler: Lanczos scaler, ");
a86c461c 2338 else if(flags&SWS_SPLINE)
2d529db5 2339 av_log(c, AV_LOG_INFO, "SwScaler: Bicubic spline scaler, ");
28bf81c9 2340 else
2d529db5 2341 av_log(c, AV_LOG_INFO, "SwScaler: ehh flags invalid?! ");
28bf81c9 2342
e9e12f0e 2343 if(dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
6a4970ab 2344 av_log(c, AV_LOG_INFO, "from %s to%s %s ",
94c4def2 2345 sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
0d9f3d85 2346 else
6a4970ab 2347 av_log(c, AV_LOG_INFO, "from %s to %s ",
94c4def2 2348 sws_format_name(srcFormat), sws_format_name(dstFormat));
28bf81c9 2349
516b1f82 2350 if(flags & SWS_CPU_CAPS_MMX2)
2d529db5 2351 av_log(c, AV_LOG_INFO, "using MMX2\n");
516b1f82 2352 else if(flags & SWS_CPU_CAPS_3DNOW)
2d529db5 2353 av_log(c, AV_LOG_INFO, "using 3DNOW\n");
516b1f82 2354 else if(flags & SWS_CPU_CAPS_MMX)
2d529db5 2355 av_log(c, AV_LOG_INFO, "using MMX\n");
a2faa401 2356 else if(flags & SWS_CPU_CAPS_ALTIVEC)
2d529db5 2357 av_log(c, AV_LOG_INFO, "using AltiVec\n");
6a4970ab 2358 else
2d529db5 2359 av_log(c, AV_LOG_INFO, "using C\n");
28bf81c9
MN		 2360 }
2361
516b1f82 2362 if(flags & SWS_PRINT_INFO)
28bf81c9 2363 {
516b1f82 2364 if(flags & SWS_CPU_CAPS_MMX)
28bf81c9
MN		 2365 {
2366 if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2d529db5 2367 av_log(c, AV_LOG_VERBOSE, "SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
28bf81c9
MN		 2368 else
2369 {
2370 if(c->hLumFilterSize==4)
2d529db5 2371 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n");
28bf81c9 2372 else if(c->hLumFilterSize==8)
2d529db5 2373 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n");
28bf81c9 2374 else
2d529db5 2375 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n");
28bf81c9
MN		 2376
2377 if(c->hChrFilterSize==4)
2d529db5 2378 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n");
28bf81c9 2379 else if(c->hChrFilterSize==8)
2d529db5 2380 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n");
28bf81c9 2381 else
2d529db5 2382 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n");
28bf81c9
MN		 2383 }
2384 }
2385 else
2386 {
3d6a30d9 2387#if defined(ARCH_X86)
2d529db5 2388 av_log(c, AV_LOG_VERBOSE, "SwScaler: using X86-Asm scaler for horizontal scaling\n");
28bf81c9
MN		 2389#else
2390 if(flags & SWS_FAST_BILINEAR)
2d529db5 2391 av_log(c, AV_LOG_VERBOSE, "SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n");
28bf81c9 2392 else
2d529db5 2393 av_log(c, AV_LOG_VERBOSE, "SwScaler: using C scaler for horizontal scaling\n");
28bf81c9
MN		 2394#endif
2395 }
6c7506de 2396 if(isPlanarYUV(dstFormat))
28bf81c9
MN		 2397 {
2398 if(c->vLumFilterSize==1)
2d529db5 2399 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9 2400 else
2d529db5 2401 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9
MN		 2402 }
2403 else
2404 {
2405 if(c->vLumFilterSize==1 && c->vChrFilterSize==2)
2d529db5 2406 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
516b1f82 2407 "SwScaler: 2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9 2408 else if(c->vLumFilterSize==2 && c->vChrFilterSize==2)
2d529db5 2409 av_log(c, AV_LOG_VERBOSE, "SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9 2410 else
2d529db5 2411 av_log(c, AV_LOG_VERBOSE, "SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9
MN		 2412 }
2413
e9e12f0e 2414 if(dstFormat==PIX_FMT_BGR24)
2d529db5 2415 av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR24 Converter\n",
516b1f82 2416 (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
e9e12f0e 2417 else if(dstFormat==PIX_FMT_RGB32)
2d529db5 2418 av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
e9e12f0e 2419 else if(dstFormat==PIX_FMT_BGR565)
2d529db5 2420 av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
e9e12f0e 2421 else if(dstFormat==PIX_FMT_BGR555)
2d529db5 2422 av_log(c, AV_LOG_VERBOSE, "SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
28bf81c9 2423
2d529db5 2424 av_log(c, AV_LOG_VERBOSE, "SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
28bf81c9 2425 }
516b1f82 2426 if(flags & SWS_PRINT_INFO)
1e621b18 2427 {
2d529db5 2428 av_log(c, AV_LOG_DEBUG, "SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1e621b18 2429 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2d529db5 2430 av_log(c, AV_LOG_DEBUG, "SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1e621b18
MN		 2431 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2432 }
37079906 2433
516b1f82 2434 c->swScale= getSwsFunc(flags);
28bf81c9
MN		 2435 return c;
2436}
2437
2438/**
77a416e8 2439 * swscale warper, so we don't need to export the SwsContext.
fccb9b2b
MN		 2440 * assumes planar YUV to be in YUV order instead of YVU
2441 */
703b56fb 2442int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
fccb9b2b 2443 int srcSliceH, uint8_t* dst[], int dstStride[]){
21c08a3f
MN		 2444 int i;
2445 uint8_t* src2[4]= {src[0], src[1], src[2]};
2446 uint32_t pal[256];
e63ac25d 2447 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2d529db5 2448 av_log(c, AV_LOG_ERROR, "swScaler: slices start in the middle!\n");
e63ac25d
JM		 2449 return 0;
2450 }
2451 if (c->sliceDir == 0) {
2452 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2453 }
2454
21c08a3f
MN		 2455 if(c->srcFormat == PIX_FMT_PAL8){
2456 for(i=0; i<256; i++){
2457 int p= ((uint32_t*)(src[1]))[i];
2458 int r= (p>>16)&0xFF;
2459 int g= (p>> 8)&0xFF;
2460 int b= p &0xFF;
2461 int y= av_clip_uint8(((RY*r + GY*g + BY*b)>>RGB2YUV_SHIFT) + 16 );
2462 int u= av_clip_uint8(((RU*r + GU*g + BU*b)>>RGB2YUV_SHIFT) + 128);
2463 int v= av_clip_uint8(((RV*r + GV*g + BV*b)>>RGB2YUV_SHIFT) + 128);
2464 pal[i]= y + (u<<8) + (v<<16);
2465 }
2466 src2[1]= pal;
2467 }
2468
e63ac25d
JM		 2469 // copy strides, so they can safely be modified
2470 if (c->sliceDir == 1) {
2471 // slices go from top to bottom
dd9bdeb9
MN		 2472 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2]};
2473 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2]};
0c840013 2474 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
e63ac25d
JM		 2475 } else {
2476 // slices go from bottom to top => we flip the image internally
dd9bdeb9 2477 uint8_t* dst2[4]= {dst[0] + (c->dstH-1)*dstStride[0],
e63ac25d
JM		 2478 dst[1] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1],
2479 dst[2] + ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2]};
dd9bdeb9
MN		 2480 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2]};
2481 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2]};
21c08a3f
MN		 2482
2483 src2[0] += (srcSliceH-1)*srcStride[0];
2484 if(c->srcFormat != PIX_FMT_PAL8)
2485 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2486 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
6a4970ab 2487
e63ac25d
JM		 2488 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2489 }
fccb9b2b
MN		 2490}
2491
2492/**
77a416e8 2493 * swscale warper, so we don't need to export the SwsContext
d4e24275 2494 */
703b56fb
LA		 2495int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2496 int srcSliceH, uint8_t* dst[], int dstStride[]){
2497 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
d4e24275 2498}
d4e24275 2499
6a4970ab 2500SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
e21206a8
MN		 2501 float lumaSharpen, float chromaSharpen,
2502 float chromaHShift, float chromaVShift,
2503 int verbose)
2504{
94c4def2 2505 SwsFilter *filter= av_malloc(sizeof(SwsFilter));
e21206a8
MN		 2506
2507 if(lumaGBlur!=0.0){
2508 filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
2509 filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
2510 }else{
2511 filter->lumH= sws_getIdentityVec();
2512 filter->lumV= sws_getIdentityVec();
2513 }
2514
2515 if(chromaGBlur!=0.0){
2516 filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
2517 filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
2518 }else{
2519 filter->chrH= sws_getIdentityVec();
2520 filter->chrV= sws_getIdentityVec();
2521 }
2522
2523 if(chromaSharpen!=0.0){
2e728364
MN		 2524 SwsVector *id= sws_getIdentityVec();
2525 sws_scaleVec(filter->chrH, -chromaSharpen);
2526 sws_scaleVec(filter->chrV, -chromaSharpen);
2527 sws_addVec(filter->chrH, id);
2528 sws_addVec(filter->chrV, id);
e21206a8
MN		 2529 sws_freeVec(id);
2530 }
2531
2532 if(lumaSharpen!=0.0){
2e728364
MN		 2533 SwsVector *id= sws_getIdentityVec();
2534 sws_scaleVec(filter->lumH, -lumaSharpen);
2535 sws_scaleVec(filter->lumV, -lumaSharpen);
2536 sws_addVec(filter->lumH, id);
2537 sws_addVec(filter->lumV, id);
e21206a8
MN		 2538 sws_freeVec(id);
2539 }
2540
2541 if(chromaHShift != 0.0)
2542 sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
2543
2544 if(chromaVShift != 0.0)
2545 sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
2546
2547 sws_normalizeVec(filter->chrH, 1.0);
2548 sws_normalizeVec(filter->chrV, 1.0);
2549 sws_normalizeVec(filter->lumH, 1.0);
2550 sws_normalizeVec(filter->lumV, 1.0);
2551
2552 if(verbose) sws_printVec(filter->chrH);
2553 if(verbose) sws_printVec(filter->lumH);
2554
2555 return filter;
2556}
2557
d4e24275 2558/**
28bf81c9
MN		 2559 * returns a normalized gaussian curve used to filter stuff
2560 * quality=3 is high quality, lowwer is lowwer quality
2561 */
d4e24275 2562SwsVector *sws_getGaussianVec(double variance, double quality){
28bf81c9
MN		 2563 const int length= (int)(variance*quality + 0.5) | 1;
2564 int i;
8dfe9094 2565 double *coeff= av_malloc(length*sizeof(double));
28bf81c9 2566 double middle= (length-1)*0.5;
94c4def2 2567 SwsVector *vec= av_malloc(sizeof(SwsVector));
c7f822d9
MN		 2568
2569 vec->coeff= coeff;
2570 vec->length= length;
28bf81c9
MN		 2571
2572 for(i=0; i<length; i++)
2573 {
2574 double dist= i-middle;
2575 coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI);
2576 }
2577
d4e24275 2578 sws_normalizeVec(vec, 1.0);
c7f822d9
MN		 2579
2580 return vec;
28bf81c9
MN		 2581}
2582
d4e24275 2583SwsVector *sws_getConstVec(double c, int length){
5521b193 2584 int i;
8dfe9094 2585 double *coeff= av_malloc(length*sizeof(double));
94c4def2 2586 SwsVector *vec= av_malloc(sizeof(SwsVector));
5521b193
MN		 2587
2588 vec->coeff= coeff;
2589 vec->length= length;
2590
2591 for(i=0; i<length; i++)
2592 coeff[i]= c;
2593
2594 return vec;
2595}
2596
2597
d4e24275 2598SwsVector *sws_getIdentityVec(void){
2e728364 2599 return sws_getConstVec(1.0, 1);
c7f822d9
MN		 2600}
2601
2e728364 2602double sws_dcVec(SwsVector *a){
28bf81c9 2603 int i;
2e728364 2604 double sum=0;
28bf81c9 2605
c7f822d9
MN		 2606 for(i=0; i<a->length; i++)
2607 sum+= a->coeff[i];
28bf81c9 2608
2e728364 2609 return sum;
28bf81c9
MN		 2610}
2611
d4e24275 2612void sws_scaleVec(SwsVector *a, double scalar){
c7f822d9
MN		 2613 int i;
2614
2615 for(i=0; i<a->length; i++)
2616 a->coeff[i]*= scalar;
2617}
2618
2e728364
MN		 2619void sws_normalizeVec(SwsVector *a, double height){
2620 sws_scaleVec(a, height/sws_dcVec(a));
2621}
2622
d4e24275 2623static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
c7f822d9 2624 int length= a->length + b->length - 1;
8dfe9094 2625 double *coeff= av_malloc(length*sizeof(double));
28bf81c9 2626 int i, j;
94c4def2 2627 SwsVector *vec= av_malloc(sizeof(SwsVector));
c7f822d9
MN		 2628
2629 vec->coeff= coeff;
2630 vec->length= length;
28bf81c9
MN		 2631
2632 for(i=0; i<length; i++) coeff[i]= 0.0;
2633
c7f822d9 2634 for(i=0; i<a->length; i++)
28bf81c9 2635 {
c7f822d9 2636 for(j=0; j<b->length; j++)
28bf81c9 2637 {
c7f822d9 2638 coeff[i+j]+= a->coeff[i]*b->coeff[j];
28bf81c9
MN		 2639 }
2640 }
2641
c7f822d9 2642 return vec;
28bf81c9
MN		 2643}
2644
d4e24275 2645static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
6f209cb5 2646 int length= FFMAX(a->length, b->length);
8dfe9094 2647 double *coeff= av_malloc(length*sizeof(double));
28bf81c9 2648 int i;
94c4def2 2649 SwsVector *vec= av_malloc(sizeof(SwsVector));
c7f822d9
MN		 2650
2651 vec->coeff= coeff;
2652 vec->length= length;
28bf81c9
MN		 2653
2654 for(i=0; i<length; i++) coeff[i]= 0.0;
2655
c7f822d9
MN		 2656 for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2657 for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
2658
2659 return vec;
28bf81c9 2660}
c7f822d9 2661
d4e24275 2662static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
6f209cb5 2663 int length= FFMAX(a->length, b->length);
8dfe9094 2664 double *coeff= av_malloc(length*sizeof(double));
c7f822d9 2665 int i;
94c4def2 2666 SwsVector *vec= av_malloc(sizeof(SwsVector));
c7f822d9
MN		 2667
2668 vec->coeff= coeff;
2669 vec->length= length;
2670
2671 for(i=0; i<length; i++) coeff[i]= 0.0;
2672
2673 for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
2674 for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
2675
2676 return vec;
2677}
2678
2679/* shift left / or right if "shift" is negative */
d4e24275 2680static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
05749630 2681 int length= a->length + FFABS(shift)*2;
8dfe9094 2682 double *coeff= av_malloc(length*sizeof(double));
ff7ba856 2683 int i;
94c4def2 2684 SwsVector *vec= av_malloc(sizeof(SwsVector));
c7f822d9
MN		 2685
2686 vec->coeff= coeff;
2687 vec->length= length;
2688
2689 for(i=0; i<length; i++) coeff[i]= 0.0;
2690
2691 for(i=0; i<a->length; i++)
2692 {
2693 coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
2694 }
2695
2696 return vec;
2697}
2698
d4e24275
MN		 2699void sws_shiftVec(SwsVector *a, int shift){
2700 SwsVector *shifted= sws_getShiftedVec(a, shift);
94c4def2 2701 av_free(a->coeff);
5cebb24b
MN		 2702 a->coeff= shifted->coeff;
2703 a->length= shifted->length;
94c4def2 2704 av_free(shifted);
5cebb24b
MN		 2705}
2706
d4e24275
MN		 2707void sws_addVec(SwsVector *a, SwsVector *b){
2708 SwsVector *sum= sws_sumVec(a, b);
94c4def2 2709 av_free(a->coeff);
5cebb24b
MN		 2710 a->coeff= sum->coeff;
2711 a->length= sum->length;
94c4def2 2712 av_free(sum);
5cebb24b
MN		 2713}
2714
d4e24275
MN		 2715void sws_subVec(SwsVector *a, SwsVector *b){
2716 SwsVector *diff= sws_diffVec(a, b);
94c4def2 2717 av_free(a->coeff);
5cebb24b
MN		 2718 a->coeff= diff->coeff;
2719 a->length= diff->length;
94c4def2 2720 av_free(diff);
5cebb24b
MN		 2721}
2722
d4e24275
MN		 2723void sws_convVec(SwsVector *a, SwsVector *b){
2724 SwsVector *conv= sws_getConvVec(a, b);
6a4970ab 2725 av_free(a->coeff);
5cebb24b
MN		 2726 a->coeff= conv->coeff;
2727 a->length= conv->length;
94c4def2 2728 av_free(conv);
5cebb24b
MN		 2729}
2730
d4e24275 2731SwsVector *sws_cloneVec(SwsVector *a){
8dfe9094 2732 double *coeff= av_malloc(a->length*sizeof(double));
5cebb24b 2733 int i;
94c4def2 2734 SwsVector *vec= av_malloc(sizeof(SwsVector));
5cebb24b
MN		 2735
2736 vec->coeff= coeff;
2737 vec->length= a->length;
2738
2739 for(i=0; i<a->length; i++) coeff[i]= a->coeff[i];
2740
2741 return vec;
2742}
2743
d4e24275 2744void sws_printVec(SwsVector *a){
c7f822d9
MN		 2745 int i;
2746 double max=0;
2747 double min=0;
2748 double range;
2749
2750 for(i=0; i<a->length; i++)
2751 if(a->coeff[i]>max) max= a->coeff[i];
2752
2753 for(i=0; i<a->length; i++)
2754 if(a->coeff[i]<min) min= a->coeff[i];
2755
2756 range= max - min;
2757
2758 for(i=0; i<a->length; i++)
2759 {
2760 int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
2d529db5
LA		 2761 av_log(NULL, AV_LOG_DEBUG, "%1.3f ", a->coeff[i]);
2762 for(;x>0; x--) av_log(NULL, AV_LOG_DEBUG, " ");
2763 av_log(NULL, AV_LOG_DEBUG, "|\n");
c7f822d9
MN		 2764 }
2765}
2766
d4e24275 2767void sws_freeVec(SwsVector *a){
c7f822d9 2768 if(!a) return;
94c4def2 2769 av_free(a->coeff);
c7f822d9
MN		 2770 a->coeff=NULL;
2771 a->length=0;
94c4def2 2772 av_free(a);
c7f822d9
MN		 2773}
2774
e21206a8
MN		 2775void sws_freeFilter(SwsFilter *filter){
2776 if(!filter) return;
2777
2778 if(filter->lumH) sws_freeVec(filter->lumH);
2779 if(filter->lumV) sws_freeVec(filter->lumV);
2780 if(filter->chrH) sws_freeVec(filter->chrH);
2781 if(filter->chrV) sws_freeVec(filter->chrV);
94c4def2 2782 av_free(filter);
e21206a8
MN		 2783}
2784
2785
d4e24275 2786void sws_freeContext(SwsContext *c){
c7f822d9 2787 int i;
c7f822d9
MN		 2788 if(!c) return;
2789
2790 if(c->lumPixBuf)
2791 {
6c7506de 2792 for(i=0; i<c->vLumBufSize; i++)
c7f822d9 2793 {
94c4def2 2794 av_free(c->lumPixBuf[i]);
c7f822d9
MN		 2795 c->lumPixBuf[i]=NULL;
2796 }
94c4def2 2797 av_free(c->lumPixBuf);
c7f822d9
MN		 2798 c->lumPixBuf=NULL;
2799 }
2800
2801 if(c->chrPixBuf)
2802 {
6c7506de 2803 for(i=0; i<c->vChrBufSize; i++)
c7f822d9 2804 {
94c4def2 2805 av_free(c->chrPixBuf[i]);
c7f822d9
MN		 2806 c->chrPixBuf[i]=NULL;
2807 }
94c4def2 2808 av_free(c->chrPixBuf);
c7f822d9
MN		 2809 c->chrPixBuf=NULL;
2810 }
2811
94c4def2 2812 av_free(c->vLumFilter);
c7f822d9 2813 c->vLumFilter = NULL;
94c4def2 2814 av_free(c->vChrFilter);
c7f822d9 2815 c->vChrFilter = NULL;
94c4def2 2816 av_free(c->hLumFilter);
c7f822d9 2817 c->hLumFilter = NULL;
94c4def2 2818 av_free(c->hChrFilter);
c7f822d9 2819 c->hChrFilter = NULL;
d33d485e 2820#ifdef HAVE_ALTIVEC
94c4def2 2821 av_free(c->vYCoeffsBank);
d33d485e 2822 c->vYCoeffsBank = NULL;
94c4def2 2823 av_free(c->vCCoeffsBank);
d33d485e
AC		 2824 c->vCCoeffsBank = NULL;
2825#endif
c7f822d9 2826
94c4def2 2827 av_free(c->vLumFilterPos);
c7f822d9 2828 c->vLumFilterPos = NULL;
94c4def2 2829 av_free(c->vChrFilterPos);
c7f822d9 2830 c->vChrFilterPos = NULL;
94c4def2 2831 av_free(c->hLumFilterPos);
c7f822d9 2832 c->hLumFilterPos = NULL;
94c4def2 2833 av_free(c->hChrFilterPos);
c7f822d9
MN		 2834 c->hChrFilterPos = NULL;
2835
9bde778e 2836#if defined(ARCH_X86) && defined(CONFIG_GPL)
0ce5275d 2837#ifdef MAP_ANONYMOUS
38d5c282
AJ		 2838 if(c->funnyYCode) munmap(c->funnyYCode, MAX_FUNNY_CODE_SIZE);
2839 if(c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
2840#else
94c4def2
LA		 2841 av_free(c->funnyYCode);
2842 av_free(c->funnyUVCode);
38d5c282
AJ		 2843#endif
2844 c->funnyYCode=NULL;
2845 c->funnyUVCode=NULL;
3d6a30d9 2846#endif /* defined(ARCH_X86) */
38d5c282 2847
94c4def2 2848 av_free(c->lumMmx2Filter);
b7dc6f66 2849 c->lumMmx2Filter=NULL;
94c4def2 2850 av_free(c->chrMmx2Filter);
b7dc6f66 2851 c->chrMmx2Filter=NULL;
94c4def2 2852 av_free(c->lumMmx2FilterPos);
b7dc6f66 2853 c->lumMmx2FilterPos=NULL;
94c4def2 2854 av_free(c->chrMmx2FilterPos);
b7dc6f66 2855 c->chrMmx2FilterPos=NULL;
94c4def2 2856 av_free(c->yuvTable);
cf7d1c1a 2857 c->yuvTable=NULL;
b7dc6f66 2858
94c4def2 2859 av_free(c);
c7f822d9
MN		 2860}
2861
22e46959
VP		 2862/**
2863 * Checks if context is valid or reallocs a new one instead.
2864 * If context is NULL, just calls sws_getContext() to get a new one.
2865 * Otherwise, checks if the parameters are the same already saved in context.
2866 * If that is the case, returns the current context.
2867 * Otherwise, frees context and gets a new one.
2868 *
2869 * Be warned that srcFilter, dstFilter are not checked, they are
2870 * asumed to remain valid.
2871 */
2872struct SwsContext *sws_getCachedContext(struct SwsContext *context,
2873 int srcW, int srcH, int srcFormat,
2874 int dstW, int dstH, int dstFormat, int flags,
2875 SwsFilter *srcFilter, SwsFilter *dstFilter, double *param)
2876{
2877 if (context != NULL) {
2878 if ((context->srcW != srcW) || (context->srcH != srcH) ||
2879 (context->srcFormat != srcFormat) ||
2880 (context->dstW != dstW) || (context->dstH != dstH) ||
2881 (context->dstFormat != dstFormat) || (context->flags != flags) ||
2882 (context->param != param))
2883 {
2884 sws_freeContext(context);
2885 context = NULL;
2886 }
2887 }
2888 if (context == NULL) {
2889 return sws_getContext(srcW, srcH, srcFormat,
2890 dstW, dstH, dstFormat, flags,
2891 srcFilter, dstFilter, param);
2892 }
2893 return context;
2894}
2895
Libav master