?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Make SwScaler recognize RGB48 BE/LE colourspaces (not support though).
[mplayer/glamo.git] / libswscale / swscale.c
blobe3bf0c3c576bf3623d92d2dcf81ba252d3e47921
1 /*
2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 *
20 * the C code (not assembly, mmx, ...) of this file can be used
21 * under the LGPL license too
22 */
23
24 /*
25 supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
26 supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
27 {BGR,RGB}{1,4,8,15,16} support dithering
28
29 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
30 YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
31 x -> x
32 YUV9 -> YV12
33 YUV9/YV12 -> Y800
34 Y800 -> YUV9/YV12
35 BGR24 -> BGR32 & RGB24 -> RGB32
36 BGR32 -> BGR24 & RGB32 -> RGB24
37 BGR15 -> BGR16
38 */
39
40 /*
41 tested special converters (most are tested actually, but I did not write it down ...)
42 YV12 -> BGR16
43 YV12 -> YV12
44 BGR15 -> BGR16
45 BGR16 -> BGR16
46 YVU9 -> YV12
47
48 untested special converters
49 YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
50 YV12/I420 -> YV12/I420
51 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
52 BGR24 -> BGR32 & RGB24 -> RGB32
53 BGR32 -> BGR24 & RGB32 -> RGB24
54 BGR24 -> YV12
55 */
56
57 #define _SVID_SOURCE //needed for MAP_ANONYMOUS
58 #include <inttypes.h>
59 #include <string.h>
60 #include <math.h>
61 #include <stdio.h>
62 #include <unistd.h>
63 #include "config.h"
64 #include <assert.h>
65 #if HAVE_SYS_MMAN_H
66 #include <sys/mman.h>
67 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
68 #define MAP_ANONYMOUS MAP_ANON
69 #endif
70 #endif
71 #if HAVE_VIRTUALALLOC
72 #define WIN32_LEAN_AND_MEAN
73 #include <windows.h>
74 #endif
75 #include "swscale.h"
76 #include "swscale_internal.h"
77 #include "rgb2rgb.h"
78 #include "libavutil/x86_cpu.h"
79 #include "libavutil/bswap.h"
80
81 unsigned swscale_version(void)
82 {
83 return LIBSWSCALE_VERSION_INT;
84 }
85
86 #undef MOVNTQ
87 #undef PAVGB
88
89 //#undef HAVE_MMX2
90 //#define HAVE_AMD3DNOW
91 //#undef HAVE_MMX
92 //#undef ARCH_X86
93 //#define WORDS_BIGENDIAN
94 #define DITHER1XBPP
95
96 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
97
98 #define RET 0xC3 //near return opcode for x86
99
100 #ifdef M_PI
101 #define PI M_PI
102 #else
103 #define PI 3.14159265358979323846
104 #endif
105
106 #define isSupportedIn(x) ( \
107 (x)==PIX_FMT_YUV420P \
108 || (x)==PIX_FMT_YUVA420P \
109 || (x)==PIX_FMT_YUYV422 \
110 || (x)==PIX_FMT_UYVY422 \
111 || (x)==PIX_FMT_RGB32 \
112 || (x)==PIX_FMT_RGB32_1 \
113 || (x)==PIX_FMT_BGR24 \
114 || (x)==PIX_FMT_BGR565 \
115 || (x)==PIX_FMT_BGR555 \
116 || (x)==PIX_FMT_BGR32 \
117 || (x)==PIX_FMT_BGR32_1 \
118 || (x)==PIX_FMT_RGB24 \
119 || (x)==PIX_FMT_RGB565 \
120 || (x)==PIX_FMT_RGB555 \
121 || (x)==PIX_FMT_GRAY8 \
122 || (x)==PIX_FMT_YUV410P \
123 || (x)==PIX_FMT_YUV440P \
124 || (x)==PIX_FMT_GRAY16BE \
125 || (x)==PIX_FMT_GRAY16LE \
126 || (x)==PIX_FMT_YUV444P \
127 || (x)==PIX_FMT_YUV422P \
128 || (x)==PIX_FMT_YUV411P \
129 || (x)==PIX_FMT_PAL8 \
130 || (x)==PIX_FMT_BGR8 \
131 || (x)==PIX_FMT_RGB8 \
132 || (x)==PIX_FMT_BGR4_BYTE \
133 || (x)==PIX_FMT_RGB4_BYTE \
134 || (x)==PIX_FMT_YUV440P \
135 || (x)==PIX_FMT_MONOWHITE \
136 || (x)==PIX_FMT_MONOBLACK \
137 || (x)==PIX_FMT_YUV420PLE \
138 || (x)==PIX_FMT_YUV422PLE \
139 || (x)==PIX_FMT_YUV444PLE \
140 || (x)==PIX_FMT_YUV420PBE \
141 || (x)==PIX_FMT_YUV422PBE \
142 || (x)==PIX_FMT_YUV444PBE \
143 )
144 #define isSupportedOut(x) ( \
145 (x)==PIX_FMT_YUV420P \
146 || (x)==PIX_FMT_YUVA420P \
147 || (x)==PIX_FMT_YUYV422 \
148 || (x)==PIX_FMT_UYVY422 \
149 || (x)==PIX_FMT_YUV444P \
150 || (x)==PIX_FMT_YUV422P \
151 || (x)==PIX_FMT_YUV411P \
152 || isRGB(x) \
153 || isBGR(x) \
154 || (x)==PIX_FMT_NV12 \
155 || (x)==PIX_FMT_NV21 \
156 || (x)==PIX_FMT_GRAY16BE \
157 || (x)==PIX_FMT_GRAY16LE \
158 || (x)==PIX_FMT_GRAY8 \
159 || (x)==PIX_FMT_YUV410P \
160 || (x)==PIX_FMT_YUV440P \
161 || (x)==PIX_FMT_YUV420PLE \
162 || (x)==PIX_FMT_YUV422PLE \
163 || (x)==PIX_FMT_YUV444PLE \
164 || (x)==PIX_FMT_YUV420PBE \
165 || (x)==PIX_FMT_YUV422PBE \
166 || (x)==PIX_FMT_YUV444PBE \
167 )
168 #define isPacked(x) ( \
169 (x)==PIX_FMT_PAL8 \
170 || (x)==PIX_FMT_YUYV422 \
171 || (x)==PIX_FMT_UYVY422 \
172 || isRGB(x) \
173 || isBGR(x) \
174 )
175 #define usePal(x) ( \
176 (x)==PIX_FMT_PAL8 \
177 || (x)==PIX_FMT_BGR4_BYTE \
178 || (x)==PIX_FMT_RGB4_BYTE \
179 || (x)==PIX_FMT_BGR8 \
180 || (x)==PIX_FMT_RGB8 \
181 )
182
183 #define RGB2YUV_SHIFT 15
184 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
185 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
186 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
187 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
188 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
189 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
190 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
191 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
192 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
193
194 extern const int32_t ff_yuv2rgb_coeffs[8][4];
195
196 static const double rgb2yuv_table[8][9]={
197 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
198 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
199 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
200 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
201 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
202 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
203 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
204 {0.701 , 0.087 , 0.212 , -0.384, 0.5 -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
205 };
206
207 /*
208 NOTES
209 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
210
211 TODO
212 more intelligent misalignment avoidance for the horizontal scaler
213 write special vertical cubic upscale version
214 optimize C code (YV12 / minmax)
215 add support for packed pixel YUV input & output
216 add support for Y8 output
217 optimize BGR24 & BGR32
218 add BGR4 output support
219 write special BGR->BGR scaler
220 */
221
222 #if ARCH_X86 && CONFIG_GPL
223 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
224 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
225 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
226 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
227 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
228 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
229 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
230 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
231
232 const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
233 0x0103010301030103LL,
234 0x0200020002000200LL,};
235
236 const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
237 0x0602060206020602LL,
238 0x0004000400040004LL,};
239
240 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
241 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
242 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
243 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
244 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
245 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
246
247 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
248 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
249 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
250
251 #ifdef FAST_BGR2YV12
252 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
253 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
254 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
255 #else
256 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
257 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
258 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
259 #endif /* FAST_BGR2YV12 */
260 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
261 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
262 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
263
264 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
265 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
266 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
267 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
268 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
269
270 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
271 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
272 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
273 };
274
275 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
276
277 #endif /* ARCH_X86 && CONFIG_GPL */
278
279 // clipping helper table for C implementations:
280 static unsigned char clip_table[768];
281
282 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
283
284 static const uint8_t __attribute__((aligned(8))) dither_2x2_4[2][8]={
285 { 1, 3, 1, 3, 1, 3, 1, 3, },
286 { 2, 0, 2, 0, 2, 0, 2, 0, },
287 };
288
289 static const uint8_t __attribute__((aligned(8))) dither_2x2_8[2][8]={
290 { 6, 2, 6, 2, 6, 2, 6, 2, },
291 { 0, 4, 0, 4, 0, 4, 0, 4, },
292 };
293
294 const uint8_t __attribute__((aligned(8))) dither_8x8_32[8][8]={
295 { 17, 9, 23, 15, 16, 8, 22, 14, },
296 { 5, 29, 3, 27, 4, 28, 2, 26, },
297 { 21, 13, 19, 11, 20, 12, 18, 10, },
298 { 0, 24, 6, 30, 1, 25, 7, 31, },
299 { 16, 8, 22, 14, 17, 9, 23, 15, },
300 { 4, 28, 2, 26, 5, 29, 3, 27, },
301 { 20, 12, 18, 10, 21, 13, 19, 11, },
302 { 1, 25, 7, 31, 0, 24, 6, 30, },
303 };
304
305 #if 0
306 const uint8_t __attribute__((aligned(8))) dither_8x8_64[8][8]={
307 { 0, 48, 12, 60, 3, 51, 15, 63, },
308 { 32, 16, 44, 28, 35, 19, 47, 31, },
309 { 8, 56, 4, 52, 11, 59, 7, 55, },
310 { 40, 24, 36, 20, 43, 27, 39, 23, },
311 { 2, 50, 14, 62, 1, 49, 13, 61, },
312 { 34, 18, 46, 30, 33, 17, 45, 29, },
313 { 10, 58, 6, 54, 9, 57, 5, 53, },
314 { 42, 26, 38, 22, 41, 25, 37, 21, },
315 };
316 #endif
317
318 const uint8_t __attribute__((aligned(8))) dither_8x8_73[8][8]={
319 { 0, 55, 14, 68, 3, 58, 17, 72, },
320 { 37, 18, 50, 32, 40, 22, 54, 35, },
321 { 9, 64, 5, 59, 13, 67, 8, 63, },
322 { 46, 27, 41, 23, 49, 31, 44, 26, },
323 { 2, 57, 16, 71, 1, 56, 15, 70, },
324 { 39, 21, 52, 34, 38, 19, 51, 33, },
325 { 11, 66, 7, 62, 10, 65, 6, 60, },
326 { 48, 30, 43, 25, 47, 29, 42, 24, },
327 };
328
329 #if 0
330 const uint8_t __attribute__((aligned(8))) dither_8x8_128[8][8]={
331 { 68, 36, 92, 60, 66, 34, 90, 58, },
332 { 20, 116, 12, 108, 18, 114, 10, 106, },
333 { 84, 52, 76, 44, 82, 50, 74, 42, },
334 { 0, 96, 24, 120, 6, 102, 30, 126, },
335 { 64, 32, 88, 56, 70, 38, 94, 62, },
336 { 16, 112, 8, 104, 22, 118, 14, 110, },
337 { 80, 48, 72, 40, 86, 54, 78, 46, },
338 { 4, 100, 28, 124, 2, 98, 26, 122, },
339 };
340 #endif
341
342 #if 1
343 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
344 {117, 62, 158, 103, 113, 58, 155, 100, },
345 { 34, 199, 21, 186, 31, 196, 17, 182, },
346 {144, 89, 131, 76, 141, 86, 127, 72, },
347 { 0, 165, 41, 206, 10, 175, 52, 217, },
348 {110, 55, 151, 96, 120, 65, 162, 107, },
349 { 28, 193, 14, 179, 38, 203, 24, 189, },
350 {138, 83, 124, 69, 148, 93, 134, 79, },
351 { 7, 172, 48, 213, 3, 168, 45, 210, },
352 };
353 #elif 1
354 // tries to correct a gamma of 1.5
355 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
356 { 0, 143, 18, 200, 2, 156, 25, 215, },
357 { 78, 28, 125, 64, 89, 36, 138, 74, },
358 { 10, 180, 3, 161, 16, 195, 8, 175, },
359 {109, 51, 93, 38, 121, 60, 105, 47, },
360 { 1, 152, 23, 210, 0, 147, 20, 205, },
361 { 85, 33, 134, 71, 81, 30, 130, 67, },
362 { 14, 190, 6, 171, 12, 185, 5, 166, },
363 {117, 57, 101, 44, 113, 54, 97, 41, },
364 };
365 #elif 1
366 // tries to correct a gamma of 2.0
367 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
368 { 0, 124, 8, 193, 0, 140, 12, 213, },
369 { 55, 14, 104, 42, 66, 19, 119, 52, },
370 { 3, 168, 1, 145, 6, 187, 3, 162, },
371 { 86, 31, 70, 21, 99, 39, 82, 28, },
372 { 0, 134, 11, 206, 0, 129, 9, 200, },
373 { 62, 17, 114, 48, 58, 16, 109, 45, },
374 { 5, 181, 2, 157, 4, 175, 1, 151, },
375 { 95, 36, 78, 26, 90, 34, 74, 24, },
376 };
377 #else
378 // tries to correct a gamma of 2.5
379 const uint8_t __attribute__((aligned(8))) dither_8x8_220[8][8]={
380 { 0, 107, 3, 187, 0, 125, 6, 212, },
381 { 39, 7, 86, 28, 49, 11, 102, 36, },
382 { 1, 158, 0, 131, 3, 180, 1, 151, },
383 { 68, 19, 52, 12, 81, 25, 64, 17, },
384 { 0, 119, 5, 203, 0, 113, 4, 195, },
385 { 45, 9, 96, 33, 42, 8, 91, 30, },
386 { 2, 172, 1, 144, 2, 165, 0, 137, },
387 { 77, 23, 60, 15, 72, 21, 56, 14, },
388 };
389 #endif
390
391 const char *sws_format_name(enum PixelFormat format)
392 {
393 switch (format) {
394 case PIX_FMT_YUV420P:
395 return "yuv420p";
396 case PIX_FMT_YUVA420P:
397 return "yuva420p";
398 case PIX_FMT_YUYV422:
399 return "yuyv422";
400 case PIX_FMT_RGB24:
401 return "rgb24";
402 case PIX_FMT_BGR24:
403 return "bgr24";
404 case PIX_FMT_YUV422P:
405 return "yuv422p";
406 case PIX_FMT_YUV444P:
407 return "yuv444p";
408 case PIX_FMT_RGB32:
409 return "rgb32";
410 case PIX_FMT_YUV410P:
411 return "yuv410p";
412 case PIX_FMT_YUV411P:
413 return "yuv411p";
414 case PIX_FMT_RGB565:
415 return "rgb565";
416 case PIX_FMT_RGB555:
417 return "rgb555";
418 case PIX_FMT_GRAY16BE:
419 return "gray16be";
420 case PIX_FMT_GRAY16LE:
421 return "gray16le";
422 case PIX_FMT_GRAY8:
423 return "gray8";
424 case PIX_FMT_MONOWHITE:
425 return "mono white";
426 case PIX_FMT_MONOBLACK:
427 return "mono black";
428 case PIX_FMT_PAL8:
429 return "Palette";
430 case PIX_FMT_YUVJ420P:
431 return "yuvj420p";
432 case PIX_FMT_YUVJ422P:
433 return "yuvj422p";
434 case PIX_FMT_YUVJ444P:
435 return "yuvj444p";
436 case PIX_FMT_XVMC_MPEG2_MC:
437 return "xvmc_mpeg2_mc";
438 case PIX_FMT_XVMC_MPEG2_IDCT:
439 return "xvmc_mpeg2_idct";
440 case PIX_FMT_UYVY422:
441 return "uyvy422";
442 case PIX_FMT_UYYVYY411:
443 return "uyyvyy411";
444 case PIX_FMT_RGB32_1:
445 return "rgb32x";
446 case PIX_FMT_BGR32_1:
447 return "bgr32x";
448 case PIX_FMT_BGR32:
449 return "bgr32";
450 case PIX_FMT_BGR565:
451 return "bgr565";
452 case PIX_FMT_BGR555:
453 return "bgr555";
454 case PIX_FMT_BGR8:
455 return "bgr8";
456 case PIX_FMT_BGR4:
457 return "bgr4";
458 case PIX_FMT_BGR4_BYTE:
459 return "bgr4 byte";
460 case PIX_FMT_RGB8:
461 return "rgb8";
462 case PIX_FMT_RGB4:
463 return "rgb4";
464 case PIX_FMT_RGB4_BYTE:
465 return "rgb4 byte";
466 case PIX_FMT_RGB48BE:
467 return "rgb48be";
468 case PIX_FMT_RGB48LE:
469 return "rgb48le";
470 case PIX_FMT_NV12:
471 return "nv12";
472 case PIX_FMT_NV21:
473 return "nv21";
474 case PIX_FMT_YUV440P:
475 return "yuv440p";
476 case PIX_FMT_VDPAU_H264:
477 return "vdpau_h264";
478 case PIX_FMT_VDPAU_MPEG1:
479 return "vdpau_mpeg1";
480 case PIX_FMT_VDPAU_MPEG2:
481 return "vdpau_mpeg2";
482 case PIX_FMT_VDPAU_WMV3:
483 return "vdpau_wmv3";
484 case PIX_FMT_VDPAU_VC1:
485 return "vdpau_vc1";
486 case PIX_FMT_YUV420PLE:
487 return "yuv420ple";
488 case PIX_FMT_YUV422PLE:
489 return "yuv422ple";
490 case PIX_FMT_YUV444PLE:
491 return "yuv444ple";
492 case PIX_FMT_YUV420PBE:
493 return "yuv420pbe";
494 case PIX_FMT_YUV422PBE:
495 return "yuv422pbe";
496 case PIX_FMT_YUV444PBE:
497 return "yuv444pbe";
498 default:
499 return "Unknown format";
500 }
501 }
502
503 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
504 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
505 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
506 {
507 //FIXME Optimize (just quickly written not optimized..)
508 int i;
509 for (i=0; i<dstW; i++)
510 {
511 int val=1<<18;
512 int j;
513 for (j=0; j<lumFilterSize; j++)
514 val += lumSrc[j][i] * lumFilter[j];
515
516 dest[i]= av_clip_uint8(val>>19);
517 }
518
519 if (uDest)
520 for (i=0; i<chrDstW; i++)
521 {
522 int u=1<<18;
523 int v=1<<18;
524 int j;
525 for (j=0; j<chrFilterSize; j++)
526 {
527 u += chrSrc[j][i] * chrFilter[j];
528 v += chrSrc[j][i + VOFW] * chrFilter[j];
529 }
530
531 uDest[i]= av_clip_uint8(u>>19);
532 vDest[i]= av_clip_uint8(v>>19);
533 }
534
535 if (CONFIG_SWSCALE_ALPHA && aDest)
536 for (i=0; i<dstW; i++){
537 int val=1<<18;
538 int j;
539 for (j=0; j<lumFilterSize; j++)
540 val += alpSrc[j][i] * lumFilter[j];
541
542 aDest[i]= av_clip_uint8(val>>19);
543 }
544
545 }
546
547 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
548 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
549 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
550 {
551 //FIXME Optimize (just quickly written not optimized..)
552 int i;
553 for (i=0; i<dstW; i++)
554 {
555 int val=1<<18;
556 int j;
557 for (j=0; j<lumFilterSize; j++)
558 val += lumSrc[j][i] * lumFilter[j];
559
560 dest[i]= av_clip_uint8(val>>19);
561 }
562
563 if (!uDest)
564 return;
565
566 if (dstFormat == PIX_FMT_NV12)
567 for (i=0; i<chrDstW; i++)
568 {
569 int u=1<<18;
570 int v=1<<18;
571 int j;
572 for (j=0; j<chrFilterSize; j++)
573 {
574 u += chrSrc[j][i] * chrFilter[j];
575 v += chrSrc[j][i + VOFW] * chrFilter[j];
576 }
577
578 uDest[2*i]= av_clip_uint8(u>>19);
579 uDest[2*i+1]= av_clip_uint8(v>>19);
580 }
581 else
582 for (i=0; i<chrDstW; i++)
583 {
584 int u=1<<18;
585 int v=1<<18;
586 int j;
587 for (j=0; j<chrFilterSize; j++)
588 {
589 u += chrSrc[j][i] * chrFilter[j];
590 v += chrSrc[j][i + VOFW] * chrFilter[j];
591 }
592
593 uDest[2*i]= av_clip_uint8(v>>19);
594 uDest[2*i+1]= av_clip_uint8(u>>19);
595 }
596 }
597
598 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
599 for (i=0; i<(dstW>>1); i++){\
600 int j;\
601 int Y1 = 1<<18;\
602 int Y2 = 1<<18;\
603 int U = 1<<18;\
604 int V = 1<<18;\
605 int av_unused A1, A2;\
606 type av_unused *r, *b, *g;\
607 const int i2= 2*i;\
608 \
609 for (j=0; j<lumFilterSize; j++)\
610 {\
611 Y1 += lumSrc[j][i2] * lumFilter[j];\
612 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
613 }\
614 for (j=0; j<chrFilterSize; j++)\
615 {\
616 U += chrSrc[j][i] * chrFilter[j];\
617 V += chrSrc[j][i+VOFW] * chrFilter[j];\
618 }\
619 Y1>>=19;\
620 Y2>>=19;\
621 U >>=19;\
622 V >>=19;\
623 if (alpha){\
624 A1 = 1<<18;\
625 A2 = 1<<18;\
626 for (j=0; j<lumFilterSize; j++){\
627 A1 += alpSrc[j][i2 ] * lumFilter[j];\
628 A2 += alpSrc[j][i2+1] * lumFilter[j];\
629 }\
630 A1>>=19;\
631 A2>>=19;\
632 }\
633
634 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
635 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
636 if ((Y1|Y2|U|V)&256)\
637 {\
638 if (Y1>255) Y1=255; \
639 else if (Y1<0)Y1=0; \
640 if (Y2>255) Y2=255; \
641 else if (Y2<0)Y2=0; \
642 if (U>255) U=255; \
643 else if (U<0) U=0; \
644 if (V>255) V=255; \
645 else if (V<0) V=0; \
646 }\
647 if (alpha && ((A1|A2)&256)){\
648 A1=av_clip_uint8(A1);\
649 A2=av_clip_uint8(A2);\
650 }
651
652 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
653 for (i=0; i<dstW; i++){\
654 int j;\
655 int Y = 0;\
656 int U = -128<<19;\
657 int V = -128<<19;\
658 int av_unused A;\
659 int R,G,B;\
660 \
661 for (j=0; j<lumFilterSize; j++){\
662 Y += lumSrc[j][i ] * lumFilter[j];\
663 }\
664 for (j=0; j<chrFilterSize; j++){\
665 U += chrSrc[j][i ] * chrFilter[j];\
666 V += chrSrc[j][i+VOFW] * chrFilter[j];\
667 }\
668 Y >>=10;\
669 U >>=10;\
670 V >>=10;\
671 if (alpha){\
672 A = rnd;\
673 for (j=0; j<lumFilterSize; j++)\
674 A += alpSrc[j][i ] * lumFilter[j];\
675 A >>=19;\
676 if (A&256)\
677 A = av_clip_uint8(A);\
678 }\
679
680 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
681 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
682 Y-= c->yuv2rgb_y_offset;\
683 Y*= c->yuv2rgb_y_coeff;\
684 Y+= rnd;\
685 R= Y + V*c->yuv2rgb_v2r_coeff;\
686 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
687 B= Y + U*c->yuv2rgb_u2b_coeff;\
688 if ((R|G|B)&(0xC0000000)){\
689 if (R>=(256<<22)) R=(256<<22)-1; \
690 else if (R<0)R=0; \
691 if (G>=(256<<22)) G=(256<<22)-1; \
692 else if (G<0)G=0; \
693 if (B>=(256<<22)) B=(256<<22)-1; \
694 else if (B<0)B=0; \
695 }\
696
697
698 #define YSCALE_YUV_2_GRAY16_C \
699 for (i=0; i<(dstW>>1); i++){\
700 int j;\
701 int Y1 = 1<<18;\
702 int Y2 = 1<<18;\
703 int U = 1<<18;\
704 int V = 1<<18;\
705 \
706 const int i2= 2*i;\
707 \
708 for (j=0; j<lumFilterSize; j++)\
709 {\
710 Y1 += lumSrc[j][i2] * lumFilter[j];\
711 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
712 }\
713 Y1>>=11;\
714 Y2>>=11;\
715 if ((Y1|Y2|U|V)&65536)\
716 {\
717 if (Y1>65535) Y1=65535; \
718 else if (Y1<0)Y1=0; \
719 if (Y2>65535) Y2=65535; \
720 else if (Y2<0)Y2=0; \
721 }
722
723 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
724 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
725 r = (type *)c->table_rV[V]; \
726 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
727 b = (type *)c->table_bU[U]; \
728
729 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
730 for (i=0; i<(dstW>>1); i++){ \
731 const int i2= 2*i; \
732 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
733 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
734 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
735 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
736 type av_unused *r, *b, *g; \
737 int av_unused A1, A2; \
738 if (alpha){\
739 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
740 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
741 }\
742
743 #define YSCALE_YUV_2_GRAY16_2_C \
744 for (i=0; i<(dstW>>1); i++){ \
745 const int i2= 2*i; \
746 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
747 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11; \
748
749 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
750 YSCALE_YUV_2_PACKED2_C(type,alpha)\
751 r = (type *)c->table_rV[V];\
752 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
753 b = (type *)c->table_bU[U];\
754
755 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
756 for (i=0; i<(dstW>>1); i++){\
757 const int i2= 2*i;\
758 int Y1= buf0[i2 ]>>7;\
759 int Y2= buf0[i2+1]>>7;\
760 int U= (uvbuf1[i ])>>7;\
761 int V= (uvbuf1[i+VOFW])>>7;\
762 type av_unused *r, *b, *g;\
763 int av_unused A1, A2;\
764 if (alpha){\
765 A1= abuf0[i2 ]>>7;\
766 A2= abuf0[i2+1]>>7;\
767 }\
768
769 #define YSCALE_YUV_2_GRAY16_1_C \
770 for (i=0; i<(dstW>>1); i++){\
771 const int i2= 2*i;\
772 int Y1= buf0[i2 ]<<1;\
773 int Y2= buf0[i2+1]<<1;\
774
775 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
776 YSCALE_YUV_2_PACKED1_C(type,alpha)\
777 r = (type *)c->table_rV[V];\
778 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
779 b = (type *)c->table_bU[U];\
780
781 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
782 for (i=0; i<(dstW>>1); i++){\
783 const int i2= 2*i;\
784 int Y1= buf0[i2 ]>>7;\
785 int Y2= buf0[i2+1]>>7;\
786 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
787 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
788 type av_unused *r, *b, *g;\
789 int av_unused A1, A2;\
790 if (alpha){\
791 A1= abuf0[i2 ]>>7;\
792 A2= abuf0[i2+1]>>7;\
793 }\
794
795 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
796 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
797 r = (type *)c->table_rV[V];\
798 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
799 b = (type *)c->table_bU[U];\
800
801 #define YSCALE_YUV_2_MONO2_C \
802 const uint8_t * const d128=dither_8x8_220[y&7];\
803 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
804 for (i=0; i<dstW-7; i+=8){\
805 int acc;\
806 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
807 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
808 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
809 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
810 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
811 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
812 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
813 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
814 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
815 dest++;\
816 }\
817
818
819 #define YSCALE_YUV_2_MONOX_C \
820 const uint8_t * const d128=dither_8x8_220[y&7];\
821 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
822 int acc=0;\
823 for (i=0; i<dstW-1; i+=2){\
824 int j;\
825 int Y1=1<<18;\
826 int Y2=1<<18;\
827 \
828 for (j=0; j<lumFilterSize; j++)\
829 {\
830 Y1 += lumSrc[j][i] * lumFilter[j];\
831 Y2 += lumSrc[j][i+1] * lumFilter[j];\
832 }\
833 Y1>>=19;\
834 Y2>>=19;\
835 if ((Y1|Y2)&256)\
836 {\
837 if (Y1>255) Y1=255;\
838 else if (Y1<0)Y1=0;\
839 if (Y2>255) Y2=255;\
840 else if (Y2<0)Y2=0;\
841 }\
842 acc+= acc + g[Y1+d128[(i+0)&7]];\
843 acc+= acc + g[Y2+d128[(i+1)&7]];\
844 if ((i&7)==6){\
845 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
846 dest++;\
847 }\
848 }
849
850
851 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
852 switch(c->dstFormat)\
853 {\
854 case PIX_FMT_RGBA:\
855 case PIX_FMT_BGRA:\
856 if (CONFIG_SMALL){\
857 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
858 func(uint32_t,needAlpha)\
859 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
860 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
861 }\
862 }else{\
863 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
864 func(uint32_t,1)\
865 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
866 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
867 }\
868 }else{\
869 func(uint32_t,0)\
870 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
871 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
872 }\
873 }\
874 }\
875 break;\
876 case PIX_FMT_ARGB:\
877 case PIX_FMT_ABGR:\
878 if (CONFIG_SMALL){\
879 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
880 func(uint32_t,needAlpha)\
881 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
882 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
883 }\
884 }else{\
885 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){\
886 func(uint32_t,1)\
887 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
888 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
889 }\
890 }else{\
891 func(uint32_t,0)\
892 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
893 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
894 }\
895 }\
896 } \
897 break;\
898 case PIX_FMT_RGB24:\
899 func(uint8_t,0)\
900 ((uint8_t*)dest)[0]= r[Y1];\
901 ((uint8_t*)dest)[1]= g[Y1];\
902 ((uint8_t*)dest)[2]= b[Y1];\
903 ((uint8_t*)dest)[3]= r[Y2];\
904 ((uint8_t*)dest)[4]= g[Y2];\
905 ((uint8_t*)dest)[5]= b[Y2];\
906 dest+=6;\
907 }\
908 break;\
909 case PIX_FMT_BGR24:\
910 func(uint8_t,0)\
911 ((uint8_t*)dest)[0]= b[Y1];\
912 ((uint8_t*)dest)[1]= g[Y1];\
913 ((uint8_t*)dest)[2]= r[Y1];\
914 ((uint8_t*)dest)[3]= b[Y2];\
915 ((uint8_t*)dest)[4]= g[Y2];\
916 ((uint8_t*)dest)[5]= r[Y2];\
917 dest+=6;\
918 }\
919 break;\
920 case PIX_FMT_RGB565:\
921 case PIX_FMT_BGR565:\
922 {\
923 const int dr1= dither_2x2_8[y&1 ][0];\
924 const int dg1= dither_2x2_4[y&1 ][0];\
925 const int db1= dither_2x2_8[(y&1)^1][0];\
926 const int dr2= dither_2x2_8[y&1 ][1];\
927 const int dg2= dither_2x2_4[y&1 ][1];\
928 const int db2= dither_2x2_8[(y&1)^1][1];\
929 func(uint16_t,0)\
930 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
931 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
932 }\
933 }\
934 break;\
935 case PIX_FMT_RGB555:\
936 case PIX_FMT_BGR555:\
937 {\
938 const int dr1= dither_2x2_8[y&1 ][0];\
939 const int dg1= dither_2x2_8[y&1 ][1];\
940 const int db1= dither_2x2_8[(y&1)^1][0];\
941 const int dr2= dither_2x2_8[y&1 ][1];\
942 const int dg2= dither_2x2_8[y&1 ][0];\
943 const int db2= dither_2x2_8[(y&1)^1][1];\
944 func(uint16_t,0)\
945 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
946 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
947 }\
948 }\
949 break;\
950 case PIX_FMT_RGB8:\
951 case PIX_FMT_BGR8:\
952 {\
953 const uint8_t * const d64= dither_8x8_73[y&7];\
954 const uint8_t * const d32= dither_8x8_32[y&7];\
955 func(uint8_t,0)\
956 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
957 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
958 }\
959 }\
960 break;\
961 case PIX_FMT_RGB4:\
962 case PIX_FMT_BGR4:\
963 {\
964 const uint8_t * const d64= dither_8x8_73 [y&7];\
965 const uint8_t * const d128=dither_8x8_220[y&7];\
966 func(uint8_t,0)\
967 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
968 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
969 }\
970 }\
971 break;\
972 case PIX_FMT_RGB4_BYTE:\
973 case PIX_FMT_BGR4_BYTE:\
974 {\
975 const uint8_t * const d64= dither_8x8_73 [y&7];\
976 const uint8_t * const d128=dither_8x8_220[y&7];\
977 func(uint8_t,0)\
978 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
979 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
980 }\
981 }\
982 break;\
983 case PIX_FMT_MONOBLACK:\
984 case PIX_FMT_MONOWHITE:\
985 {\
986 func_monoblack\
987 }\
988 break;\
989 case PIX_FMT_YUYV422:\
990 func2\
991 ((uint8_t*)dest)[2*i2+0]= Y1;\
992 ((uint8_t*)dest)[2*i2+1]= U;\
993 ((uint8_t*)dest)[2*i2+2]= Y2;\
994 ((uint8_t*)dest)[2*i2+3]= V;\
995 } \
996 break;\
997 case PIX_FMT_UYVY422:\
998 func2\
999 ((uint8_t*)dest)[2*i2+0]= U;\
1000 ((uint8_t*)dest)[2*i2+1]= Y1;\
1001 ((uint8_t*)dest)[2*i2+2]= V;\
1002 ((uint8_t*)dest)[2*i2+3]= Y2;\
1003 } \
1004 break;\
1005 case PIX_FMT_GRAY16BE:\
1006 func_g16\
1007 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
1008 ((uint8_t*)dest)[2*i2+1]= Y1;\
1009 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
1010 ((uint8_t*)dest)[2*i2+3]= Y2;\
1011 } \
1012 break;\
1013 case PIX_FMT_GRAY16LE:\
1014 func_g16\
1015 ((uint8_t*)dest)[2*i2+0]= Y1;\
1016 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
1017 ((uint8_t*)dest)[2*i2+2]= Y2;\
1018 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
1019 } \
1020 break;\
1021 }\
1022
1023
1024 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1025 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1026 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1027 {
1028 int i;
1029 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
1030 }
1031
1032 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1033 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1034 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
1035 {
1036 int i;
1037 int step= fmt_depth(c->dstFormat)/8;
1038 int aidx= 3;
1039
1040 switch(c->dstFormat){
1041 case PIX_FMT_ARGB:
1042 dest++;
1043 aidx= 0;
1044 case PIX_FMT_RGB24:
1045 aidx--;
1046 case PIX_FMT_RGBA:
1047 if (CONFIG_SMALL){
1048 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1049 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1050 dest[aidx]= needAlpha ? A : 255;
1051 dest[0]= R>>22;
1052 dest[1]= G>>22;
1053 dest[2]= B>>22;
1054 dest+= step;
1055 }
1056 }else{
1057 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1058 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1059 dest[aidx]= A;
1060 dest[0]= R>>22;
1061 dest[1]= G>>22;
1062 dest[2]= B>>22;
1063 dest+= step;
1064 }
1065 }else{
1066 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1067 dest[aidx]= 255;
1068 dest[0]= R>>22;
1069 dest[1]= G>>22;
1070 dest[2]= B>>22;
1071 dest+= step;
1072 }
1073 }
1074 }
1075 break;
1076 case PIX_FMT_ABGR:
1077 dest++;
1078 aidx= 0;
1079 case PIX_FMT_BGR24:
1080 aidx--;
1081 case PIX_FMT_BGRA:
1082 if (CONFIG_SMALL){
1083 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1084 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1085 dest[aidx]= needAlpha ? A : 255;
1086 dest[0]= B>>22;
1087 dest[1]= G>>22;
1088 dest[2]= R>>22;
1089 dest+= step;
1090 }
1091 }else{
1092 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
1093 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1094 dest[aidx]= A;
1095 dest[0]= B>>22;
1096 dest[1]= G>>22;
1097 dest[2]= R>>22;
1098 dest+= step;
1099 }
1100 }else{
1101 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1102 dest[aidx]= 255;
1103 dest[0]= B>>22;
1104 dest[1]= G>>22;
1105 dest[2]= R>>22;
1106 dest+= step;
1107 }
1108 }
1109 }
1110 break;
1111 default:
1112 assert(0);
1113 }
1114 }
1115
1116 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val){
1117 int i;
1118 uint8_t *ptr = plane + stride*y;
1119 for (i=0; i<height; i++){
1120 memset(ptr, val, width);
1121 ptr += stride;
1122 }
1123 }
1124
1125 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1126 //Plain C versions
1127 #if !HAVE_MMX || CONFIG_RUNTIME_CPUDETECT || !CONFIG_GPL
1128 #define COMPILE_C
1129 #endif
1130
1131 #if ARCH_PPC
1132 #if (HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1133 #undef COMPILE_C
1134 #define COMPILE_ALTIVEC
1135 #endif
1136 #endif //ARCH_PPC
1137
1138 #if ARCH_X86
1139
1140 #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1141 #define COMPILE_MMX
1142 #endif
1143
1144 #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1145 #define COMPILE_MMX2
1146 #endif
1147
1148 #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1149 #define COMPILE_3DNOW
1150 #endif
1151 #endif //ARCH_X86
1152
1153 #undef HAVE_MMX
1154 #undef HAVE_MMX2
1155 #undef HAVE_AMD3DNOW
1156 #undef HAVE_ALTIVEC
1157 #define HAVE_MMX 0
1158 #define HAVE_MMX2 0
1159 #define HAVE_AMD3DNOW 0
1160 #define HAVE_ALTIVEC 0
1161
1162 #ifdef COMPILE_C
1163 #define RENAME(a) a ## _C
1164 #include "swscale_template.c"
1165 #endif
1166
1167 #ifdef COMPILE_ALTIVEC
1168 #undef RENAME
1169 #undef HAVE_ALTIVEC
1170 #define HAVE_ALTIVEC 1
1171 #define RENAME(a) a ## _altivec
1172 #include "swscale_template.c"
1173 #endif
1174
1175 #if ARCH_X86
1176
1177 //MMX versions
1178 #ifdef COMPILE_MMX
1179 #undef RENAME
1180 #undef HAVE_MMX
1181 #undef HAVE_MMX2
1182 #undef HAVE_AMD3DNOW
1183 #define HAVE_MMX 1
1184 #define HAVE_MMX2 0
1185 #define HAVE_AMD3DNOW 0
1186 #define RENAME(a) a ## _MMX
1187 #include "swscale_template.c"
1188 #endif
1189
1190 //MMX2 versions
1191 #ifdef COMPILE_MMX2
1192 #undef RENAME
1193 #undef HAVE_MMX
1194 #undef HAVE_MMX2
1195 #undef HAVE_AMD3DNOW
1196 #define HAVE_MMX 1
1197 #define HAVE_MMX2 1
1198 #define HAVE_AMD3DNOW 0
1199 #define RENAME(a) a ## _MMX2
1200 #include "swscale_template.c"
1201 #endif
1202
1203 //3DNOW versions
1204 #ifdef COMPILE_3DNOW
1205 #undef RENAME
1206 #undef HAVE_MMX
1207 #undef HAVE_MMX2
1208 #undef HAVE_AMD3DNOW
1209 #define HAVE_MMX 1
1210 #define HAVE_MMX2 0
1211 #define HAVE_AMD3DNOW 1
1212 #define RENAME(a) a ## _3DNow
1213 #include "swscale_template.c"
1214 #endif
1215
1216 #endif //ARCH_X86
1217
1218 // minor note: the HAVE_xyz are messed up after this line so don't use them
1219
1220 static double getSplineCoeff(double a, double b, double c, double d, double dist)
1221 {
1222 // printf("%f %f %f %f %f\n", a,b,c,d,dist);
1223 if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
1224 else return getSplineCoeff( 0.0,
1225 b+ 2.0*c + 3.0*d,
1226 c + 3.0*d,
1227 -b- 3.0*c - 6.0*d,
1228 dist-1.0);
1229 }
1230
1231 static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
1232 int srcW, int dstW, int filterAlign, int one, int flags,
1233 SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
1234 {
1235 int i;
1236 int filterSize;
1237 int filter2Size;
1238 int minFilterSize;
1239 int64_t *filter=NULL;
1240 int64_t *filter2=NULL;
1241 const int64_t fone= 1LL<<54;
1242 int ret= -1;
1243 #if ARCH_X86
1244 if (flags & SWS_CPU_CAPS_MMX)
1245 __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
1246 #endif
1247
1248 // NOTE: the +1 is for the MMX scaler which reads over the end
1249 *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1250
1251 if (FFABS(xInc - 0x10000) <10) // unscaled
1252 {
1253 int i;
1254 filterSize= 1;
1255 filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
1256
1257 for (i=0; i<dstW; i++)
1258 {
1259 filter[i*filterSize]= fone;
1260 (*filterPos)[i]=i;
1261 }
1262
1263 }
1264 else if (flags&SWS_POINT) // lame looking point sampling mode
1265 {
1266 int i;
1267 int xDstInSrc;
1268 filterSize= 1;
1269 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1270
1271 xDstInSrc= xInc/2 - 0x8000;
1272 for (i=0; i<dstW; i++)
1273 {
1274 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1275
1276 (*filterPos)[i]= xx;
1277 filter[i]= fone;
1278 xDstInSrc+= xInc;
1279 }
1280 }
1281 else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
1282 {
1283 int i;
1284 int xDstInSrc;
1285 if (flags&SWS_BICUBIC) filterSize= 4;
1286 else if (flags&SWS_X ) filterSize= 4;
1287 else filterSize= 2; // SWS_BILINEAR / SWS_AREA
1288 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1289
1290 xDstInSrc= xInc/2 - 0x8000;
1291 for (i=0; i<dstW; i++)
1292 {
1293 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1294 int j;
1295
1296 (*filterPos)[i]= xx;
1297 //bilinear upscale / linear interpolate / area averaging
1298 for (j=0; j<filterSize; j++)
1299 {
1300 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1301 if (coeff<0) coeff=0;
1302 filter[i*filterSize + j]= coeff;
1303 xx++;
1304 }
1305 xDstInSrc+= xInc;
1306 }
1307 }
1308 else
1309 {
1310 int xDstInSrc;
1311 int sizeFactor;
1312
1313 if (flags&SWS_BICUBIC) sizeFactor= 4;
1314 else if (flags&SWS_X) sizeFactor= 8;
1315 else if (flags&SWS_AREA) sizeFactor= 1; //downscale only, for upscale it is bilinear
1316 else if (flags&SWS_GAUSS) sizeFactor= 8; // infinite ;)
1317 else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1318 else if (flags&SWS_SINC) sizeFactor= 20; // infinite ;)
1319 else if (flags&SWS_SPLINE) sizeFactor= 20; // infinite ;)
1320 else if (flags&SWS_BILINEAR) sizeFactor= 2;
1321 else {
1322 sizeFactor= 0; //GCC warning killer
1323 assert(0);
1324 }
1325
1326 if (xInc <= 1<<16) filterSize= 1 + sizeFactor; // upscale
1327 else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
1328
1329 if (filterSize > srcW-2) filterSize=srcW-2;
1330
1331 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
1332
1333 xDstInSrc= xInc - 0x10000;
1334 for (i=0; i<dstW; i++)
1335 {
1336 int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
1337 int j;
1338 (*filterPos)[i]= xx;
1339 for (j=0; j<filterSize; j++)
1340 {
1341 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1342 double floatd;
1343 int64_t coeff;
1344
1345 if (xInc > 1<<16)
1346 d= d*dstW/srcW;
1347 floatd= d * (1.0/(1<<30));
1348
1349 if (flags & SWS_BICUBIC)
1350 {
1351 int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1<<24);
1352 int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1353 int64_t dd = ( d*d)>>30;
1354 int64_t ddd= (dd*d)>>30;
1355
1356 if (d < 1LL<<30)
1357 coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1358 else if (d < 1LL<<31)
1359 coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
1360 else
1361 coeff=0.0;
1362 coeff *= fone>>(30+24);
1363 }
1364 /* else if (flags & SWS_X)
1365 {
1366 double p= param ? param*0.01 : 0.3;
1367 coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1368 coeff*= pow(2.0, - p*d*d);
1369 }*/
1370 else if (flags & SWS_X)
1371 {
1372 double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
1373 double c;
1374
1375 if (floatd<1.0)
1376 c = cos(floatd*PI);
1377 else
1378 c=-1.0;
1379 if (c<0.0) c= -pow(-c, A);
1380 else c= pow( c, A);
1381 coeff= (c*0.5 + 0.5)*fone;
1382 }
1383 else if (flags & SWS_AREA)
1384 {
1385 int64_t d2= d - (1<<29);
1386 if (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1387 else if (d2*xInc < (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
1388 else coeff=0.0;
1389 coeff *= fone>>(30+16);
1390 }
1391 else if (flags & SWS_GAUSS)
1392 {
1393 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1394 coeff = (pow(2.0, - p*floatd*floatd))*fone;
1395 }
1396 else if (flags & SWS_SINC)
1397 {
1398 coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
1399 }
1400 else if (flags & SWS_LANCZOS)
1401 {
1402 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
1403 coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1404 if (floatd>p) coeff=0;
1405 }
1406 else if (flags & SWS_BILINEAR)
1407 {
1408 coeff= (1<<30) - d;
1409 if (coeff<0) coeff=0;
1410 coeff *= fone >> 30;
1411 }
1412 else if (flags & SWS_SPLINE)
1413 {
1414 double p=-2.196152422706632;
1415 coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
1416 }
1417 else {
1418 coeff= 0.0; //GCC warning killer
1419 assert(0);
1420 }
1421
1422 filter[i*filterSize + j]= coeff;
1423 xx++;
1424 }
1425 xDstInSrc+= 2*xInc;
1426 }
1427 }
1428
1429 /* apply src & dst Filter to filter -> filter2
1430 av_free(filter);
1431 */
1432 assert(filterSize>0);
1433 filter2Size= filterSize;
1434 if (srcFilter) filter2Size+= srcFilter->length - 1;
1435 if (dstFilter) filter2Size+= dstFilter->length - 1;
1436 assert(filter2Size>0);
1437 filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
1438
1439 for (i=0; i<dstW; i++)
1440 {
1441 int j, k;
1442
1443 if(srcFilter){
1444 for (k=0; k<srcFilter->length; k++){
1445 for (j=0; j<filterSize; j++)
1446 filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1447 }
1448 }else{
1449 for (j=0; j<filterSize; j++)
1450 filter2[i*filter2Size + j]= filter[i*filterSize + j];
1451 }
1452 //FIXME dstFilter
1453
1454 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
1455 }
1456 av_freep(&filter);
1457
1458 /* try to reduce the filter-size (step1 find size and shift left) */
1459 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
1460 minFilterSize= 0;
1461 for (i=dstW-1; i>=0; i--)
1462 {
1463 int min= filter2Size;
1464 int j;
1465 int64_t cutOff=0.0;
1466
1467 /* get rid off near zero elements on the left by shifting left */
1468 for (j=0; j<filter2Size; j++)
1469 {
1470 int k;
1471 cutOff += FFABS(filter2[i*filter2Size]);
1472
1473 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1474
1475 /* preserve monotonicity because the core can't handle the filter otherwise */
1476 if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1477
1478 // move filter coefficients left
1479 for (k=1; k<filter2Size; k++)
1480 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
1481 filter2[i*filter2Size + k - 1]= 0;
1482 (*filterPos)[i]++;
1483 }
1484
1485 cutOff=0;
1486 /* count near zeros on the right */
1487 for (j=filter2Size-1; j>0; j--)
1488 {
1489 cutOff += FFABS(filter2[i*filter2Size + j]);
1490
1491 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
1492 min--;
1493 }
1494
1495 if (min>minFilterSize) minFilterSize= min;
1496 }
1497
1498 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1499 // we can handle the special case 4,
1500 // so we don't want to go to the full 8
1501 if (minFilterSize < 5)
1502 filterAlign = 4;
1503
1504 // We really don't want to waste our time
1505 // doing useless computation, so fall back on
1506 // the scalar C code for very small filters.
1507 // Vectorizing is worth it only if you have a
1508 // decent-sized vector.
1509 if (minFilterSize < 3)
1510 filterAlign = 1;
1511 }
1512
1513 if (flags & SWS_CPU_CAPS_MMX) {
1514 // special case for unscaled vertical filtering
1515 if (minFilterSize == 1 && filterAlign == 2)
1516 filterAlign= 1;
1517 }
1518
1519 assert(minFilterSize > 0);
1520 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
1521 assert(filterSize > 0);
1522 filter= av_malloc(filterSize*dstW*sizeof(*filter));
1523 if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
1524 goto error;
1525 *outFilterSize= filterSize;
1526
1527 if (flags&SWS_PRINT_INFO)
1528 av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1529 /* try to reduce the filter-size (step2 reduce it) */
1530 for (i=0; i<dstW; i++)
1531 {
1532 int j;
1533
1534 for (j=0; j<filterSize; j++)
1535 {
1536 if (j>=filter2Size) filter[i*filterSize + j]= 0;
1537 else filter[i*filterSize + j]= filter2[i*filter2Size + j];
1538 if((flags & SWS_BITEXACT) && j>=minFilterSize)
1539 filter[i*filterSize + j]= 0;
1540 }
1541 }
1542
1543
1544 //FIXME try to align filterPos if possible
1545
1546 //fix borders
1547 for (i=0; i<dstW; i++)
1548 {
1549 int j;
1550 if ((*filterPos)[i] < 0)
1551 {
1552 // move filter coefficients left to compensate for filterPos
1553 for (j=1; j<filterSize; j++)
1554 {
1555 int left= FFMAX(j + (*filterPos)[i], 0);
1556 filter[i*filterSize + left] += filter[i*filterSize + j];
1557 filter[i*filterSize + j]=0;
1558 }
1559 (*filterPos)[i]= 0;
1560 }
1561
1562 if ((*filterPos)[i] + filterSize > srcW)
1563 {
1564 int shift= (*filterPos)[i] + filterSize - srcW;
1565 // move filter coefficients right to compensate for filterPos
1566 for (j=filterSize-2; j>=0; j--)
1567 {
1568 int right= FFMIN(j + shift, filterSize-1);
1569 filter[i*filterSize +right] += filter[i*filterSize +j];
1570 filter[i*filterSize +j]=0;
1571 }
1572 (*filterPos)[i]= srcW - filterSize;
1573 }
1574 }
1575
1576 // Note the +1 is for the MMX scaler which reads over the end
1577 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1578 *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1579
1580 /* normalize & store in outFilter */
1581 for (i=0; i<dstW; i++)
1582 {
1583 int j;
1584 int64_t error=0;
1585 int64_t sum=0;
1586
1587 for (j=0; j<filterSize; j++)
1588 {
1589 sum+= filter[i*filterSize + j];
1590 }
1591 sum= (sum + one/2)/ one;
1592 for (j=0; j<*outFilterSize; j++)
1593 {
1594 int64_t v= filter[i*filterSize + j] + error;
1595 int intV= ROUNDED_DIV(v, sum);
1596 (*outFilter)[i*(*outFilterSize) + j]= intV;
1597 error= v - intV*sum;
1598 }
1599 }
1600
1601 (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
1602 for (i=0; i<*outFilterSize; i++)
1603 {
1604 int j= dstW*(*outFilterSize);
1605 (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1606 }
1607
1608 ret=0;
1609 error:
1610 av_free(filter);
1611 av_free(filter2);
1612 return ret;
1613 }
1614
1615 #ifdef COMPILE_MMX2
1616 static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits)
1617 {
1618 uint8_t *fragmentA;
1619 x86_reg imm8OfPShufW1A;
1620 x86_reg imm8OfPShufW2A;
1621 x86_reg fragmentLengthA;
1622 uint8_t *fragmentB;
1623 x86_reg imm8OfPShufW1B;
1624 x86_reg imm8OfPShufW2B;
1625 x86_reg fragmentLengthB;
1626 int fragmentPos;
1627
1628 int xpos, i;
1629
1630 // create an optimized horizontal scaling routine
1631
1632 //code fragment
1633
1634 __asm__ volatile(
1635 "jmp 9f \n\t"
1636 // Begin
1637 "0: \n\t"
1638 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
1639 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
1640 "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
1641 "punpcklbw %%mm7, %%mm1 \n\t"
1642 "punpcklbw %%mm7, %%mm0 \n\t"
1643 "pshufw $0xFF, %%mm1, %%mm1 \n\t"
1644 "1: \n\t"
1645 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1646 "2: \n\t"
1647 "psubw %%mm1, %%mm0 \n\t"
1648 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
1649 "pmullw %%mm3, %%mm0 \n\t"
1650 "psllw $7, %%mm1 \n\t"
1651 "paddw %%mm1, %%mm0 \n\t"
1652
1653 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1654
1655 "add $8, %%"REG_a" \n\t"
1656 // End
1657 "9: \n\t"
1658 // "int $3 \n\t"
1659 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
1660 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
1661 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
1662 "dec %1 \n\t"
1663 "dec %2 \n\t"
1664 "sub %0, %1 \n\t"
1665 "sub %0, %2 \n\t"
1666 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
1667 "sub %0, %3 \n\t"
1668
1669
1670 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1671 "=r" (fragmentLengthA)
1672 );
1673
1674 __asm__ volatile(
1675 "jmp 9f \n\t"
1676 // Begin
1677 "0: \n\t"
1678 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
1679 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
1680 "punpcklbw %%mm7, %%mm0 \n\t"
1681 "pshufw $0xFF, %%mm0, %%mm1 \n\t"
1682 "1: \n\t"
1683 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1684 "2: \n\t"
1685 "psubw %%mm1, %%mm0 \n\t"
1686 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
1687 "pmullw %%mm3, %%mm0 \n\t"
1688 "psllw $7, %%mm1 \n\t"
1689 "paddw %%mm1, %%mm0 \n\t"
1690
1691 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1692
1693 "add $8, %%"REG_a" \n\t"
1694 // End
1695 "9: \n\t"
1696 // "int $3 \n\t"
1697 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
1698 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
1699 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
1700 "dec %1 \n\t"
1701 "dec %2 \n\t"
1702 "sub %0, %1 \n\t"
1703 "sub %0, %2 \n\t"
1704 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
1705 "sub %0, %3 \n\t"
1706
1707
1708 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1709 "=r" (fragmentLengthB)
1710 );
1711
1712 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1713 fragmentPos=0;
1714
1715 for (i=0; i<dstW/numSplits; i++)
1716 {
1717 int xx=xpos>>16;
1718
1719 if ((i&3) == 0)
1720 {
1721 int a=0;
1722 int b=((xpos+xInc)>>16) - xx;
1723 int c=((xpos+xInc*2)>>16) - xx;
1724 int d=((xpos+xInc*3)>>16) - xx;
1725
1726 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9;
1727 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9;
1728 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1729 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1730 filterPos[i/2]= xx;
1731
1732 if (d+1<4)
1733 {
1734 int maxShift= 3-(d+1);
1735 int shift=0;
1736
1737 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);
1738
1739 funnyCode[fragmentPos + imm8OfPShufW1B]=
1740 (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
1741 funnyCode[fragmentPos + imm8OfPShufW2B]=
1742 a | (b<<2) | (c<<4) | (d<<6);
1743
1744 if (i+3>=dstW) shift=maxShift; //avoid overread
1745 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
1746
1747 if (shift && i>=shift)
1748 {
1749 funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
1750 funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
1751 filterPos[i/2]-=shift;
1752 }
1753
1754 fragmentPos+= fragmentLengthB;
1755 }
1756 else
1757 {
1758 int maxShift= 3-d;
1759 int shift=0;
1760
1761 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);
1762
1763 funnyCode[fragmentPos + imm8OfPShufW1A]=
1764 funnyCode[fragmentPos + imm8OfPShufW2A]=
1765 a | (b<<2) | (c<<4) | (d<<6);
1766
1767 if (i+4>=dstW) shift=maxShift; //avoid overread
1768 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align
1769
1770 if (shift && i>=shift)
1771 {
1772 funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
1773 funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
1774 filterPos[i/2]-=shift;
1775 }
1776
1777 fragmentPos+= fragmentLengthA;
1778 }
1779
1780 funnyCode[fragmentPos]= RET;
1781 }
1782 xpos+=xInc;
1783 }
1784 filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
1785 }
1786 #endif /* COMPILE_MMX2 */
1787
1788 static void globalInit(void){
1789 // generating tables:
1790 int i;
1791 for (i=0; i<768; i++){
1792 int c= av_clip_uint8(i-256);
1793 clip_table[i]=c;
1794 }
1795 }
1796
1797 static SwsFunc getSwsFunc(SwsContext *c)
1798 {
1799 #if CONFIG_RUNTIME_CPUDETECT && CONFIG_GPL
1800 int flags = c->flags;
1801
1802 #if ARCH_X86
1803 // ordered per speed fastest first
1804 if (flags & SWS_CPU_CAPS_MMX2) {
1805 sws_init_swScale_MMX2(c);
1806 return swScale_MMX2;
1807 } else if (flags & SWS_CPU_CAPS_3DNOW) {
1808 sws_init_swScale_3DNow(c);
1809 return swScale_3DNow;
1810 } else if (flags & SWS_CPU_CAPS_MMX) {
1811 sws_init_swScale_MMX(c);
1812 return swScale_MMX;
1813 } else {
1814 sws_init_swScale_C(c);
1815 return swScale_C;
1816 }
1817
1818 #else
1819 #if ARCH_PPC
1820 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1821 sws_init_swScale_altivec(c);
1822 return swScale_altivec;
1823 } else {
1824 sws_init_swScale_C(c);
1825 return swScale_C;
1826 }
1827 #endif
1828 sws_init_swScale_C(c);
1829 return swScale_C;
1830 #endif /* ARCH_X86 */
1831 #else //CONFIG_RUNTIME_CPUDETECT
1832 #if HAVE_MMX2
1833 sws_init_swScale_MMX2(c);
1834 return swScale_MMX2;
1835 #elif HAVE_AMD3DNOW
1836 sws_init_swScale_3DNow(c);
1837 return swScale_3DNow;
1838 #elif HAVE_MMX
1839 sws_init_swScale_MMX(c);
1840 return swScale_MMX;
1841 #elif HAVE_ALTIVEC
1842 sws_init_swScale_altivec(c);
1843 return swScale_altivec;
1844 #else
1845 sws_init_swScale_C(c);
1846 return swScale_C;
1847 #endif
1848 #endif //!CONFIG_RUNTIME_CPUDETECT
1849 }
1850
1851 static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1852 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1853 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1854 /* Copy Y plane */
1855 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1856 memcpy(dst, src[0], srcSliceH*dstStride[0]);
1857 else
1858 {
1859 int i;
1860 const uint8_t *srcPtr= src[0];
1861 uint8_t *dstPtr= dst;
1862 for (i=0; i<srcSliceH; i++)
1863 {
1864 memcpy(dstPtr, srcPtr, c->srcW);
1865 srcPtr+= srcStride[0];
1866 dstPtr+= dstStride[0];
1867 }
1868 }
1869 dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1870 if (c->dstFormat == PIX_FMT_NV12)
1871 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1872 else
1873 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1874
1875 return srcSliceH;
1876 }
1877
1878 static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1879 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1880 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1881
1882 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1883
1884 return srcSliceH;
1885 }
1886
1887 static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1888 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1889 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1890
1891 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1892
1893 return srcSliceH;
1894 }
1895
1896 static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1897 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1898 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1899
1900 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1901
1902 return srcSliceH;
1903 }
1904
1905 static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1906 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1907 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1908
1909 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1910
1911 return srcSliceH;
1912 }
1913
1914 static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1915 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1916 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1917 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1918 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1919
1920 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1921
1922 if (dstParam[3])
1923 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1924
1925 return srcSliceH;
1926 }
1927
1928 static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1929 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1930 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1931 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1932 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1933
1934 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1935
1936 return srcSliceH;
1937 }
1938
1939 static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1940 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1941 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1942 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1943 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1944
1945 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1946
1947 if (dstParam[3])
1948 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1949
1950 return srcSliceH;
1951 }
1952
1953 static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1954 int srcSliceH, uint8_t* dstParam[], int dstStride[]){
1955 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1956 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1957 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1958
1959 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1960
1961 return srcSliceH;
1962 }
1963
1964 static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
1965 int srcSliceH, uint8_t* dst[], int dstStride[]){
1966 const enum PixelFormat srcFormat= c->srcFormat;
1967 const enum PixelFormat dstFormat= c->dstFormat;
1968 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1969 const uint8_t *palette)=NULL;
1970 int i;
1971 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1972 uint8_t *srcPtr= src[0];
1973
1974 if (!usePal(srcFormat))
1975 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1976 sws_format_name(srcFormat), sws_format_name(dstFormat));
1977
1978 switch(dstFormat){
1979 case PIX_FMT_RGB32 : conv = palette8topacked32; break;
1980 case PIX_FMT_BGR32 : conv = palette8topacked32; break;
1981 case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1982 case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1983 case PIX_FMT_RGB24 : conv = palette8topacked24; break;
1984 case PIX_FMT_BGR24 : conv = palette8topacked24; break;
1985 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1986 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
1987 }
1988
1989
1990 for (i=0; i<srcSliceH; i++) {
1991 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1992 srcPtr+= srcStride[0];
1993 dstPtr+= dstStride[0];
1994 }
1995
1996 return srcSliceH;
1997 }
1998
1999 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
2000 static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2001 int srcSliceH, uint8_t* dst[], int dstStride[]){
2002 const enum PixelFormat srcFormat= c->srcFormat;
2003 const enum PixelFormat dstFormat= c->dstFormat;
2004 const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2005 const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2006 const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2007 const int dstId= fmt_depth(dstFormat) >> 2;
2008 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2009
2010 /* BGR -> BGR */
2011 if ( (isBGR(srcFormat) && isBGR(dstFormat))
2012 || (isRGB(srcFormat) && isRGB(dstFormat))){
2013 switch(srcId | (dstId<<4)){
2014 case 0x34: conv= rgb16to15; break;
2015 case 0x36: conv= rgb24to15; break;
2016 case 0x38: conv= rgb32to15; break;
2017 case 0x43: conv= rgb15to16; break;
2018 case 0x46: conv= rgb24to16; break;
2019 case 0x48: conv= rgb32to16; break;
2020 case 0x63: conv= rgb15to24; break;
2021 case 0x64: conv= rgb16to24; break;
2022 case 0x68: conv= rgb32to24; break;
2023 case 0x83: conv= rgb15to32; break;
2024 case 0x84: conv= rgb16to32; break;
2025 case 0x86: conv= rgb24to32; break;
2026 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2027 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2028 }
2029 }else if ( (isBGR(srcFormat) && isRGB(dstFormat))
2030 || (isRGB(srcFormat) && isBGR(dstFormat))){
2031 switch(srcId | (dstId<<4)){
2032 case 0x33: conv= rgb15tobgr15; break;
2033 case 0x34: conv= rgb16tobgr15; break;
2034 case 0x36: conv= rgb24tobgr15; break;
2035 case 0x38: conv= rgb32tobgr15; break;
2036 case 0x43: conv= rgb15tobgr16; break;
2037 case 0x44: conv= rgb16tobgr16; break;
2038 case 0x46: conv= rgb24tobgr16; break;
2039 case 0x48: conv= rgb32tobgr16; break;
2040 case 0x63: conv= rgb15tobgr24; break;
2041 case 0x64: conv= rgb16tobgr24; break;
2042 case 0x66: conv= rgb24tobgr24; break;
2043 case 0x68: conv= rgb32tobgr24; break;
2044 case 0x83: conv= rgb15tobgr32; break;
2045 case 0x84: conv= rgb16tobgr32; break;
2046 case 0x86: conv= rgb24tobgr32; break;
2047 case 0x88: conv= rgb32tobgr32; break;
2048 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2049 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2050 }
2051 }else{
2052 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2053 sws_format_name(srcFormat), sws_format_name(dstFormat));
2054 }
2055
2056 if(conv)
2057 {
2058 uint8_t *srcPtr= src[0];
2059 if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2060 srcPtr += ALT32_CORR;
2061
2062 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
2063 conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
2064 else
2065 {
2066 int i;
2067 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2068
2069 for (i=0; i<srcSliceH; i++)
2070 {
2071 conv(srcPtr, dstPtr, c->srcW*srcBpp);
2072 srcPtr+= srcStride[0];
2073 dstPtr+= dstStride[0];
2074 }
2075 }
2076 }
2077 return srcSliceH;
2078 }
2079
2080 static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2081 int srcSliceH, uint8_t* dst[], int dstStride[]){
2082
2083 rgb24toyv12(
2084 src[0],
2085 dst[0]+ srcSliceY *dstStride[0],
2086 dst[1]+(srcSliceY>>1)*dstStride[1],
2087 dst[2]+(srcSliceY>>1)*dstStride[2],
2088 c->srcW, srcSliceH,
2089 dstStride[0], dstStride[1], srcStride[0]);
2090 if (dst[3])
2091 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2092 return srcSliceH;
2093 }
2094
2095 static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2096 int srcSliceH, uint8_t* dst[], int dstStride[]){
2097 int i;
2098
2099 /* copy Y */
2100 if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2101 memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
2102 else{
2103 uint8_t *srcPtr= src[0];
2104 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2105
2106 for (i=0; i<srcSliceH; i++)
2107 {
2108 memcpy(dstPtr, srcPtr, c->srcW);
2109 srcPtr+= srcStride[0];
2110 dstPtr+= dstStride[0];
2111 }
2112 }
2113
2114 if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P){
2115 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2116 srcSliceH >> 2, srcStride[1], dstStride[1]);
2117 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2118 srcSliceH >> 2, srcStride[2], dstStride[2]);
2119 }else{
2120 planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2121 srcSliceH >> 2, srcStride[1], dstStride[2]);
2122 planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2123 srcSliceH >> 2, srcStride[2], dstStride[1]);
2124 }
2125 if (dst[3])
2126 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2127 return srcSliceH;
2128 }
2129
2130 /* unscaled copy like stuff (assumes nearly identical formats) */
2131 static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2132 int srcSliceH, uint8_t* dst[], int dstStride[])
2133 {
2134 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2135 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
2136 else
2137 {
2138 int i;
2139 uint8_t *srcPtr= src[0];
2140 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2141 int length=0;
2142
2143 /* universal length finder */
2144 while(length+c->srcW <= FFABS(dstStride[0])
2145 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
2146 assert(length!=0);
2147
2148 for (i=0; i<srcSliceH; i++)
2149 {
2150 memcpy(dstPtr, srcPtr, length);
2151 srcPtr+= srcStride[0];
2152 dstPtr+= dstStride[0];
2153 }
2154 }
2155 return srcSliceH;
2156 }
2157
2158 static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2159 int srcSliceH, uint8_t* dst[], int dstStride[])
2160 {
2161 int plane, i, j;
2162 for (plane=0; plane<4; plane++)
2163 {
2164 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
2165 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2166 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
2167 uint8_t *srcPtr= src[plane];
2168 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2169
2170 if (!dst[plane]) continue;
2171 // ignore palette for GRAY8
2172 if (plane == 1 && !dst[2]) continue;
2173 if (!src[plane] || (plane == 1 && !src[2])){
2174 if(is16BPS(c->dstFormat))
2175 length*=2;
2176 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
2177 }else
2178 {
2179 if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)){
2180 if (!isBE(c->srcFormat)) srcPtr++;
2181 for (i=0; i<height; i++){
2182 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2183 srcPtr+= srcStride[plane];
2184 dstPtr+= dstStride[plane];
2185 }
2186 }else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)){
2187 for (i=0; i<height; i++){
2188 for (j=0; j<length; j++){
2189 dstPtr[ j<<1 ] = srcPtr[j];
2190 dstPtr[(j<<1)+1] = srcPtr[j];
2191 }
2192 srcPtr+= srcStride[plane];
2193 dstPtr+= dstStride[plane];
2194 }
2195 }else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2196 && isBE(c->srcFormat) != isBE(c->dstFormat)){
2197
2198 for (i=0; i<height; i++){
2199 for (j=0; j<length; j++)
2200 ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2201 srcPtr+= srcStride[plane];
2202 dstPtr+= dstStride[plane];
2203 }
2204 } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
2205 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
2206 else
2207 {
2208 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2209 length*=2;
2210 for (i=0; i<height; i++)
2211 {
2212 memcpy(dstPtr, srcPtr, length);
2213 srcPtr+= srcStride[plane];
2214 dstPtr+= dstStride[plane];
2215 }
2216 }
2217 }
2218 }
2219 return srcSliceH;
2220 }
2221
2222
2223 static void getSubSampleFactors(int *h, int *v, int format){
2224 switch(format){
2225 case PIX_FMT_UYVY422:
2226 case PIX_FMT_YUYV422:
2227 *h=1;
2228 *v=0;
2229 break;
2230 case PIX_FMT_YUV420P:
2231 case PIX_FMT_YUV420PLE:
2232 case PIX_FMT_YUV420PBE:
2233 case PIX_FMT_YUVA420P:
2234 case PIX_FMT_GRAY16BE:
2235 case PIX_FMT_GRAY16LE:
2236 case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2237 case PIX_FMT_NV12:
2238 case PIX_FMT_NV21:
2239 *h=1;
2240 *v=1;
2241 break;
2242 case PIX_FMT_YUV440P:
2243 *h=0;
2244 *v=1;
2245 break;
2246 case PIX_FMT_YUV410P:
2247 *h=2;
2248 *v=2;
2249 break;
2250 case PIX_FMT_YUV444P:
2251 case PIX_FMT_YUV444PLE:
2252 case PIX_FMT_YUV444PBE:
2253 *h=0;
2254 *v=0;
2255 break;
2256 case PIX_FMT_YUV422P:
2257 case PIX_FMT_YUV422PLE:
2258 case PIX_FMT_YUV422PBE:
2259 *h=1;
2260 *v=0;
2261 break;
2262 case PIX_FMT_YUV411P:
2263 *h=2;
2264 *v=0;
2265 break;
2266 default:
2267 *h=0;
2268 *v=0;
2269 break;
2270 }
2271 }
2272
2273 static uint16_t roundToInt16(int64_t f){
2274 int r= (f + (1<<15))>>16;
2275 if (r<-0x7FFF) return 0x8000;
2276 else if (r> 0x7FFF) return 0x7FFF;
2277 else return r;
2278 }
2279
2280 int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation){
2281 int64_t crv = inv_table[0];
2282 int64_t cbu = inv_table[1];
2283 int64_t cgu = -inv_table[2];
2284 int64_t cgv = -inv_table[3];
2285 int64_t cy = 1<<16;
2286 int64_t oy = 0;
2287
2288 memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2289 memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
2290
2291 c->brightness= brightness;
2292 c->contrast = contrast;
2293 c->saturation= saturation;
2294 c->srcRange = srcRange;
2295 c->dstRange = dstRange;
2296 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2297
2298 c->uOffset= 0x0400040004000400LL;
2299 c->vOffset= 0x0400040004000400LL;
2300
2301 if (!srcRange){
2302 cy= (cy*255) / 219;
2303 oy= 16<<16;
2304 }else{
2305 crv= (crv*224) / 255;
2306 cbu= (cbu*224) / 255;
2307 cgu= (cgu*224) / 255;
2308 cgv= (cgv*224) / 255;
2309 }
2310
2311 cy = (cy *contrast )>>16;
2312 crv= (crv*contrast * saturation)>>32;
2313 cbu= (cbu*contrast * saturation)>>32;
2314 cgu= (cgu*contrast * saturation)>>32;
2315 cgv= (cgv*contrast * saturation)>>32;
2316
2317 oy -= 256*brightness;
2318
2319 c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL;
2320 c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL;
2321 c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2322 c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2323 c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2324 c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL;
2325
2326 c->yuv2rgb_y_coeff = (int16_t)roundToInt16(cy <<13);
2327 c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2328 c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2329 c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2330 c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2331 c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
2332
2333 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
2334 //FIXME factorize
2335
2336 #ifdef COMPILE_ALTIVEC
2337 if (c->flags & SWS_CPU_CAPS_ALTIVEC)
2338 ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
2339 #endif
2340 return 0;
2341 }
2342
2343 int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation){
2344 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
2345
2346 *inv_table = c->srcColorspaceTable;
2347 *table = c->dstColorspaceTable;
2348 *srcRange = c->srcRange;
2349 *dstRange = c->dstRange;
2350 *brightness= c->brightness;
2351 *contrast = c->contrast;
2352 *saturation= c->saturation;
2353
2354 return 0;
2355 }
2356
2357 static int handle_jpeg(enum PixelFormat *format)
2358 {
2359 switch (*format) {
2360 case PIX_FMT_YUVJ420P:
2361 *format = PIX_FMT_YUV420P;
2362 return 1;
2363 case PIX_FMT_YUVJ422P:
2364 *format = PIX_FMT_YUV422P;
2365 return 1;
2366 case PIX_FMT_YUVJ444P:
2367 *format = PIX_FMT_YUV444P;
2368 return 1;
2369 case PIX_FMT_YUVJ440P:
2370 *format = PIX_FMT_YUV440P;
2371 return 1;
2372 default:
2373 return 0;
2374 }
2375 }
2376
2377 SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
2378 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2379 {
2380
2381 SwsContext *c;
2382 int i;
2383 int usesVFilter, usesHFilter;
2384 int unscaled, needsDither;
2385 int srcRange, dstRange;
2386 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
2387 #if ARCH_X86
2388 if (flags & SWS_CPU_CAPS_MMX)
2389 __asm__ volatile("emms\n\t"::: "memory");
2390 #endif
2391
2392 #if !CONFIG_RUNTIME_CPUDETECT || !CONFIG_GPL //ensure that the flags match the compiled variant if cpudetect is off
2393 flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
2394 #if HAVE_MMX2
2395 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
2396 #elif HAVE_AMD3DNOW
2397 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
2398 #elif HAVE_MMX
2399 flags |= SWS_CPU_CAPS_MMX;
2400 #elif HAVE_ALTIVEC
2401 flags |= SWS_CPU_CAPS_ALTIVEC;
2402 #elif ARCH_BFIN
2403 flags |= SWS_CPU_CAPS_BFIN;
2404 #endif
2405 #endif /* CONFIG_RUNTIME_CPUDETECT */
2406 if (clip_table[512] != 255) globalInit();
2407 if (!rgb15to16) sws_rgb2rgb_init(flags);
2408
2409 unscaled = (srcW == dstW && srcH == dstH);
2410 needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2411 && (fmt_depth(dstFormat))<24
2412 && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2413
2414 srcRange = handle_jpeg(&srcFormat);
2415 dstRange = handle_jpeg(&dstFormat);
2416
2417 if (!isSupportedIn(srcFormat))
2418 {
2419 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
2420 return NULL;
2421 }
2422 if (!isSupportedOut(dstFormat))
2423 {
2424 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
2425 return NULL;
2426 }
2427
2428 i= flags & ( SWS_POINT
2429 |SWS_AREA
2430 |SWS_BILINEAR
2431 |SWS_FAST_BILINEAR
2432 |SWS_BICUBIC
2433 |SWS_X
2434 |SWS_GAUSS
2435 |SWS_LANCZOS
2436 |SWS_SINC
2437 |SWS_SPLINE
2438 |SWS_BICUBLIN);
2439 if(!i || (i & (i-1)))
2440 {
2441 av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
2442 return NULL;
2443 }
2444
2445 /* sanity check */
2446 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
2447 {
2448 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2449 srcW, srcH, dstW, dstH);
2450 return NULL;
2451 }
2452 if(srcW > VOFW || dstW > VOFW){
2453 av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
2454 return NULL;
2455 }
2456
2457 if (!dstFilter) dstFilter= &dummyFilter;
2458 if (!srcFilter) srcFilter= &dummyFilter;
2459
2460 c= av_mallocz(sizeof(SwsContext));
2461
2462 c->av_class = &sws_context_class;
2463 c->srcW= srcW;
2464 c->srcH= srcH;
2465 c->dstW= dstW;
2466 c->dstH= dstH;
2467 c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2468 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2469 c->flags= flags;
2470 c->dstFormat= dstFormat;
2471 c->srcFormat= srcFormat;
2472 c->vRounder= 4* 0x0001000100010001ULL;
2473
2474 usesHFilter= usesVFilter= 0;
2475 if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2476 if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2477 if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2478 if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2479 if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2480 if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2481 if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2482 if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
2483
2484 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2485 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2486
2487 // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
2488 if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2489
2490 // drop some chroma lines if the user wants it
2491 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2492 c->chrSrcVSubSample+= c->vChrDrop;
2493
2494 // drop every other pixel for chroma calculation unless user wants full chroma
2495 if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2496 && srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8
2497 && srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4
2498 && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2499 && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2500 c->chrSrcHSubSample=1;
2501
2502 if (param){
2503 c->param[0] = param[0];
2504 c->param[1] = param[1];
2505 }else{
2506 c->param[0] =
2507 c->param[1] = SWS_PARAM_DEFAULT;
2508 }
2509
2510 c->chrIntHSubSample= c->chrDstHSubSample;
2511 c->chrIntVSubSample= c->chrSrcVSubSample;
2512
2513 // Note the -((-x)>>y) is so that we always round toward +inf.
2514 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2515 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2516 c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2517 c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2518
2519 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
2520
2521 /* unscaled special cases */
2522 if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat)))
2523 {
2524 /* yv12_to_nv12 */
2525 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21))
2526 {
2527 c->swScale= PlanarToNV12Wrapper;
2528 }
2529 /* yuv2bgr */
2530 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
2531 && !(flags & SWS_ACCURATE_RND) && !(dstH&1))
2532 {
2533 c->swScale= ff_yuv2rgb_get_func_ptr(c);
2534 }
2535
2536 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT))
2537 {
2538 c->swScale= yvu9toyv12Wrapper;
2539 }
2540
2541 /* bgr24toYV12 */
2542 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
2543 c->swScale= bgr24toyv12Wrapper;
2544
2545 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
2546 if ( (isBGR(srcFormat) || isRGB(srcFormat))
2547 && (isBGR(dstFormat) || isRGB(dstFormat))
2548 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
2549 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
2550 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
2551 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
2552 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2553 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2554 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
2555 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
2556 && dstFormat != PIX_FMT_RGB32_1
2557 && dstFormat != PIX_FMT_BGR32_1
2558 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
2559 c->swScale= rgb2rgbWrapper;
2560
2561 if ((usePal(srcFormat) && (
2562 dstFormat == PIX_FMT_RGB32 ||
2563 dstFormat == PIX_FMT_RGB32_1 ||
2564 dstFormat == PIX_FMT_RGB24 ||
2565 dstFormat == PIX_FMT_BGR32 ||
2566 dstFormat == PIX_FMT_BGR32_1 ||
2567 dstFormat == PIX_FMT_BGR24)))
2568 c->swScale= pal2rgbWrapper;
2569
2570 if (srcFormat == PIX_FMT_YUV422P)
2571 {
2572 if (dstFormat == PIX_FMT_YUYV422)
2573 c->swScale= YUV422PToYuy2Wrapper;
2574 else if (dstFormat == PIX_FMT_UYVY422)
2575 c->swScale= YUV422PToUyvyWrapper;
2576 }
2577
2578 /* LQ converters if -sws 0 or -sws 4*/
2579 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
2580 /* yv12_to_yuy2 */
2581 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P)
2582 {
2583 if (dstFormat == PIX_FMT_YUYV422)
2584 c->swScale= PlanarToYuy2Wrapper;
2585 else if (dstFormat == PIX_FMT_UYVY422)
2586 c->swScale= PlanarToUyvyWrapper;
2587 }
2588 }
2589 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2590 c->swScale= YUYV2YUV420Wrapper;
2591 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
2592 c->swScale= UYVY2YUV420Wrapper;
2593 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2594 c->swScale= YUYV2YUV422Wrapper;
2595 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2596 c->swScale= UYVY2YUV422Wrapper;
2597
2598 #ifdef COMPILE_ALTIVEC
2599 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
2600 !(c->flags & SWS_BITEXACT) &&
2601 srcFormat == PIX_FMT_YUV420P) {
2602 // unscaled YV12 -> packed YUV, we want speed
2603 if (dstFormat == PIX_FMT_YUYV422)
2604 c->swScale= yv12toyuy2_unscaled_altivec;
2605 else if (dstFormat == PIX_FMT_UYVY422)
2606 c->swScale= yv12touyvy_unscaled_altivec;
2607 }
2608 #endif
2609
2610 /* simple copy */
2611 if ( srcFormat == dstFormat
2612 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
2613 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
2614 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
2615 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
2616 || (isGray(dstFormat) && isGray(srcFormat))
2617 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2618 && c->chrDstHSubSample == c->chrSrcHSubSample
2619 && c->chrDstVSubSample == c->chrSrcVSubSample))
2620 {
2621 if (isPacked(c->srcFormat))
2622 c->swScale= packedCopy;
2623 else /* Planar YUV or gray */
2624 c->swScale= planarCopy;
2625 }
2626 #if ARCH_BFIN
2627 if (flags & SWS_CPU_CAPS_BFIN)
2628 ff_bfin_get_unscaled_swscale (c);
2629 #endif
2630
2631 if (c->swScale){
2632 if (flags&SWS_PRINT_INFO)
2633 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
2634 sws_format_name(srcFormat), sws_format_name(dstFormat));
2635 return c;
2636 }
2637 }
2638
2639 if (flags & SWS_CPU_CAPS_MMX2)
2640 {
2641 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
2642 if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
2643 {
2644 if (flags&SWS_PRINT_INFO)
2645 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
2646 }
2647 if (usesHFilter) c->canMMX2BeUsed=0;
2648 }
2649 else
2650 c->canMMX2BeUsed=0;
2651
2652 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2653 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2654
2655 // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2656 // but only for the FAST_BILINEAR mode otherwise do correct scaling
2657 // n-2 is the last chrominance sample available
2658 // this is not perfect, but no one should notice the difference, the more correct variant
2659 // would be like the vertical one, but that would require some special code for the
2660 // first and last pixel
2661 if (flags&SWS_FAST_BILINEAR)
2662 {
2663 if (c->canMMX2BeUsed)
2664 {
2665 c->lumXInc+= 20;
2666 c->chrXInc+= 20;
2667 }
2668 //we don't use the x86 asm scaler if MMX is available
2669 else if (flags & SWS_CPU_CAPS_MMX)
2670 {
2671 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2672 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2673 }
2674 }
2675
2676 /* precalculate horizontal scaler filter coefficients */
2677 {
2678 const int filterAlign=
2679 (flags & SWS_CPU_CAPS_MMX) ? 4 :
2680 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2681 1;
2682
2683 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2684 srcW , dstW, filterAlign, 1<<14,
2685 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
2686 srcFilter->lumH, dstFilter->lumH, c->param);
2687 initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2688 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2689 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2690 srcFilter->chrH, dstFilter->chrH, c->param);
2691
2692 #define MAX_FUNNY_CODE_SIZE 10000
2693 #if defined(COMPILE_MMX2)
2694 // can't downscale !!!
2695 if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
2696 {
2697 #ifdef MAP_ANONYMOUS
2698 c->funnyYCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2699 c->funnyUVCode = mmap(NULL, MAX_FUNNY_CODE_SIZE, PROT_EXEC | PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2700 #elif HAVE_VIRTUALALLOC
2701 c->funnyYCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2702 c->funnyUVCode = VirtualAlloc(NULL, MAX_FUNNY_CODE_SIZE, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2703 #else
2704 c->funnyYCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2705 c->funnyUVCode = av_malloc(MAX_FUNNY_CODE_SIZE);
2706 #endif
2707
2708 c->lumMmx2Filter = av_malloc((dstW /8+8)*sizeof(int16_t));
2709 c->chrMmx2Filter = av_malloc((c->chrDstW /4+8)*sizeof(int16_t));
2710 c->lumMmx2FilterPos= av_malloc((dstW /2/8+8)*sizeof(int32_t));
2711 c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
2712
2713 initMMX2HScaler( dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2714 initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
2715 }
2716 #endif /* defined(COMPILE_MMX2) */
2717 } // initialize horizontal stuff
2718
2719
2720
2721 /* precalculate vertical scaler filter coefficients */
2722 {
2723 const int filterAlign=
2724 (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2725 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2726 1;
2727
2728 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
2729 srcH , dstH, filterAlign, (1<<12),
2730 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
2731 srcFilter->lumV, dstFilter->lumV, c->param);
2732 initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
2733 c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
2734 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2735 srcFilter->chrV, dstFilter->chrV, c->param);
2736
2737 #if HAVE_ALTIVEC
2738 c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2739 c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2740
2741 for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2742 int j;
2743 short *p = (short *)&c->vYCoeffsBank[i];
2744 for (j=0;j<8;j++)
2745 p[j] = c->vLumFilter[i];
2746 }
2747
2748 for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2749 int j;
2750 short *p = (short *)&c->vCCoeffsBank[i];
2751 for (j=0;j<8;j++)
2752 p[j] = c->vChrFilter[i];
2753 }
2754 #endif
2755 }
2756
2757 // calculate buffer sizes so that they won't run out while handling these damn slices
2758 c->vLumBufSize= c->vLumFilterSize;
2759 c->vChrBufSize= c->vChrFilterSize;
2760 for (i=0; i<dstH; i++)
2761 {
2762 int chrI= i*c->chrDstH / dstH;
2763 int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
2764 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2765
2766 nextSlice>>= c->chrSrcVSubSample;
2767 nextSlice<<= c->chrSrcVSubSample;
2768 if (c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice)
2769 c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
2770 if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2771 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2772 }
2773
2774 // allocate pixbufs (we use dynamic allocation because otherwise we would need to
2775 c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2776 c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
2777 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2778 c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2779 //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)
2780 /* align at 16 bytes for AltiVec */
2781 for (i=0; i<c->vLumBufSize; i++)
2782 c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2783 for (i=0; i<c->vChrBufSize; i++)
2784 c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
2785 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2786 for (i=0; i<c->vLumBufSize; i++)
2787 c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
2788
2789 //try to avoid drawing green stuff between the right end and the stride end
2790 for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2791
2792 assert(2*VOFW == VOF);
2793
2794 assert(c->chrDstH <= dstH);
2795
2796 if (flags&SWS_PRINT_INFO)
2797 {
2798 #ifdef DITHER1XBPP
2799 const char *dither= " dithered";
2800 #else
2801 const char *dither= "";
2802 #endif
2803 if (flags&SWS_FAST_BILINEAR)
2804 av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
2805 else if (flags&SWS_BILINEAR)
2806 av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
2807 else if (flags&SWS_BICUBIC)
2808 av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
2809 else if (flags&SWS_X)
2810 av_log(c, AV_LOG_INFO, "Experimental scaler, ");
2811 else if (flags&SWS_POINT)
2812 av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
2813 else if (flags&SWS_AREA)
2814 av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
2815 else if (flags&SWS_BICUBLIN)
2816 av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
2817 else if (flags&SWS_GAUSS)
2818 av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
2819 else if (flags&SWS_SINC)
2820 av_log(c, AV_LOG_INFO, "Sinc scaler, ");
2821 else if (flags&SWS_LANCZOS)
2822 av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
2823 else if (flags&SWS_SPLINE)
2824 av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
2825 else
2826 av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
2827
2828 if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2829 av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2830 sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2831 else
2832 av_log(c, AV_LOG_INFO, "from %s to %s ",
2833 sws_format_name(srcFormat), sws_format_name(dstFormat));
2834
2835 if (flags & SWS_CPU_CAPS_MMX2)
2836 av_log(c, AV_LOG_INFO, "using MMX2\n");
2837 else if (flags & SWS_CPU_CAPS_3DNOW)
2838 av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2839 else if (flags & SWS_CPU_CAPS_MMX)
2840 av_log(c, AV_LOG_INFO, "using MMX\n");
2841 else if (flags & SWS_CPU_CAPS_ALTIVEC)
2842 av_log(c, AV_LOG_INFO, "using AltiVec\n");
2843 else
2844 av_log(c, AV_LOG_INFO, "using C\n");
2845 }
2846
2847 if (flags & SWS_PRINT_INFO)
2848 {
2849 if (flags & SWS_CPU_CAPS_MMX)
2850 {
2851 if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
2852 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
2853 else
2854 {
2855 if (c->hLumFilterSize==4)
2856 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
2857 else if (c->hLumFilterSize==8)
2858 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
2859 else
2860 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
2861
2862 if (c->hChrFilterSize==4)
2863 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
2864 else if (c->hChrFilterSize==8)
2865 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
2866 else
2867 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
2868 }
2869 }
2870 else
2871 {
2872 #if ARCH_X86
2873 av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
2874 #else
2875 if (flags & SWS_FAST_BILINEAR)
2876 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
2877 else
2878 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
2879 #endif
2880 }
2881 if (isPlanarYUV(dstFormat))
2882 {
2883 if (c->vLumFilterSize==1)
2884 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2885 else
2886 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2887 }
2888 else
2889 {
2890 if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
2891 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
2892 " 2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2893 else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
2894 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2895 else
2896 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2897 }
2898
2899 if (dstFormat==PIX_FMT_BGR24)
2900 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
2901 (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
2902 else if (dstFormat==PIX_FMT_RGB32)
2903 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2904 else if (dstFormat==PIX_FMT_BGR565)
2905 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2906 else if (dstFormat==PIX_FMT_BGR555)
2907 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
2908
2909 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
2910 }
2911 if (flags & SWS_PRINT_INFO)
2912 {
2913 av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2914 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
2915 av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
2916 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
2917 }
2918
2919 c->swScale= getSwsFunc(c);
2920 return c;
2921 }
2922
2923 static void reset_ptr(uint8_t* src[], int format){
2924 if(!isALPHA(format))
2925 src[3]=NULL;
2926 if(!isPlanarYUV(format)){
2927 src[3]=src[2]=NULL;
2928 if( format != PIX_FMT_PAL8
2929 && format != PIX_FMT_RGB8
2930 && format != PIX_FMT_BGR8
2931 && format != PIX_FMT_RGB4_BYTE
2932 && format != PIX_FMT_BGR4_BYTE
2933 )
2934 src[1]= NULL;
2935 }
2936 }
2937
2938 /**
2939 * swscale wrapper, so we don't need to export the SwsContext.
2940 * Assumes planar YUV to be in YUV order instead of YVU.
2941 */
2942 int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2943 int srcSliceH, uint8_t* dst[], int dstStride[]){
2944 int i;
2945 uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
2946 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
2947
2948 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2949 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2950 return 0;
2951 }
2952 if (c->sliceDir == 0) {
2953 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2954 }
2955
2956 if (usePal(c->srcFormat)){
2957 for (i=0; i<256; i++){
2958 int p, r, g, b,y,u,v;
2959 if(c->srcFormat == PIX_FMT_PAL8){
2960 p=((uint32_t*)(src[1]))[i];
2961 r= (p>>16)&0xFF;
2962 g= (p>> 8)&0xFF;
2963 b= p &0xFF;
2964 }else if(c->srcFormat == PIX_FMT_RGB8){
2965 r= (i>>5 )*36;
2966 g= ((i>>2)&7)*36;
2967 b= (i&3 )*85;
2968 }else if(c->srcFormat == PIX_FMT_BGR8){
2969 b= (i>>6 )*85;
2970 g= ((i>>3)&7)*36;
2971 r= (i&7 )*36;
2972 }else if(c->srcFormat == PIX_FMT_RGB4_BYTE){
2973 r= (i>>3 )*255;
2974 g= ((i>>1)&3)*85;
2975 b= (i&1 )*255;
2976 }else {
2977 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2978 b= (i>>3 )*255;
2979 g= ((i>>1)&3)*85;
2980 r= (i&1 )*255;
2981 }
2982 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2983 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2984 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2985 c->pal_yuv[i]= y + (u<<8) + (v<<16);
2986
2987
2988 switch(c->dstFormat) {
2989 case PIX_FMT_BGR32:
2990 #ifndef WORDS_BIGENDIAN
2991 case PIX_FMT_RGB24:
2992 #endif
2993 c->pal_rgb[i]= r + (g<<8) + (b<<16);
2994 break;
2995 case PIX_FMT_BGR32_1:
2996 #ifdef WORDS_BIGENDIAN
2997 case PIX_FMT_BGR24:
2998 #endif
2999 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
3000 break;
3001 case PIX_FMT_RGB32_1:
3002 #ifdef WORDS_BIGENDIAN
3003 case PIX_FMT_RGB24:
3004 #endif
3005 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3006 break;
3007 case PIX_FMT_RGB32:
3008 #ifndef WORDS_BIGENDIAN
3009 case PIX_FMT_BGR24:
3010 #endif
3011 default:
3012 c->pal_rgb[i]= b + (g<<8) + (r<<16);
3013 }
3014 }
3015 }
3016
3017 // copy strides, so they can safely be modified
3018 if (c->sliceDir == 1) {
3019 // slices go from top to bottom
3020 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3021 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
3022
3023 reset_ptr(src2, c->srcFormat);
3024 reset_ptr(dst2, c->dstFormat);
3025
3026 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
3027 } else {
3028 // slices go from bottom to top => we flip the image internally
3029 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3030 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
3031
3032 src2[0] += (srcSliceH-1)*srcStride[0];
3033 if (!usePal(c->srcFormat))
3034 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3035 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
3036 src2[3] += (srcSliceH-1)*srcStride[3];
3037 dst2[0] += ( c->dstH -1)*dstStride[0];
3038 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3039 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3040 dst2[3] += ( c->dstH -1)*dstStride[3];
3041
3042 reset_ptr(src2, c->srcFormat);
3043 reset_ptr(dst2, c->dstFormat);
3044
3045 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3046 }
3047 }
3048
3049 #if LIBSWSCALE_VERSION_MAJOR < 1
3050 int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
3051 int srcSliceH, uint8_t* dst[], int dstStride[]){
3052 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
3053 }
3054 #endif
3055
3056 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
3057 float lumaSharpen, float chromaSharpen,
3058 float chromaHShift, float chromaVShift,
3059 int verbose)
3060 {
3061 SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3062
3063 if (lumaGBlur!=0.0){
3064 filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3065 filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
3066 }else{
3067 filter->lumH= sws_getIdentityVec();
3068 filter->lumV= sws_getIdentityVec();
3069 }
3070
3071 if (chromaGBlur!=0.0){
3072 filter->chrH= sws_getGaussianVec(chromaGBlur, 3.0);
3073 filter->chrV= sws_getGaussianVec(chromaGBlur, 3.0);
3074 }else{
3075 filter->chrH= sws_getIdentityVec();
3076 filter->chrV= sws_getIdentityVec();
3077 }
3078
3079 if (chromaSharpen!=0.0){
3080 SwsVector *id= sws_getIdentityVec();
3081 sws_scaleVec(filter->chrH, -chromaSharpen);
3082 sws_scaleVec(filter->chrV, -chromaSharpen);
3083 sws_addVec(filter->chrH, id);
3084 sws_addVec(filter->chrV, id);
3085 sws_freeVec(id);
3086 }
3087
3088 if (lumaSharpen!=0.0){
3089 SwsVector *id= sws_getIdentityVec();
3090 sws_scaleVec(filter->lumH, -lumaSharpen);
3091 sws_scaleVec(filter->lumV, -lumaSharpen);
3092 sws_addVec(filter->lumH, id);
3093 sws_addVec(filter->lumV, id);
3094 sws_freeVec(id);
3095 }
3096
3097 if (chromaHShift != 0.0)
3098 sws_shiftVec(filter->chrH, (int)(chromaHShift+0.5));
3099
3100 if (chromaVShift != 0.0)
3101 sws_shiftVec(filter->chrV, (int)(chromaVShift+0.5));
3102
3103 sws_normalizeVec(filter->chrH, 1.0);
3104 sws_normalizeVec(filter->chrV, 1.0);
3105 sws_normalizeVec(filter->lumH, 1.0);
3106 sws_normalizeVec(filter->lumV, 1.0);
3107
3108 if (verbose) sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
3109 if (verbose) sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
3110
3111 return filter;
3112 }
3113
3114 SwsVector *sws_getGaussianVec(double variance, double quality){
3115 const int length= (int)(variance*quality + 0.5) | 1;
3116 int i;
3117 double *coeff= av_malloc(length*sizeof(double));
3118 double middle= (length-1)*0.5;
3119 SwsVector *vec= av_malloc(sizeof(SwsVector));
3120
3121 vec->coeff= coeff;
3122 vec->length= length;
3123
3124 for (i=0; i<length; i++)
3125 {
3126 double dist= i-middle;
3127 coeff[i]= exp(-dist*dist/(2*variance*variance)) / sqrt(2*variance*PI);
3128 }
3129
3130 sws_normalizeVec(vec, 1.0);
3131
3132 return vec;
3133 }
3134
3135 SwsVector *sws_getConstVec(double c, int length){
3136 int i;
3137 double *coeff= av_malloc(length*sizeof(double));
3138 SwsVector *vec= av_malloc(sizeof(SwsVector));
3139
3140 vec->coeff= coeff;
3141 vec->length= length;
3142
3143 for (i=0; i<length; i++)
3144 coeff[i]= c;
3145
3146 return vec;
3147 }
3148
3149
3150 SwsVector *sws_getIdentityVec(void){
3151 return sws_getConstVec(1.0, 1);
3152 }
3153
3154 double sws_dcVec(SwsVector *a){
3155 int i;
3156 double sum=0;
3157
3158 for (i=0; i<a->length; i++)
3159 sum+= a->coeff[i];
3160
3161 return sum;
3162 }
3163
3164 void sws_scaleVec(SwsVector *a, double scalar){
3165 int i;
3166
3167 for (i=0; i<a->length; i++)
3168 a->coeff[i]*= scalar;
3169 }
3170
3171 void sws_normalizeVec(SwsVector *a, double height){
3172 sws_scaleVec(a, height/sws_dcVec(a));
3173 }
3174
3175 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b){
3176 int length= a->length + b->length - 1;
3177 double *coeff= av_malloc(length*sizeof(double));
3178 int i, j;
3179 SwsVector *vec= av_malloc(sizeof(SwsVector));
3180
3181 vec->coeff= coeff;
3182 vec->length= length;
3183
3184 for (i=0; i<length; i++) coeff[i]= 0.0;
3185
3186 for (i=0; i<a->length; i++)
3187 {
3188 for (j=0; j<b->length; j++)
3189 {
3190 coeff[i+j]+= a->coeff[i]*b->coeff[j];
3191 }
3192 }
3193
3194 return vec;
3195 }
3196
3197 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b){
3198 int length= FFMAX(a->length, b->length);
3199 double *coeff= av_malloc(length*sizeof(double));
3200 int i;
3201 SwsVector *vec= av_malloc(sizeof(SwsVector));
3202
3203 vec->coeff= coeff;
3204 vec->length= length;
3205
3206 for (i=0; i<length; i++) coeff[i]= 0.0;
3207
3208 for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3209 for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i];
3210
3211 return vec;
3212 }
3213
3214 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b){
3215 int length= FFMAX(a->length, b->length);
3216 double *coeff= av_malloc(length*sizeof(double));
3217 int i;
3218 SwsVector *vec= av_malloc(sizeof(SwsVector));
3219
3220 vec->coeff= coeff;
3221 vec->length= length;
3222
3223 for (i=0; i<length; i++) coeff[i]= 0.0;
3224
3225 for (i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i];
3226 for (i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i];
3227
3228 return vec;
3229 }
3230
3231 /* shift left / or right if "shift" is negative */
3232 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift){
3233 int length= a->length + FFABS(shift)*2;
3234 double *coeff= av_malloc(length*sizeof(double));
3235 int i;
3236 SwsVector *vec= av_malloc(sizeof(SwsVector));
3237
3238 vec->coeff= coeff;
3239 vec->length= length;
3240
3241 for (i=0; i<length; i++) coeff[i]= 0.0;
3242
3243 for (i=0; i<a->length; i++)
3244 {
3245 coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i];
3246 }
3247
3248 return vec;
3249 }
3250
3251 void sws_shiftVec(SwsVector *a, int shift){
3252 SwsVector *shifted= sws_getShiftedVec(a, shift);
3253 av_free(a->coeff);
3254 a->coeff= shifted->coeff;
3255 a->length= shifted->length;
3256 av_free(shifted);
3257 }
3258
3259 void sws_addVec(SwsVector *a, SwsVector *b){
3260 SwsVector *sum= sws_sumVec(a, b);
3261 av_free(a->coeff);
3262 a->coeff= sum->coeff;
3263 a->length= sum->length;
3264 av_free(sum);
3265 }
3266
3267 void sws_subVec(SwsVector *a, SwsVector *b){
3268 SwsVector *diff= sws_diffVec(a, b);
3269 av_free(a->coeff);
3270 a->coeff= diff->coeff;
3271 a->length= diff->length;
3272 av_free(diff);
3273 }
3274
3275 void sws_convVec(SwsVector *a, SwsVector *b){
3276 SwsVector *conv= sws_getConvVec(a, b);
3277 av_free(a->coeff);
3278 a->coeff= conv->coeff;
3279 a->length= conv->length;
3280 av_free(conv);
3281 }
3282
3283 SwsVector *sws_cloneVec(SwsVector *a){
3284 double *coeff= av_malloc(a->length*sizeof(double));
3285 int i;
3286 SwsVector *vec= av_malloc(sizeof(SwsVector));
3287
3288 vec->coeff= coeff;
3289 vec->length= a->length;
3290
3291 for (i=0; i<a->length; i++) coeff[i]= a->coeff[i];
3292
3293 return vec;
3294 }
3295
3296 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level){
3297 int i;
3298 double max=0;
3299 double min=0;
3300 double range;
3301
3302 for (i=0; i<a->length; i++)
3303 if (a->coeff[i]>max) max= a->coeff[i];
3304
3305 for (i=0; i<a->length; i++)
3306 if (a->coeff[i]<min) min= a->coeff[i];
3307
3308 range= max - min;
3309
3310 for (i=0; i<a->length; i++)
3311 {
3312 int x= (int)((a->coeff[i]-min)*60.0/range +0.5);
3313 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
3314 for (;x>0; x--) av_log(log_ctx, log_level, " ");
3315 av_log(log_ctx, log_level, "|\n");
3316 }
3317 }
3318
3319 #if LIBSWSCALE_VERSION_MAJOR < 1
3320 void sws_printVec(SwsVector *a){
3321 sws_printVec2(a, NULL, AV_LOG_DEBUG);
3322 }
3323 #endif
3324
3325 void sws_freeVec(SwsVector *a){
3326 if (!a) return;
3327 av_freep(&a->coeff);
3328 a->length=0;
3329 av_free(a);
3330 }
3331
3332 void sws_freeFilter(SwsFilter *filter){
3333 if (!filter) return;
3334
3335 if (filter->lumH) sws_freeVec(filter->lumH);
3336 if (filter->lumV) sws_freeVec(filter->lumV);
3337 if (filter->chrH) sws_freeVec(filter->chrH);
3338 if (filter->chrV) sws_freeVec(filter->chrV);
3339 av_free(filter);
3340 }
3341
3342
3343 void sws_freeContext(SwsContext *c){
3344 int i;
3345 if (!c) return;
3346
3347 if (c->lumPixBuf)
3348 {
3349 for (i=0; i<c->vLumBufSize; i++)
3350 av_freep(&c->lumPixBuf[i]);
3351 av_freep(&c->lumPixBuf);
3352 }
3353
3354 if (c->chrPixBuf)
3355 {
3356 for (i=0; i<c->vChrBufSize; i++)
3357 av_freep(&c->chrPixBuf[i]);
3358 av_freep(&c->chrPixBuf);
3359 }
3360
3361 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf){
3362 for (i=0; i<c->vLumBufSize; i++)
3363 av_freep(&c->alpPixBuf[i]);
3364 av_freep(&c->alpPixBuf);
3365 }
3366
3367 av_freep(&c->vLumFilter);
3368 av_freep(&c->vChrFilter);
3369 av_freep(&c->hLumFilter);
3370 av_freep(&c->hChrFilter);
3371 #if HAVE_ALTIVEC
3372 av_freep(&c->vYCoeffsBank);
3373 av_freep(&c->vCCoeffsBank);
3374 #endif
3375
3376 av_freep(&c->vLumFilterPos);
3377 av_freep(&c->vChrFilterPos);
3378 av_freep(&c->hLumFilterPos);
3379 av_freep(&c->hChrFilterPos);
3380
3381 #if ARCH_X86 && CONFIG_GPL
3382 #ifdef MAP_ANONYMOUS
3383 if (c->funnyYCode ) munmap(c->funnyYCode , MAX_FUNNY_CODE_SIZE);
3384 if (c->funnyUVCode) munmap(c->funnyUVCode, MAX_FUNNY_CODE_SIZE);
3385 #elif HAVE_VIRTUALALLOC
3386 if (c->funnyYCode ) VirtualFree(c->funnyYCode , MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3387 if (c->funnyUVCode) VirtualFree(c->funnyUVCode, MAX_FUNNY_CODE_SIZE, MEM_RELEASE);
3388 #else
3389 av_free(c->funnyYCode );
3390 av_free(c->funnyUVCode);
3391 #endif
3392 c->funnyYCode=NULL;
3393 c->funnyUVCode=NULL;
3394 #endif /* ARCH_X86 && CONFIG_GPL */
3395
3396 av_freep(&c->lumMmx2Filter);
3397 av_freep(&c->chrMmx2Filter);
3398 av_freep(&c->lumMmx2FilterPos);
3399 av_freep(&c->chrMmx2FilterPos);
3400 av_freep(&c->yuvTable);
3401
3402 av_free(c);
3403 }
3404
3405 struct SwsContext *sws_getCachedContext(struct SwsContext *context,
3406 int srcW, int srcH, enum PixelFormat srcFormat,
3407 int dstW, int dstH, enum PixelFormat dstFormat, int flags,
3408 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
3409 {
3410 static const double default_param[2] = {SWS_PARAM_DEFAULT, SWS_PARAM_DEFAULT};
3411
3412 if (!param)
3413 param = default_param;
3414
3415 if (context) {
3416 if (context->srcW != srcW || context->srcH != srcH ||
3417 context->srcFormat != srcFormat ||
3418 context->dstW != dstW || context->dstH != dstH ||
3419 context->dstFormat != dstFormat || context->flags != flags ||
3420 context->param[0] != param[0] || context->param[1] != param[1])
3421 {
3422 sws_freeContext(context);
3423 context = NULL;
3424 }
3425 }
3426 if (!context) {
3427 return sws_getContext(srcW, srcH, srcFormat,
3428 dstW, dstH, dstFormat, flags,
3429 srcFilter, dstFilter, param);
3430 }
3431 return context;
3432 }
3433
mplayer with SMedia Glamo drivers