2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
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 GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #define _DEFAULT_SOURCE
24 #define _SVID_SOURCE // needed for MAP_ANONYMOUS
25 #define _DARWIN_C_SOURCE // needed for MAP_ANON
32 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
33 #define MAP_ANONYMOUS MAP_ANON
37 #define WIN32_LEAN_AND_MEAN
41 #include "libavutil/attributes.h"
42 #include "libavutil/avassert.h"
43 #include "libavutil/avutil.h"
44 #include "libavutil/bswap.h"
45 #include "libavutil/cpu.h"
46 #include "libavutil/imgutils.h"
47 #include "libavutil/intreadwrite.h"
48 #include "libavutil/libm.h"
49 #include "libavutil/mathematics.h"
50 #include "libavutil/opt.h"
51 #include "libavutil/pixdesc.h"
52 #include "libavutil/ppc/cpu.h"
53 #include "libavutil/x86/asm.h"
54 #include "libavutil/x86/cpu.h"
57 #include "swscale_internal.h"
59 static void handle_formats(SwsContext *c);
61 unsigned swscale_version(void)
63 av_assert0(LIBSWSCALE_VERSION_MICRO >= 100);
64 return LIBSWSCALE_VERSION_INT;
67 const char *swscale_configuration(void)
69 return FFMPEG_CONFIGURATION;
72 const char *swscale_license(void)
74 #define LICENSE_PREFIX "libswscale license: "
75 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
78 typedef struct FormatEntry {
79 uint8_t is_supported_in :1;
80 uint8_t is_supported_out :1;
81 uint8_t is_supported_endianness :1;
84 static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
85 [AV_PIX_FMT_YUV420P] = { 1, 1 },
86 [AV_PIX_FMT_YUYV422] = { 1, 1 },
87 [AV_PIX_FMT_RGB24] = { 1, 1 },
88 [AV_PIX_FMT_BGR24] = { 1, 1 },
89 [AV_PIX_FMT_YUV422P] = { 1, 1 },
90 [AV_PIX_FMT_YUV444P] = { 1, 1 },
91 [AV_PIX_FMT_YUV410P] = { 1, 1 },
92 [AV_PIX_FMT_YUV411P] = { 1, 1 },
93 [AV_PIX_FMT_GRAY8] = { 1, 1 },
94 [AV_PIX_FMT_MONOWHITE] = { 1, 1 },
95 [AV_PIX_FMT_MONOBLACK] = { 1, 1 },
96 [AV_PIX_FMT_PAL8] = { 1, 0 },
97 [AV_PIX_FMT_YUVJ420P] = { 1, 1 },
98 [AV_PIX_FMT_YUVJ411P] = { 1, 1 },
99 [AV_PIX_FMT_YUVJ422P] = { 1, 1 },
100 [AV_PIX_FMT_YUVJ444P] = { 1, 1 },
101 [AV_PIX_FMT_YVYU422] = { 1, 1 },
102 [AV_PIX_FMT_UYVY422] = { 1, 1 },
103 [AV_PIX_FMT_UYYVYY411] = { 0, 0 },
104 [AV_PIX_FMT_BGR8] = { 1, 1 },
105 [AV_PIX_FMT_BGR4] = { 0, 1 },
106 [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
107 [AV_PIX_FMT_RGB8] = { 1, 1 },
108 [AV_PIX_FMT_RGB4] = { 0, 1 },
109 [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
110 [AV_PIX_FMT_NV12] = { 1, 1 },
111 [AV_PIX_FMT_NV21] = { 1, 1 },
112 [AV_PIX_FMT_ARGB] = { 1, 1 },
113 [AV_PIX_FMT_RGBA] = { 1, 1 },
114 [AV_PIX_FMT_ABGR] = { 1, 1 },
115 [AV_PIX_FMT_BGRA] = { 1, 1 },
116 [AV_PIX_FMT_0RGB] = { 1, 1 },
117 [AV_PIX_FMT_RGB0] = { 1, 1 },
118 [AV_PIX_FMT_0BGR] = { 1, 1 },
119 [AV_PIX_FMT_BGR0] = { 1, 1 },
120 [AV_PIX_FMT_GRAY16BE] = { 1, 1 },
121 [AV_PIX_FMT_GRAY16LE] = { 1, 1 },
122 [AV_PIX_FMT_YUV440P] = { 1, 1 },
123 [AV_PIX_FMT_YUVJ440P] = { 1, 1 },
124 [AV_PIX_FMT_YUV440P10LE] = { 1, 1 },
125 [AV_PIX_FMT_YUV440P10BE] = { 1, 1 },
126 [AV_PIX_FMT_YUV440P12LE] = { 1, 1 },
127 [AV_PIX_FMT_YUV440P12BE] = { 1, 1 },
128 [AV_PIX_FMT_YUVA420P] = { 1, 1 },
129 [AV_PIX_FMT_YUVA422P] = { 1, 1 },
130 [AV_PIX_FMT_YUVA444P] = { 1, 1 },
131 [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
132 [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
133 [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
134 [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
135 [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
136 [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
137 [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
138 [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
139 [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
140 [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
141 [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
142 [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
143 [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
144 [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
145 [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
146 [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
147 [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
148 [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
149 [AV_PIX_FMT_RGB48BE] = { 1, 1 },
150 [AV_PIX_FMT_RGB48LE] = { 1, 1 },
151 [AV_PIX_FMT_RGBA64BE] = { 1, 1, 1 },
152 [AV_PIX_FMT_RGBA64LE] = { 1, 1, 1 },
153 [AV_PIX_FMT_RGB565BE] = { 1, 1 },
154 [AV_PIX_FMT_RGB565LE] = { 1, 1 },
155 [AV_PIX_FMT_RGB555BE] = { 1, 1 },
156 [AV_PIX_FMT_RGB555LE] = { 1, 1 },
157 [AV_PIX_FMT_BGR565BE] = { 1, 1 },
158 [AV_PIX_FMT_BGR565LE] = { 1, 1 },
159 [AV_PIX_FMT_BGR555BE] = { 1, 1 },
160 [AV_PIX_FMT_BGR555LE] = { 1, 1 },
161 [AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
162 [AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
163 [AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
164 [AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
165 [AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
166 [AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
167 [AV_PIX_FMT_RGB444LE] = { 1, 1 },
168 [AV_PIX_FMT_RGB444BE] = { 1, 1 },
169 [AV_PIX_FMT_BGR444LE] = { 1, 1 },
170 [AV_PIX_FMT_BGR444BE] = { 1, 1 },
171 [AV_PIX_FMT_YA8] = { 1, 1 },
172 [AV_PIX_FMT_YA16BE] = { 1, 0 },
173 [AV_PIX_FMT_YA16LE] = { 1, 0 },
174 [AV_PIX_FMT_BGR48BE] = { 1, 1 },
175 [AV_PIX_FMT_BGR48LE] = { 1, 1 },
176 [AV_PIX_FMT_BGRA64BE] = { 1, 1, 1 },
177 [AV_PIX_FMT_BGRA64LE] = { 1, 1, 1 },
178 [AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
179 [AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
180 [AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
181 [AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
182 [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
183 [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
184 [AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
185 [AV_PIX_FMT_YUV420P14LE] = { 1, 1 },
186 [AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
187 [AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
188 [AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
189 [AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
190 [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
191 [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
192 [AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
193 [AV_PIX_FMT_YUV422P14LE] = { 1, 1 },
194 [AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
195 [AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
196 [AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
197 [AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
198 [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
199 [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
200 [AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
201 [AV_PIX_FMT_YUV444P14LE] = { 1, 1 },
202 [AV_PIX_FMT_GBRP] = { 1, 1 },
203 [AV_PIX_FMT_GBRP9LE] = { 1, 1 },
204 [AV_PIX_FMT_GBRP9BE] = { 1, 1 },
205 [AV_PIX_FMT_GBRP10LE] = { 1, 1 },
206 [AV_PIX_FMT_GBRP10BE] = { 1, 1 },
207 [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
208 [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
209 [AV_PIX_FMT_GBRAP12LE] = { 1, 0 },
210 [AV_PIX_FMT_GBRAP12BE] = { 1, 0 },
211 [AV_PIX_FMT_GBRP14LE] = { 1, 1 },
212 [AV_PIX_FMT_GBRP14BE] = { 1, 1 },
213 [AV_PIX_FMT_GBRP16LE] = { 1, 0 },
214 [AV_PIX_FMT_GBRP16BE] = { 1, 0 },
215 [AV_PIX_FMT_GBRAP] = { 1, 1 },
216 [AV_PIX_FMT_GBRAP16LE] = { 1, 0 },
217 [AV_PIX_FMT_GBRAP16BE] = { 1, 0 },
218 [AV_PIX_FMT_BAYER_BGGR8] = { 1, 0 },
219 [AV_PIX_FMT_BAYER_RGGB8] = { 1, 0 },
220 [AV_PIX_FMT_BAYER_GBRG8] = { 1, 0 },
221 [AV_PIX_FMT_BAYER_GRBG8] = { 1, 0 },
222 [AV_PIX_FMT_BAYER_BGGR16LE] = { 1, 0 },
223 [AV_PIX_FMT_BAYER_BGGR16BE] = { 1, 0 },
224 [AV_PIX_FMT_BAYER_RGGB16LE] = { 1, 0 },
225 [AV_PIX_FMT_BAYER_RGGB16BE] = { 1, 0 },
226 [AV_PIX_FMT_BAYER_GBRG16LE] = { 1, 0 },
227 [AV_PIX_FMT_BAYER_GBRG16BE] = { 1, 0 },
228 [AV_PIX_FMT_BAYER_GRBG16LE] = { 1, 0 },
229 [AV_PIX_FMT_BAYER_GRBG16BE] = { 1, 0 },
230 [AV_PIX_FMT_XYZ12BE] = { 1, 1, 1 },
231 [AV_PIX_FMT_XYZ12LE] = { 1, 1, 1 },
232 [AV_PIX_FMT_AYUV64LE] = { 1, 1},
233 [AV_PIX_FMT_P010LE] = { 1, 0 },
234 [AV_PIX_FMT_P010BE] = { 1, 0 },
237 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
239 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
240 format_entries[pix_fmt].is_supported_in : 0;
243 int sws_isSupportedOutput(enum AVPixelFormat pix_fmt)
245 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
246 format_entries[pix_fmt].is_supported_out : 0;
249 int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
251 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
252 format_entries[pix_fmt].is_supported_endianness : 0;
255 static double getSplineCoeff(double a, double b, double c, double d,
259 return ((d * dist + c) * dist + b) * dist + a;
261 return getSplineCoeff(0.0,
262 b + 2.0 * c + 3.0 * d,
264 -b - 3.0 * c - 6.0 * d,
268 static av_cold int get_local_pos(SwsContext *s, int chr_subsample, int pos, int dir)
270 if (pos == -1 || pos <= -513) {
271 pos = (128 << chr_subsample) - 128;
273 pos += 128; // relative to ideal left edge
274 return pos >> chr_subsample;
278 int flag; ///< flag associated to the algorithm
279 const char *description; ///< human-readable description
280 int size_factor; ///< size factor used when initing the filters
283 static const ScaleAlgorithm scale_algorithms[] = {
284 { SWS_AREA, "area averaging", 1 /* downscale only, for upscale it is bilinear */ },
285 { SWS_BICUBIC, "bicubic", 4 },
286 { SWS_BICUBLIN, "luma bicubic / chroma bilinear", -1 },
287 { SWS_BILINEAR, "bilinear", 2 },
288 { SWS_FAST_BILINEAR, "fast bilinear", -1 },
289 { SWS_GAUSS, "Gaussian", 8 /* infinite ;) */ },
290 { SWS_LANCZOS, "Lanczos", -1 /* custom */ },
291 { SWS_POINT, "nearest neighbor / point", -1 },
292 { SWS_SINC, "sinc", 20 /* infinite ;) */ },
293 { SWS_SPLINE, "bicubic spline", 20 /* infinite :)*/ },
294 { SWS_X, "experimental", 8 },
297 static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
298 int *outFilterSize, int xInc, int srcW,
299 int dstW, int filterAlign, int one,
300 int flags, int cpu_flags,
301 SwsVector *srcFilter, SwsVector *dstFilter,
302 double param[2], int srcPos, int dstPos)
308 int64_t *filter = NULL;
309 int64_t *filter2 = NULL;
310 const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
313 emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
315 // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
316 FF_ALLOC_ARRAY_OR_GOTO(NULL, *filterPos, (dstW + 3), sizeof(**filterPos), fail);
318 if (FFABS(xInc - 0x10000) < 10 && srcPos == dstPos) { // unscaled
321 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter,
322 dstW, sizeof(*filter) * filterSize, fail);
324 for (i = 0; i < dstW; i++) {
325 filter[i * filterSize] = fone;
328 } else if (flags & SWS_POINT) { // lame looking point sampling mode
332 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
333 dstW, sizeof(*filter) * filterSize, fail);
335 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
336 for (i = 0; i < dstW; i++) {
337 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
339 (*filterPos)[i] = xx;
343 } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
344 (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
348 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
349 dstW, sizeof(*filter) * filterSize, fail);
351 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
352 for (i = 0; i < dstW; i++) {
353 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
356 (*filterPos)[i] = xx;
357 // bilinear upscale / linear interpolate / area averaging
358 for (j = 0; j < filterSize; j++) {
359 int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
362 filter[i * filterSize + j] = coeff;
371 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
372 if (flags & scale_algorithms[i].flag && scale_algorithms[i].size_factor > 0) {
373 sizeFactor = scale_algorithms[i].size_factor;
377 if (flags & SWS_LANCZOS)
378 sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
379 av_assert0(sizeFactor > 0);
382 filterSize = 1 + sizeFactor; // upscale
384 filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
386 filterSize = FFMIN(filterSize, srcW - 2);
387 filterSize = FFMAX(filterSize, 1);
389 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
390 dstW, sizeof(*filter) * filterSize, fail);
392 xDstInSrc = ((dstPos*(int64_t)xInc)>>7) - ((srcPos*0x10000LL)>>7);
393 for (i = 0; i < dstW; i++) {
394 int xx = (xDstInSrc - (filterSize - 2) * (1LL<<16)) / (1 << 17);
396 (*filterPos)[i] = xx;
397 for (j = 0; j < filterSize; j++) {
398 int64_t d = (FFABS(((int64_t)xx * (1 << 17)) - xDstInSrc)) << 13;
404 floatd = d * (1.0 / (1 << 30));
406 if (flags & SWS_BICUBIC) {
407 int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
408 int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
410 if (d >= 1LL << 31) {
413 int64_t dd = (d * d) >> 30;
414 int64_t ddd = (dd * d) >> 30;
417 coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
418 (-18 * (1 << 24) + 12 * B + 6 * C) * dd +
419 (6 * (1 << 24) - 2 * B) * (1 << 30);
421 coeff = (-B - 6 * C) * ddd +
422 (6 * B + 30 * C) * dd +
423 (-12 * B - 48 * C) * d +
424 (8 * B + 24 * C) * (1 << 30);
426 coeff /= (1LL<<54)/fone;
429 else if (flags & SWS_X) {
430 double p = param ? param * 0.01 : 0.3;
431 coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
432 coeff *= pow(2.0, -p * d * d);
435 else if (flags & SWS_X) {
436 double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
440 c = cos(floatd * M_PI);
447 coeff = (c * 0.5 + 0.5) * fone;
448 } else if (flags & SWS_AREA) {
449 int64_t d2 = d - (1 << 29);
450 if (d2 * xInc < -(1LL << (29 + 16)))
451 coeff = 1.0 * (1LL << (30 + 16));
452 else if (d2 * xInc < (1LL << (29 + 16)))
453 coeff = -d2 * xInc + (1LL << (29 + 16));
456 coeff *= fone >> (30 + 16);
457 } else if (flags & SWS_GAUSS) {
458 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
459 coeff = exp2(-p * floatd * floatd) * fone;
460 } else if (flags & SWS_SINC) {
461 coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
462 } else if (flags & SWS_LANCZOS) {
463 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
464 coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
465 (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
468 } else if (flags & SWS_BILINEAR) {
469 coeff = (1 << 30) - d;
473 } else if (flags & SWS_SPLINE) {
474 double p = -2.196152422706632;
475 coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
480 filter[i * filterSize + j] = coeff;
483 xDstInSrc += 2 * xInc;
487 /* apply src & dst Filter to filter -> filter2
490 av_assert0(filterSize > 0);
491 filter2Size = filterSize;
493 filter2Size += srcFilter->length - 1;
495 filter2Size += dstFilter->length - 1;
496 av_assert0(filter2Size > 0);
497 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter2, dstW, filter2Size * sizeof(*filter2), fail);
499 for (i = 0; i < dstW; i++) {
503 for (k = 0; k < srcFilter->length; k++) {
504 for (j = 0; j < filterSize; j++)
505 filter2[i * filter2Size + k + j] +=
506 srcFilter->coeff[k] * filter[i * filterSize + j];
509 for (j = 0; j < filterSize; j++)
510 filter2[i * filter2Size + j] = filter[i * filterSize + j];
514 (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
518 /* try to reduce the filter-size (step1 find size and shift left) */
519 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
521 for (i = dstW - 1; i >= 0; i--) {
522 int min = filter2Size;
524 int64_t cutOff = 0.0;
526 /* get rid of near zero elements on the left by shifting left */
527 for (j = 0; j < filter2Size; j++) {
529 cutOff += FFABS(filter2[i * filter2Size]);
531 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
534 /* preserve monotonicity because the core can't handle the
535 * filter otherwise */
536 if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
539 // move filter coefficients left
540 for (k = 1; k < filter2Size; k++)
541 filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
542 filter2[i * filter2Size + k - 1] = 0;
547 /* count near zeros on the right */
548 for (j = filter2Size - 1; j > 0; j--) {
549 cutOff += FFABS(filter2[i * filter2Size + j]);
551 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
556 if (min > minFilterSize)
560 if (PPC_ALTIVEC(cpu_flags)) {
561 // we can handle the special case 4, so we don't want to go the full 8
562 if (minFilterSize < 5)
565 /* We really don't want to waste our time doing useless computation, so
566 * fall back on the scalar C code for very small filters.
567 * Vectorizing is worth it only if you have a decent-sized vector. */
568 if (minFilterSize < 3)
572 if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
573 // special case for unscaled vertical filtering
574 if (minFilterSize == 1 && filterAlign == 2)
578 av_assert0(minFilterSize > 0);
579 filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
580 av_assert0(filterSize > 0);
581 filter = av_malloc_array(dstW, filterSize * sizeof(*filter));
584 if (filterSize >= MAX_FILTER_SIZE * 16 /
585 ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16)) {
586 ret = RETCODE_USE_CASCADE;
589 *outFilterSize = filterSize;
591 if (flags & SWS_PRINT_INFO)
592 av_log(NULL, AV_LOG_VERBOSE,
593 "SwScaler: reducing / aligning filtersize %d -> %d\n",
594 filter2Size, filterSize);
595 /* try to reduce the filter-size (step2 reduce it) */
596 for (i = 0; i < dstW; i++) {
599 for (j = 0; j < filterSize; j++) {
600 if (j >= filter2Size)
601 filter[i * filterSize + j] = 0;
603 filter[i * filterSize + j] = filter2[i * filter2Size + j];
604 if ((flags & SWS_BITEXACT) && j >= minFilterSize)
605 filter[i * filterSize + j] = 0;
609 // FIXME try to align filterPos if possible
612 for (i = 0; i < dstW; i++) {
614 if ((*filterPos)[i] < 0) {
615 // move filter coefficients left to compensate for filterPos
616 for (j = 1; j < filterSize; j++) {
617 int left = FFMAX(j + (*filterPos)[i], 0);
618 filter[i * filterSize + left] += filter[i * filterSize + j];
619 filter[i * filterSize + j] = 0;
624 if ((*filterPos)[i] + filterSize > srcW) {
625 int shift = (*filterPos)[i] + FFMIN(filterSize - srcW, 0);
628 for (j = filterSize - 1; j >= 0; j--) {
629 if ((*filterPos)[i] + j >= srcW) {
630 acc += filter[i * filterSize + j];
631 filter[i * filterSize + j] = 0;
634 for (j = filterSize - 1; j >= 0; j--) {
636 filter[i * filterSize + j] = 0;
638 filter[i * filterSize + j] = filter[i * filterSize + j - shift];
642 (*filterPos)[i]-= shift;
643 filter[i * filterSize + srcW - 1 - (*filterPos)[i]] += acc;
645 av_assert0((*filterPos)[i] >= 0);
646 av_assert0((*filterPos)[i] < srcW);
647 if ((*filterPos)[i] + filterSize > srcW) {
648 for (j = 0; j < filterSize; j++) {
649 av_assert0((*filterPos)[i] + j < srcW || !filter[i * filterSize + j]);
654 // Note the +1 is for the MMX scaler which reads over the end
655 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
656 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, *outFilter,
657 (dstW + 3), *outFilterSize * sizeof(int16_t), fail);
659 /* normalize & store in outFilter */
660 for (i = 0; i < dstW; i++) {
665 for (j = 0; j < filterSize; j++) {
666 sum += filter[i * filterSize + j];
668 sum = (sum + one / 2) / one;
670 av_log(NULL, AV_LOG_WARNING, "SwScaler: zero vector in scaling\n");
673 for (j = 0; j < *outFilterSize; j++) {
674 int64_t v = filter[i * filterSize + j] + error;
675 int intV = ROUNDED_DIV(v, sum);
676 (*outFilter)[i * (*outFilterSize) + j] = intV;
677 error = v - intV * sum;
681 (*filterPos)[dstW + 0] =
682 (*filterPos)[dstW + 1] =
683 (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
684 * read over the end */
685 for (i = 0; i < *outFilterSize; i++) {
686 int k = (dstW - 1) * (*outFilterSize) + i;
687 (*outFilter)[k + 1 * (*outFilterSize)] =
688 (*outFilter)[k + 2 * (*outFilterSize)] =
689 (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
696 av_log(NULL, ret == RETCODE_USE_CASCADE ? AV_LOG_DEBUG : AV_LOG_ERROR, "sws: initFilter failed\n");
702 static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange)
704 int64_t W, V, Z, Cy, Cu, Cv;
705 int64_t vr = table[0];
706 int64_t ub = table[1];
707 int64_t ug = -table[2];
708 int64_t vg = -table[3];
711 uint8_t *p = (uint8_t*)c->input_rgb2yuv_table;
713 static const int8_t map[] = {
714 BY_IDX, GY_IDX, -1 , BY_IDX, BY_IDX, GY_IDX, -1 , BY_IDX,
715 RY_IDX, -1 , GY_IDX, RY_IDX, RY_IDX, -1 , GY_IDX, RY_IDX,
716 RY_IDX, GY_IDX, -1 , RY_IDX, RY_IDX, GY_IDX, -1 , RY_IDX,
717 BY_IDX, -1 , GY_IDX, BY_IDX, BY_IDX, -1 , GY_IDX, BY_IDX,
718 BU_IDX, GU_IDX, -1 , BU_IDX, BU_IDX, GU_IDX, -1 , BU_IDX,
719 RU_IDX, -1 , GU_IDX, RU_IDX, RU_IDX, -1 , GU_IDX, RU_IDX,
720 RU_IDX, GU_IDX, -1 , RU_IDX, RU_IDX, GU_IDX, -1 , RU_IDX,
721 BU_IDX, -1 , GU_IDX, BU_IDX, BU_IDX, -1 , GU_IDX, BU_IDX,
722 BV_IDX, GV_IDX, -1 , BV_IDX, BV_IDX, GV_IDX, -1 , BV_IDX,
723 RV_IDX, -1 , GV_IDX, RV_IDX, RV_IDX, -1 , GV_IDX, RV_IDX,
724 RV_IDX, GV_IDX, -1 , RV_IDX, RV_IDX, GV_IDX, -1 , RV_IDX,
725 BV_IDX, -1 , GV_IDX, BV_IDX, BV_IDX, -1 , GV_IDX, BV_IDX,
726 RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX,
727 BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX,
728 GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 ,
729 -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX,
730 RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX,
731 BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX,
732 GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 ,
733 -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX,
734 RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX,
735 BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX,
736 GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 ,
737 -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, //23
738 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //24
739 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //25
740 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //26
741 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //27
742 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //28
743 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //29
744 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //30
745 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //31
746 BY_IDX, GY_IDX, RY_IDX, -1 , -1 , -1 , -1 , -1 , //32
747 BU_IDX, GU_IDX, RU_IDX, -1 , -1 , -1 , -1 , -1 , //33
748 BV_IDX, GV_IDX, RV_IDX, -1 , -1 , -1 , -1 , -1 , //34
751 dstRange = 0; //FIXME range = 1 is handled elsewhere
761 W = ROUNDED_DIV(ONE*ONE*ug, ub);
762 V = ROUNDED_DIV(ONE*ONE*vg, vr);
765 Cy = ROUNDED_DIV(cy*Z, ONE);
766 Cu = ROUNDED_DIV(ub*Z, ONE);
767 Cv = ROUNDED_DIV(vr*Z, ONE);
769 c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cy);
770 c->input_rgb2yuv_table[GY_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cy);
771 c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cy);
773 c->input_rgb2yuv_table[RU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cu);
774 c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cu);
775 c->input_rgb2yuv_table[BU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W) , Cu);
777 c->input_rgb2yuv_table[RV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z) , Cv);
778 c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cv);
779 c->input_rgb2yuv_table[BV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cv);
781 if(/*!dstRange && */!memcmp(table, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sizeof(ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]))) {
782 c->input_rgb2yuv_table[BY_IDX] = ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
783 c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
784 c->input_rgb2yuv_table[BU_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
785 c->input_rgb2yuv_table[GY_IDX] = ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
786 c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
787 c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
788 c->input_rgb2yuv_table[RY_IDX] = ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
789 c->input_rgb2yuv_table[RV_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
790 c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
792 for(i=0; i<FF_ARRAY_ELEMS(map); i++)
793 AV_WL16(p + 16*4 + 2*i, map[i] >= 0 ? c->input_rgb2yuv_table[map[i]] : 0);
796 static void fill_xyztables(struct SwsContext *c)
799 double xyzgamma = XYZ_GAMMA;
800 double rgbgamma = 1.0 / RGB_GAMMA;
801 double xyzgammainv = 1.0 / XYZ_GAMMA;
802 double rgbgammainv = RGB_GAMMA;
803 static const int16_t xyz2rgb_matrix[3][4] = {
804 {13270, -6295, -2041},
806 { 228, -835, 4329} };
807 static const int16_t rgb2xyz_matrix[3][4] = {
811 static int16_t xyzgamma_tab[4096], rgbgamma_tab[4096], xyzgammainv_tab[4096], rgbgammainv_tab[4096];
813 memcpy(c->xyz2rgb_matrix, xyz2rgb_matrix, sizeof(c->xyz2rgb_matrix));
814 memcpy(c->rgb2xyz_matrix, rgb2xyz_matrix, sizeof(c->rgb2xyz_matrix));
815 c->xyzgamma = xyzgamma_tab;
816 c->rgbgamma = rgbgamma_tab;
817 c->xyzgammainv = xyzgammainv_tab;
818 c->rgbgammainv = rgbgammainv_tab;
820 if (rgbgamma_tab[4095])
823 /* set gamma vectors */
824 for (i = 0; i < 4096; i++) {
825 xyzgamma_tab[i] = lrint(pow(i / 4095.0, xyzgamma) * 4095.0);
826 rgbgamma_tab[i] = lrint(pow(i / 4095.0, rgbgamma) * 4095.0);
827 xyzgammainv_tab[i] = lrint(pow(i / 4095.0, xyzgammainv) * 4095.0);
828 rgbgammainv_tab[i] = lrint(pow(i / 4095.0, rgbgammainv) * 4095.0);
832 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
833 int srcRange, const int table[4], int dstRange,
834 int brightness, int contrast, int saturation)
836 const AVPixFmtDescriptor *desc_dst;
837 const AVPixFmtDescriptor *desc_src;
841 desc_dst = av_pix_fmt_desc_get(c->dstFormat);
842 desc_src = av_pix_fmt_desc_get(c->srcFormat);
844 if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
846 if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
849 if (c->srcRange != srcRange ||
850 c->dstRange != dstRange ||
851 c->brightness != brightness ||
852 c->contrast != contrast ||
853 c->saturation != saturation ||
854 memcmp(c->srcColorspaceTable, inv_table, sizeof(int) * 4) ||
855 memcmp(c->dstColorspaceTable, table, sizeof(int) * 4)
859 memmove(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
860 memmove(c->dstColorspaceTable, table, sizeof(int) * 4);
864 c->brightness = brightness;
865 c->contrast = contrast;
866 c->saturation = saturation;
867 c->srcRange = srcRange;
868 c->dstRange = dstRange;
870 //The srcBpc check is possibly wrong but we seem to lack a definitive reference to test this
871 //and what we have in ticket 2939 looks better with this check
872 if (need_reinit && (c->srcBpc == 8 || !isYUV(c->srcFormat)))
873 ff_sws_init_range_convert(c);
875 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
876 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
878 if (c->cascaded_context[c->cascaded_mainindex])
879 return sws_setColorspaceDetails(c->cascaded_context[c->cascaded_mainindex],inv_table, srcRange,table, dstRange, brightness, contrast, saturation);
884 if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat))) {
885 if (!c->cascaded_context[0] &&
886 memcmp(c->dstColorspaceTable, c->srcColorspaceTable, sizeof(int) * 4) &&
887 c->srcW && c->srcH && c->dstW && c->dstH) {
888 enum AVPixelFormat tmp_format;
889 int tmp_width, tmp_height;
895 av_log(c, AV_LOG_VERBOSE, "YUV color matrix differs for YUV->YUV, using intermediate RGB to convert\n");
897 if (isNBPS(c->dstFormat) || is16BPS(c->dstFormat)) {
898 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
899 tmp_format = AV_PIX_FMT_BGRA64;
901 tmp_format = AV_PIX_FMT_BGR48;
904 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
905 tmp_format = AV_PIX_FMT_BGRA;
907 tmp_format = AV_PIX_FMT_BGR24;
911 if (srcW*srcH > dstW*dstH) {
919 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
920 tmp_width, tmp_height, tmp_format, 64);
924 c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, c->srcFormat,
925 tmp_width, tmp_height, tmp_format,
927 if (!c->cascaded_context[0])
930 c->cascaded_context[0]->alphablend = c->alphablend;
931 ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
934 //we set both src and dst depending on that the RGB side will be ignored
935 sws_setColorspaceDetails(c->cascaded_context[0], inv_table,
936 srcRange, table, dstRange,
937 brightness, contrast, saturation);
939 c->cascaded_context[1] = sws_getContext(tmp_width, tmp_height, tmp_format,
940 dstW, dstH, c->dstFormat,
941 c->flags, NULL, NULL, c->param);
942 if (!c->cascaded_context[1])
944 sws_setColorspaceDetails(c->cascaded_context[1], inv_table,
945 srcRange, table, dstRange,
946 0, 1 << 16, 1 << 16);
952 if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) {
953 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
954 contrast, saturation);
958 ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
959 contrast, saturation);
962 fill_rgb2yuv_table(c, table, dstRange);
967 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
968 int *srcRange, int **table, int *dstRange,
969 int *brightness, int *contrast, int *saturation)
974 *inv_table = c->srcColorspaceTable;
975 *table = c->dstColorspaceTable;
976 *srcRange = c->srcRange;
977 *dstRange = c->dstRange;
978 *brightness = c->brightness;
979 *contrast = c->contrast;
980 *saturation = c->saturation;
985 static int handle_jpeg(enum AVPixelFormat *format)
988 case AV_PIX_FMT_YUVJ420P:
989 *format = AV_PIX_FMT_YUV420P;
991 case AV_PIX_FMT_YUVJ411P:
992 *format = AV_PIX_FMT_YUV411P;
994 case AV_PIX_FMT_YUVJ422P:
995 *format = AV_PIX_FMT_YUV422P;
997 case AV_PIX_FMT_YUVJ444P:
998 *format = AV_PIX_FMT_YUV444P;
1000 case AV_PIX_FMT_YUVJ440P:
1001 *format = AV_PIX_FMT_YUV440P;
1003 case AV_PIX_FMT_GRAY8:
1004 case AV_PIX_FMT_YA8:
1005 case AV_PIX_FMT_GRAY16LE:
1006 case AV_PIX_FMT_GRAY16BE:
1007 case AV_PIX_FMT_YA16BE:
1008 case AV_PIX_FMT_YA16LE:
1015 static int handle_0alpha(enum AVPixelFormat *format)
1018 case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
1019 case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
1020 case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
1021 case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4;
1026 static int handle_xyz(enum AVPixelFormat *format)
1029 case AV_PIX_FMT_XYZ12BE : *format = AV_PIX_FMT_RGB48BE; return 1;
1030 case AV_PIX_FMT_XYZ12LE : *format = AV_PIX_FMT_RGB48LE; return 1;
1035 static void handle_formats(SwsContext *c)
1037 c->src0Alpha |= handle_0alpha(&c->srcFormat);
1038 c->dst0Alpha |= handle_0alpha(&c->dstFormat);
1039 c->srcXYZ |= handle_xyz(&c->srcFormat);
1040 c->dstXYZ |= handle_xyz(&c->dstFormat);
1041 if (c->srcXYZ || c->dstXYZ)
1045 SwsContext *sws_alloc_context(void)
1047 SwsContext *c = av_mallocz(sizeof(SwsContext));
1049 av_assert0(offsetof(SwsContext, redDither) + DITHER32_INT == offsetof(SwsContext, dither32));
1052 c->av_class = &ff_sws_context_class;
1053 av_opt_set_defaults(c);
1059 static uint16_t * alloc_gamma_tbl(double e)
1063 tbl = (uint16_t*)av_malloc(sizeof(uint16_t) * 1 << 16);
1067 for (i = 0; i < 65536; ++i) {
1068 tbl[i] = pow(i / 65535.0, e) * 65535.0;
1073 static enum AVPixelFormat alphaless_fmt(enum AVPixelFormat fmt)
1076 case AV_PIX_FMT_ARGB: return AV_PIX_FMT_RGB24;
1077 case AV_PIX_FMT_RGBA: return AV_PIX_FMT_RGB24;
1078 case AV_PIX_FMT_ABGR: return AV_PIX_FMT_BGR24;
1079 case AV_PIX_FMT_BGRA: return AV_PIX_FMT_BGR24;
1080 case AV_PIX_FMT_YA8: return AV_PIX_FMT_GRAY8;
1082 case AV_PIX_FMT_YUVA420P: return AV_PIX_FMT_YUV420P;
1083 case AV_PIX_FMT_YUVA422P: return AV_PIX_FMT_YUV422P;
1084 case AV_PIX_FMT_YUVA444P: return AV_PIX_FMT_YUV444P;
1086 case AV_PIX_FMT_GBRAP: return AV_PIX_FMT_GBRP;
1088 case AV_PIX_FMT_GBRAP12LE: return AV_PIX_FMT_GBRP12;
1089 case AV_PIX_FMT_GBRAP12BE: return AV_PIX_FMT_GBRP12;
1091 case AV_PIX_FMT_GBRAP16LE: return AV_PIX_FMT_GBRP16;
1092 case AV_PIX_FMT_GBRAP16BE: return AV_PIX_FMT_GBRP16;
1094 case AV_PIX_FMT_RGBA64LE: return AV_PIX_FMT_RGB48;
1095 case AV_PIX_FMT_RGBA64BE: return AV_PIX_FMT_RGB48;
1096 case AV_PIX_FMT_BGRA64LE: return AV_PIX_FMT_BGR48;
1097 case AV_PIX_FMT_BGRA64BE: return AV_PIX_FMT_BGR48;
1099 case AV_PIX_FMT_YA16BE: return AV_PIX_FMT_GRAY16;
1100 case AV_PIX_FMT_YA16LE: return AV_PIX_FMT_GRAY16;
1102 case AV_PIX_FMT_YUVA420P9BE: return AV_PIX_FMT_YUV420P9;
1103 case AV_PIX_FMT_YUVA422P9BE: return AV_PIX_FMT_YUV422P9;
1104 case AV_PIX_FMT_YUVA444P9BE: return AV_PIX_FMT_YUV444P9;
1105 case AV_PIX_FMT_YUVA420P9LE: return AV_PIX_FMT_YUV420P9;
1106 case AV_PIX_FMT_YUVA422P9LE: return AV_PIX_FMT_YUV422P9;
1107 case AV_PIX_FMT_YUVA444P9LE: return AV_PIX_FMT_YUV444P9;
1108 case AV_PIX_FMT_YUVA420P10BE: return AV_PIX_FMT_YUV420P10;
1109 case AV_PIX_FMT_YUVA422P10BE: return AV_PIX_FMT_YUV422P10;
1110 case AV_PIX_FMT_YUVA444P10BE: return AV_PIX_FMT_YUV444P10;
1111 case AV_PIX_FMT_YUVA420P10LE: return AV_PIX_FMT_YUV420P10;
1112 case AV_PIX_FMT_YUVA422P10LE: return AV_PIX_FMT_YUV422P10;
1113 case AV_PIX_FMT_YUVA444P10LE: return AV_PIX_FMT_YUV444P10;
1114 case AV_PIX_FMT_YUVA420P16BE: return AV_PIX_FMT_YUV420P16;
1115 case AV_PIX_FMT_YUVA422P16BE: return AV_PIX_FMT_YUV422P16;
1116 case AV_PIX_FMT_YUVA444P16BE: return AV_PIX_FMT_YUV444P16;
1117 case AV_PIX_FMT_YUVA420P16LE: return AV_PIX_FMT_YUV420P16;
1118 case AV_PIX_FMT_YUVA422P16LE: return AV_PIX_FMT_YUV422P16;
1119 case AV_PIX_FMT_YUVA444P16LE: return AV_PIX_FMT_YUV444P16;
1121 // case AV_PIX_FMT_AYUV64LE:
1122 // case AV_PIX_FMT_AYUV64BE:
1123 // case AV_PIX_FMT_PAL8:
1124 default: return AV_PIX_FMT_NONE;
1128 av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
1129 SwsFilter *dstFilter)
1132 int usesVFilter, usesHFilter;
1134 SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
1139 int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
1140 int flags, cpu_flags;
1141 enum AVPixelFormat srcFormat = c->srcFormat;
1142 enum AVPixelFormat dstFormat = c->dstFormat;
1143 const AVPixFmtDescriptor *desc_src;
1144 const AVPixFmtDescriptor *desc_dst;
1146 enum AVPixelFormat tmpFmt;
1148 cpu_flags = av_get_cpu_flags();
1152 ff_sws_rgb2rgb_init();
1154 unscaled = (srcW == dstW && srcH == dstH);
1156 c->srcRange |= handle_jpeg(&c->srcFormat);
1157 c->dstRange |= handle_jpeg(&c->dstFormat);
1159 if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
1160 av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
1162 if (!c->contrast && !c->saturation && !c->dstFormatBpp)
1163 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
1164 ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
1165 c->dstRange, 0, 1 << 16, 1 << 16);
1168 srcFormat = c->srcFormat;
1169 dstFormat = c->dstFormat;
1170 desc_src = av_pix_fmt_desc_get(srcFormat);
1171 desc_dst = av_pix_fmt_desc_get(dstFormat);
1173 // If the source has no alpha then disable alpha blendaway
1175 c->alphablend = SWS_ALPHA_BLEND_NONE;
1177 if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
1178 av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
1179 if (!sws_isSupportedInput(srcFormat)) {
1180 av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
1181 av_get_pix_fmt_name(srcFormat));
1182 return AVERROR(EINVAL);
1184 if (!sws_isSupportedOutput(dstFormat)) {
1185 av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
1186 av_get_pix_fmt_name(dstFormat));
1187 return AVERROR(EINVAL);
1190 av_assert2(desc_src && desc_dst);
1192 i = flags & (SWS_POINT |
1204 /* provide a default scaler if not set by caller */
1206 if (dstW < srcW && dstH < srcH)
1207 flags |= SWS_BICUBIC;
1208 else if (dstW > srcW && dstH > srcH)
1209 flags |= SWS_BICUBIC;
1211 flags |= SWS_BICUBIC;
1213 } else if (i & (i - 1)) {
1214 av_log(c, AV_LOG_ERROR,
1215 "Exactly one scaler algorithm must be chosen, got %X\n", i);
1216 return AVERROR(EINVAL);
1219 if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
1220 /* FIXME check if these are enough and try to lower them after
1221 * fixing the relevant parts of the code */
1222 av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
1223 srcW, srcH, dstW, dstH);
1224 return AVERROR(EINVAL);
1226 if (flags & SWS_FAST_BILINEAR) {
1227 if (srcW < 8 || dstW < 8) {
1228 flags ^= SWS_FAST_BILINEAR | SWS_BILINEAR;
1234 dstFilter = &dummyFilter;
1236 srcFilter = &dummyFilter;
1238 c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
1239 c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
1240 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
1241 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
1242 c->vRounder = 4 * 0x0001000100010001ULL;
1244 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
1245 (srcFilter->chrV && srcFilter->chrV->length > 1) ||
1246 (dstFilter->lumV && dstFilter->lumV->length > 1) ||
1247 (dstFilter->chrV && dstFilter->chrV->length > 1);
1248 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
1249 (srcFilter->chrH && srcFilter->chrH->length > 1) ||
1250 (dstFilter->lumH && dstFilter->lumH->length > 1) ||
1251 (dstFilter->chrH && dstFilter->chrH->length > 1);
1253 av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
1254 av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);
1256 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
1258 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
1259 flags |= SWS_FULL_CHR_H_INT;
1263 if ( c->chrSrcHSubSample == 0
1264 && c->chrSrcVSubSample == 0
1265 && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
1266 && !(c->flags & SWS_FAST_BILINEAR)
1268 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
1269 flags |= SWS_FULL_CHR_H_INT;
1274 if (c->dither == SWS_DITHER_AUTO) {
1275 if (flags & SWS_ERROR_DIFFUSION)
1276 c->dither = SWS_DITHER_ED;
1279 if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1280 dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1281 dstFormat == AV_PIX_FMT_BGR8 ||
1282 dstFormat == AV_PIX_FMT_RGB8) {
1283 if (c->dither == SWS_DITHER_AUTO)
1284 c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
1285 if (!(flags & SWS_FULL_CHR_H_INT)) {
1286 if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER) {
1287 av_log(c, AV_LOG_DEBUG,
1288 "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
1289 av_get_pix_fmt_name(dstFormat));
1290 flags |= SWS_FULL_CHR_H_INT;
1294 if (flags & SWS_FULL_CHR_H_INT) {
1295 if (c->dither == SWS_DITHER_BAYER) {
1296 av_log(c, AV_LOG_DEBUG,
1297 "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1298 av_get_pix_fmt_name(dstFormat));
1299 c->dither = SWS_DITHER_ED;
1303 if (isPlanarRGB(dstFormat)) {
1304 if (!(flags & SWS_FULL_CHR_H_INT)) {
1305 av_log(c, AV_LOG_DEBUG,
1306 "%s output is not supported with half chroma resolution, switching to full\n",
1307 av_get_pix_fmt_name(dstFormat));
1308 flags |= SWS_FULL_CHR_H_INT;
1313 /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1314 * chroma interpolation */
1315 if (flags & SWS_FULL_CHR_H_INT &&
1316 isAnyRGB(dstFormat) &&
1317 !isPlanarRGB(dstFormat) &&
1318 dstFormat != AV_PIX_FMT_RGBA64LE &&
1319 dstFormat != AV_PIX_FMT_RGBA64BE &&
1320 dstFormat != AV_PIX_FMT_BGRA64LE &&
1321 dstFormat != AV_PIX_FMT_BGRA64BE &&
1322 dstFormat != AV_PIX_FMT_RGB48LE &&
1323 dstFormat != AV_PIX_FMT_RGB48BE &&
1324 dstFormat != AV_PIX_FMT_BGR48LE &&
1325 dstFormat != AV_PIX_FMT_BGR48BE &&
1326 dstFormat != AV_PIX_FMT_RGBA &&
1327 dstFormat != AV_PIX_FMT_ARGB &&
1328 dstFormat != AV_PIX_FMT_BGRA &&
1329 dstFormat != AV_PIX_FMT_ABGR &&
1330 dstFormat != AV_PIX_FMT_RGB24 &&
1331 dstFormat != AV_PIX_FMT_BGR24 &&
1332 dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1333 dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1334 dstFormat != AV_PIX_FMT_BGR8 &&
1335 dstFormat != AV_PIX_FMT_RGB8
1337 av_log(c, AV_LOG_WARNING,
1338 "full chroma interpolation for destination format '%s' not yet implemented\n",
1339 av_get_pix_fmt_name(dstFormat));
1340 flags &= ~SWS_FULL_CHR_H_INT;
1343 if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1344 c->chrDstHSubSample = 1;
1346 // drop some chroma lines if the user wants it
1347 c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1348 SWS_SRC_V_CHR_DROP_SHIFT;
1349 c->chrSrcVSubSample += c->vChrDrop;
1351 /* drop every other pixel for chroma calculation unless user
1352 * wants full chroma */
1353 if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) &&
1354 srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1355 srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1356 srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1357 srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
1358 srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
1359 srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
1360 srcFormat != AV_PIX_FMT_GBRAP12BE && srcFormat != AV_PIX_FMT_GBRAP12LE &&
1361 srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE &&
1362 srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
1363 srcFormat != AV_PIX_FMT_GBRAP16BE && srcFormat != AV_PIX_FMT_GBRAP16LE &&
1364 ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1365 (flags & SWS_FAST_BILINEAR)))
1366 c->chrSrcHSubSample = 1;
1368 // Note the AV_CEIL_RSHIFT is so that we always round toward +inf.
1369 c->chrSrcW = AV_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
1370 c->chrSrcH = AV_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
1371 c->chrDstW = AV_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
1372 c->chrDstH = AV_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
1374 FF_ALLOCZ_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
1376 c->srcBpc = desc_src->comp[0].depth;
1379 c->dstBpc = desc_dst->comp[0].depth;
1382 if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1384 if (c->dstBpc == 16)
1387 if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1388 c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1389 c->chrDstW >= c->chrSrcW &&
1391 if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1393 && (flags & SWS_FAST_BILINEAR)) {
1394 if (flags & SWS_PRINT_INFO)
1395 av_log(c, AV_LOG_INFO,
1396 "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1398 if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1399 c->canMMXEXTBeUsed = 0;
1401 c->canMMXEXTBeUsed = 0;
1403 c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1404 c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1406 /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1407 * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1409 * n-2 is the last chrominance sample available.
1410 * This is not perfect, but no one should notice the difference, the more
1411 * correct variant would be like the vertical one, but that would require
1412 * some special code for the first and last pixel */
1413 if (flags & SWS_FAST_BILINEAR) {
1414 if (c->canMMXEXTBeUsed) {
1418 // we don't use the x86 asm scaler if MMX is available
1419 else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1420 c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
1421 c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1425 // hardcoded for now
1426 c->gamma_value = 2.2;
1427 tmpFmt = AV_PIX_FMT_RGBA64LE;
1430 if (!unscaled && c->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
1432 c->cascaded_context[0] = NULL;
1434 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1435 srcW, srcH, tmpFmt, 64);
1439 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1441 flags, NULL, NULL, c->param);
1442 if (!c->cascaded_context[0]) {
1446 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
1448 flags, srcFilter, dstFilter, c->param);
1450 if (!c->cascaded_context[1])
1453 c2 = c->cascaded_context[1];
1454 c2->is_internal_gamma = 1;
1455 c2->gamma = alloc_gamma_tbl( c->gamma_value);
1456 c2->inv_gamma = alloc_gamma_tbl(1.f/c->gamma_value);
1457 if (!c2->gamma || !c2->inv_gamma)
1458 return AVERROR(ENOMEM);
1460 // is_internal_flag is set after creating the context
1461 // to properly create the gamma convert FilterDescriptor
1462 // we have to re-initialize it
1463 ff_free_filters(c2);
1464 if (ff_init_filters(c2) < 0) {
1465 sws_freeContext(c2);
1469 c->cascaded_context[2] = NULL;
1470 if (dstFormat != tmpFmt) {
1471 ret = av_image_alloc(c->cascaded1_tmp, c->cascaded1_tmpStride,
1472 dstW, dstH, tmpFmt, 64);
1476 c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
1477 dstW, dstH, dstFormat,
1478 flags, NULL, NULL, c->param);
1479 if (!c->cascaded_context[2])
1485 if (isBayer(srcFormat)) {
1487 (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P)) {
1488 enum AVPixelFormat tmpFormat = AV_PIX_FMT_RGB24;
1490 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1491 srcW, srcH, tmpFormat, 64);
1495 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1496 srcW, srcH, tmpFormat,
1497 flags, srcFilter, NULL, c->param);
1498 if (!c->cascaded_context[0])
1501 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
1502 dstW, dstH, dstFormat,
1503 flags, NULL, dstFilter, c->param);
1504 if (!c->cascaded_context[1])
1510 if (CONFIG_SWSCALE_ALPHA && isALPHA(srcFormat) && !isALPHA(dstFormat)) {
1511 enum AVPixelFormat tmpFormat = alphaless_fmt(srcFormat);
1513 if (tmpFormat != AV_PIX_FMT_NONE && c->alphablend != SWS_ALPHA_BLEND_NONE)
1515 dstFormat != tmpFormat ||
1516 usesHFilter || usesVFilter ||
1517 c->srcRange != c->dstRange
1519 c->cascaded_mainindex = 1;
1520 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1521 srcW, srcH, tmpFormat, 64);
1525 c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, srcFormat,
1526 srcW, srcH, tmpFormat,
1528 if (!c->cascaded_context[0])
1530 c->cascaded_context[0]->alphablend = c->alphablend;
1531 ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
1535 c->cascaded_context[1] = sws_alloc_set_opts(srcW, srcH, tmpFormat,
1536 dstW, dstH, dstFormat,
1538 if (!c->cascaded_context[1])
1541 c->cascaded_context[1]->srcRange = c->srcRange;
1542 c->cascaded_context[1]->dstRange = c->dstRange;
1543 ret = sws_init_context(c->cascaded_context[1], srcFilter , dstFilter);
1551 #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
1553 /* precalculate horizontal scaler filter coefficients */
1555 #if HAVE_MMXEXT_INLINE
1556 // can't downscale !!!
1557 if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1558 c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1560 c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1561 NULL, NULL, NULL, 4);
1564 c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
1565 PROT_READ | PROT_WRITE,
1566 MAP_PRIVATE | MAP_ANONYMOUS,
1568 c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
1569 PROT_READ | PROT_WRITE,
1570 MAP_PRIVATE | MAP_ANONYMOUS,
1572 #elif HAVE_VIRTUALALLOC
1573 c->lumMmxextFilterCode = VirtualAlloc(NULL,
1574 c->lumMmxextFilterCodeSize,
1576 PAGE_EXECUTE_READWRITE);
1577 c->chrMmxextFilterCode = VirtualAlloc(NULL,
1578 c->chrMmxextFilterCodeSize,
1580 PAGE_EXECUTE_READWRITE);
1582 c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
1583 c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
1586 #ifdef MAP_ANONYMOUS
1587 if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1589 if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
1592 av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1593 return AVERROR(ENOMEM);
1596 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
1597 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
1598 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
1599 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);
1601 ff_init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
1602 c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1603 ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1604 c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1607 if ( mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
1608 || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
1609 av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
1614 #endif /* HAVE_MMXEXT_INLINE */
1616 const int filterAlign = X86_MMX(cpu_flags) ? 4 :
1617 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1619 if ((ret = initFilter(&c->hLumFilter, &c->hLumFilterPos,
1620 &c->hLumFilterSize, c->lumXInc,
1621 srcW, dstW, filterAlign, 1 << 14,
1622 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1623 cpu_flags, srcFilter->lumH, dstFilter->lumH,
1625 get_local_pos(c, 0, 0, 0),
1626 get_local_pos(c, 0, 0, 0))) < 0)
1628 if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
1629 &c->hChrFilterSize, c->chrXInc,
1630 c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1631 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1632 cpu_flags, srcFilter->chrH, dstFilter->chrH,
1634 get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
1635 get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
1638 } // initialize horizontal stuff
1640 /* precalculate vertical scaler filter coefficients */
1642 const int filterAlign = X86_MMX(cpu_flags) ? 2 :
1643 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1645 if ((ret = initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
1646 c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1647 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1648 cpu_flags, srcFilter->lumV, dstFilter->lumV,
1650 get_local_pos(c, 0, 0, 1),
1651 get_local_pos(c, 0, 0, 1))) < 0)
1653 if ((ret = initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
1654 c->chrYInc, c->chrSrcH, c->chrDstH,
1655 filterAlign, (1 << 12),
1656 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1657 cpu_flags, srcFilter->chrV, dstFilter->chrV,
1659 get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
1660 get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1))) < 0)
1665 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
1666 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);
1668 for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1670 short *p = (short *)&c->vYCoeffsBank[i];
1671 for (j = 0; j < 8; j++)
1672 p[j] = c->vLumFilter[i];
1675 for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1677 short *p = (short *)&c->vCCoeffsBank[i];
1678 for (j = 0; j < 8; j++)
1679 p[j] = c->vChrFilter[i];
1684 // calculate buffer sizes so that they won't run out while handling these damn slices
1685 c->vLumBufSize = c->vLumFilterSize;
1686 c->vChrBufSize = c->vChrFilterSize;
1687 for (i = 0; i < dstH; i++) {
1688 int chrI = (int64_t)i * c->chrDstH / dstH;
1689 int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
1690 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
1691 << c->chrSrcVSubSample));
1693 nextSlice >>= c->chrSrcVSubSample;
1694 nextSlice <<= c->chrSrcVSubSample;
1695 if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
1696 c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
1697 if (c->vChrFilterPos[chrI] + c->vChrBufSize <
1698 (nextSlice >> c->chrSrcVSubSample))
1699 c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
1700 c->vChrFilterPos[chrI];
1703 for (i = 0; i < 4; i++)
1704 FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);
1706 /* Allocate pixbufs (we use dynamic allocation because otherwise we would
1707 * need to allocate several megabytes to handle all possible cases) */
1708 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1709 FF_ALLOCZ_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1710 FF_ALLOCZ_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1711 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
1712 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1713 /* Note we need at least one pixel more at the end because of the MMX code
1714 * (just in case someone wants to replace the 4000/8000). */
1715 /* align at 16 bytes for AltiVec */
1716 for (i = 0; i < c->vLumBufSize; i++) {
1717 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
1718 dst_stride + 16, fail);
1719 c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
1721 // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1722 c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1723 c->uv_offx2 = dst_stride + 16;
1724 for (i = 0; i < c->vChrBufSize; i++) {
1725 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
1726 dst_stride * 2 + 32, fail);
1727 c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
1728 c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
1729 = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
1731 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1732 for (i = 0; i < c->vLumBufSize; i++) {
1733 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
1734 dst_stride + 16, fail);
1735 c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
1738 // try to avoid drawing green stuff between the right end and the stride end
1739 for (i = 0; i < c->vChrBufSize; i++)
1740 if(desc_dst->comp[0].depth == 16){
1741 av_assert0(c->dstBpc > 14);
1742 for(j=0; j<dst_stride/2+1; j++)
1743 ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
1745 for(j=0; j<dst_stride+1; j++)
1746 ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
1748 av_assert0(c->chrDstH <= dstH);
1750 if (flags & SWS_PRINT_INFO) {
1751 const char *scaler = NULL, *cpucaps;
1753 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
1754 if (flags & scale_algorithms[i].flag) {
1755 scaler = scale_algorithms[i].description;
1760 scaler = "ehh flags invalid?!";
1761 av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
1763 av_get_pix_fmt_name(srcFormat),
1765 dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
1766 dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1767 dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1772 av_get_pix_fmt_name(dstFormat));
1774 if (INLINE_MMXEXT(cpu_flags))
1776 else if (INLINE_AMD3DNOW(cpu_flags))
1778 else if (INLINE_MMX(cpu_flags))
1780 else if (PPC_ALTIVEC(cpu_flags))
1781 cpucaps = "AltiVec";
1785 av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);
1787 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1788 av_log(c, AV_LOG_DEBUG,
1789 "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1790 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1791 av_log(c, AV_LOG_DEBUG,
1792 "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1793 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1794 c->chrXInc, c->chrYInc);
1797 /* alpha blend special case, note this has been split via cascaded contexts if its scaled */
1798 if (unscaled && !usesHFilter && !usesVFilter &&
1799 c->alphablend != SWS_ALPHA_BLEND_NONE &&
1800 isALPHA(srcFormat) &&
1801 (c->srcRange == c->dstRange || isAnyRGB(dstFormat)) &&
1802 alphaless_fmt(srcFormat) == dstFormat
1804 c->swscale = ff_sws_alphablendaway;
1806 if (flags & SWS_PRINT_INFO)
1807 av_log(c, AV_LOG_INFO,
1808 "using alpha blendaway %s -> %s special converter\n",
1809 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1813 /* unscaled special cases */
1814 if (unscaled && !usesHFilter && !usesVFilter &&
1815 (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
1816 ff_get_unscaled_swscale(c);
1819 if (flags & SWS_PRINT_INFO)
1820 av_log(c, AV_LOG_INFO,
1821 "using unscaled %s -> %s special converter\n",
1822 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1827 c->swscale = ff_getSwsFunc(c);
1828 return ff_init_filters(c);
1829 fail: // FIXME replace things by appropriate error codes
1830 if (ret == RETCODE_USE_CASCADE) {
1831 int tmpW = sqrt(srcW * (int64_t)dstW);
1832 int tmpH = sqrt(srcH * (int64_t)dstH);
1833 enum AVPixelFormat tmpFormat = AV_PIX_FMT_YUV420P;
1835 if (isALPHA(srcFormat))
1836 tmpFormat = AV_PIX_FMT_YUVA420P;
1838 if (srcW*(int64_t)srcH <= 4LL*dstW*dstH)
1839 return AVERROR(EINVAL);
1841 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1842 tmpW, tmpH, tmpFormat, 64);
1846 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1847 tmpW, tmpH, tmpFormat,
1848 flags, srcFilter, NULL, c->param);
1849 if (!c->cascaded_context[0])
1852 c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
1853 dstW, dstH, dstFormat,
1854 flags, NULL, dstFilter, c->param);
1855 if (!c->cascaded_context[1])
1862 SwsContext *sws_alloc_set_opts(int srcW, int srcH, enum AVPixelFormat srcFormat,
1863 int dstW, int dstH, enum AVPixelFormat dstFormat,
1864 int flags, const double *param)
1868 if (!(c = sws_alloc_context()))
1876 c->srcFormat = srcFormat;
1877 c->dstFormat = dstFormat;
1880 c->param[0] = param[0];
1881 c->param[1] = param[1];
1887 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
1888 int dstW, int dstH, enum AVPixelFormat dstFormat,
1889 int flags, SwsFilter *srcFilter,
1890 SwsFilter *dstFilter, const double *param)
1894 c = sws_alloc_set_opts(srcW, srcH, srcFormat,
1895 dstW, dstH, dstFormat,
1900 if (sws_init_context(c, srcFilter, dstFilter) < 0) {
1908 static int isnan_vec(SwsVector *a)
1911 for (i=0; i<a->length; i++)
1912 if (isnan(a->coeff[i]))
1917 static void makenan_vec(SwsVector *a)
1920 for (i=0; i<a->length; i++)
1924 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
1925 float lumaSharpen, float chromaSharpen,
1926 float chromaHShift, float chromaVShift,
1929 SwsFilter *filter = av_malloc(sizeof(SwsFilter));
1933 if (lumaGBlur != 0.0) {
1934 filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
1935 filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
1937 filter->lumH = sws_getIdentityVec();
1938 filter->lumV = sws_getIdentityVec();
1941 if (chromaGBlur != 0.0) {
1942 filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
1943 filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
1945 filter->chrH = sws_getIdentityVec();
1946 filter->chrV = sws_getIdentityVec();
1949 if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
1952 if (chromaSharpen != 0.0) {
1953 SwsVector *id = sws_getIdentityVec();
1956 sws_scaleVec(filter->chrH, -chromaSharpen);
1957 sws_scaleVec(filter->chrV, -chromaSharpen);
1958 sws_addVec(filter->chrH, id);
1959 sws_addVec(filter->chrV, id);
1963 if (lumaSharpen != 0.0) {
1964 SwsVector *id = sws_getIdentityVec();
1967 sws_scaleVec(filter->lumH, -lumaSharpen);
1968 sws_scaleVec(filter->lumV, -lumaSharpen);
1969 sws_addVec(filter->lumH, id);
1970 sws_addVec(filter->lumV, id);
1974 if (chromaHShift != 0.0)
1975 sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
1977 if (chromaVShift != 0.0)
1978 sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
1980 sws_normalizeVec(filter->chrH, 1.0);
1981 sws_normalizeVec(filter->chrV, 1.0);
1982 sws_normalizeVec(filter->lumH, 1.0);
1983 sws_normalizeVec(filter->lumV, 1.0);
1985 if (isnan_vec(filter->chrH) ||
1986 isnan_vec(filter->chrV) ||
1987 isnan_vec(filter->lumH) ||
1988 isnan_vec(filter->lumV))
1992 sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
1994 sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
1999 sws_freeVec(filter->lumH);
2000 sws_freeVec(filter->lumV);
2001 sws_freeVec(filter->chrH);
2002 sws_freeVec(filter->chrV);
2007 SwsVector *sws_allocVec(int length)
2011 if(length <= 0 || length > INT_MAX/ sizeof(double))
2014 vec = av_malloc(sizeof(SwsVector));
2017 vec->length = length;
2018 vec->coeff = av_malloc(sizeof(double) * length);
2024 SwsVector *sws_getGaussianVec(double variance, double quality)
2026 const int length = (int)(variance * quality + 0.5) | 1;
2028 double middle = (length - 1) * 0.5;
2031 if(variance < 0 || quality < 0)
2034 vec = sws_allocVec(length);
2039 for (i = 0; i < length; i++) {
2040 double dist = i - middle;
2041 vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
2042 sqrt(2 * variance * M_PI);
2045 sws_normalizeVec(vec, 1.0);
2050 SwsVector *sws_getConstVec(double c, int length)
2053 SwsVector *vec = sws_allocVec(length);
2058 for (i = 0; i < length; i++)
2064 SwsVector *sws_getIdentityVec(void)
2066 return sws_getConstVec(1.0, 1);
2069 static double sws_dcVec(SwsVector *a)
2074 for (i = 0; i < a->length; i++)
2080 void sws_scaleVec(SwsVector *a, double scalar)
2084 for (i = 0; i < a->length; i++)
2085 a->coeff[i] *= scalar;
2088 void sws_normalizeVec(SwsVector *a, double height)
2090 sws_scaleVec(a, height / sws_dcVec(a));
2093 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
2095 int length = a->length + b->length - 1;
2097 SwsVector *vec = sws_getConstVec(0.0, length);
2102 for (i = 0; i < a->length; i++) {
2103 for (j = 0; j < b->length; j++) {
2104 vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
2111 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
2113 int length = FFMAX(a->length, b->length);
2115 SwsVector *vec = sws_getConstVec(0.0, length);
2120 for (i = 0; i < a->length; i++)
2121 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
2122 for (i = 0; i < b->length; i++)
2123 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
2128 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
2130 int length = FFMAX(a->length, b->length);
2132 SwsVector *vec = sws_getConstVec(0.0, length);
2137 for (i = 0; i < a->length; i++)
2138 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
2139 for (i = 0; i < b->length; i++)
2140 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];
2145 /* shift left / or right if "shift" is negative */
2146 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
2148 int length = a->length + FFABS(shift) * 2;
2150 SwsVector *vec = sws_getConstVec(0.0, length);
2155 for (i = 0; i < a->length; i++) {
2156 vec->coeff[i + (length - 1) / 2 -
2157 (a->length - 1) / 2 - shift] = a->coeff[i];
2163 void sws_shiftVec(SwsVector *a, int shift)
2165 SwsVector *shifted = sws_getShiftedVec(a, shift);
2171 a->coeff = shifted->coeff;
2172 a->length = shifted->length;
2176 void sws_addVec(SwsVector *a, SwsVector *b)
2178 SwsVector *sum = sws_sumVec(a, b);
2184 a->coeff = sum->coeff;
2185 a->length = sum->length;
2189 void sws_subVec(SwsVector *a, SwsVector *b)
2191 SwsVector *diff = sws_diffVec(a, b);
2197 a->coeff = diff->coeff;
2198 a->length = diff->length;
2202 void sws_convVec(SwsVector *a, SwsVector *b)
2204 SwsVector *conv = sws_getConvVec(a, b);
2210 a->coeff = conv->coeff;
2211 a->length = conv->length;
2215 SwsVector *sws_cloneVec(SwsVector *a)
2217 SwsVector *vec = sws_allocVec(a->length);
2222 memcpy(vec->coeff, a->coeff, a->length * sizeof(*a->coeff));
2227 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
2234 for (i = 0; i < a->length; i++)
2235 if (a->coeff[i] > max)
2238 for (i = 0; i < a->length; i++)
2239 if (a->coeff[i] < min)
2244 for (i = 0; i < a->length; i++) {
2245 int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
2246 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
2248 av_log(log_ctx, log_level, " ");
2249 av_log(log_ctx, log_level, "|\n");
2253 void sws_freeVec(SwsVector *a)
2257 av_freep(&a->coeff);
2262 void sws_freeFilter(SwsFilter *filter)
2267 sws_freeVec(filter->lumH);
2268 sws_freeVec(filter->lumV);
2269 sws_freeVec(filter->chrH);
2270 sws_freeVec(filter->chrV);
2274 void sws_freeContext(SwsContext *c)
2281 for (i = 0; i < c->vLumBufSize; i++)
2282 av_freep(&c->lumPixBuf[i]);
2283 av_freep(&c->lumPixBuf);
2286 if (c->chrUPixBuf) {
2287 for (i = 0; i < c->vChrBufSize; i++)
2288 av_freep(&c->chrUPixBuf[i]);
2289 av_freep(&c->chrUPixBuf);
2290 av_freep(&c->chrVPixBuf);
2293 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
2294 for (i = 0; i < c->vLumBufSize; i++)
2295 av_freep(&c->alpPixBuf[i]);
2296 av_freep(&c->alpPixBuf);
2299 for (i = 0; i < 4; i++)
2300 av_freep(&c->dither_error[i]);
2302 av_freep(&c->vLumFilter);
2303 av_freep(&c->vChrFilter);
2304 av_freep(&c->hLumFilter);
2305 av_freep(&c->hChrFilter);
2307 av_freep(&c->vYCoeffsBank);
2308 av_freep(&c->vCCoeffsBank);
2311 av_freep(&c->vLumFilterPos);
2312 av_freep(&c->vChrFilterPos);
2313 av_freep(&c->hLumFilterPos);
2314 av_freep(&c->hChrFilterPos);
2318 if (c->lumMmxextFilterCode)
2319 munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
2320 if (c->chrMmxextFilterCode)
2321 munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize);
2322 #elif HAVE_VIRTUALALLOC
2323 if (c->lumMmxextFilterCode)
2324 VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
2325 if (c->chrMmxextFilterCode)
2326 VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
2328 av_free(c->lumMmxextFilterCode);
2329 av_free(c->chrMmxextFilterCode);
2331 c->lumMmxextFilterCode = NULL;
2332 c->chrMmxextFilterCode = NULL;
2333 #endif /* HAVE_MMX_INLINE */
2335 av_freep(&c->yuvTable);
2336 av_freep(&c->formatConvBuffer);
2338 sws_freeContext(c->cascaded_context[0]);
2339 sws_freeContext(c->cascaded_context[1]);
2340 sws_freeContext(c->cascaded_context[2]);
2341 memset(c->cascaded_context, 0, sizeof(c->cascaded_context));
2342 av_freep(&c->cascaded_tmp[0]);
2343 av_freep(&c->cascaded1_tmp[0]);
2345 av_freep(&c->gamma);
2346 av_freep(&c->inv_gamma);
2353 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
2354 int srcH, enum AVPixelFormat srcFormat,
2356 enum AVPixelFormat dstFormat, int flags,
2357 SwsFilter *srcFilter,
2358 SwsFilter *dstFilter,
2359 const double *param)
2361 static const double default_param[2] = { SWS_PARAM_DEFAULT,
2362 SWS_PARAM_DEFAULT };
2363 int64_t src_h_chr_pos = -513, dst_h_chr_pos = -513,
2364 src_v_chr_pos = -513, dst_v_chr_pos = -513;
2367 param = default_param;
2370 (context->srcW != srcW ||
2371 context->srcH != srcH ||
2372 context->srcFormat != srcFormat ||
2373 context->dstW != dstW ||
2374 context->dstH != dstH ||
2375 context->dstFormat != dstFormat ||
2376 context->flags != flags ||
2377 context->param[0] != param[0] ||
2378 context->param[1] != param[1])) {
2380 av_opt_get_int(context, "src_h_chr_pos", 0, &src_h_chr_pos);
2381 av_opt_get_int(context, "src_v_chr_pos", 0, &src_v_chr_pos);
2382 av_opt_get_int(context, "dst_h_chr_pos", 0, &dst_h_chr_pos);
2383 av_opt_get_int(context, "dst_v_chr_pos", 0, &dst_v_chr_pos);
2384 sws_freeContext(context);
2389 if (!(context = sws_alloc_context()))
2391 context->srcW = srcW;
2392 context->srcH = srcH;
2393 context->srcFormat = srcFormat;
2394 context->dstW = dstW;
2395 context->dstH = dstH;
2396 context->dstFormat = dstFormat;
2397 context->flags = flags;
2398 context->param[0] = param[0];
2399 context->param[1] = param[1];
2401 av_opt_set_int(context, "src_h_chr_pos", src_h_chr_pos, 0);
2402 av_opt_set_int(context, "src_v_chr_pos", src_v_chr_pos, 0);
2403 av_opt_set_int(context, "dst_h_chr_pos", dst_h_chr_pos, 0);
2404 av_opt_set_int(context, "dst_v_chr_pos", dst_v_chr_pos, 0);
2406 if (sws_init_context(context, srcFilter, dstFilter) < 0) {
2407 sws_freeContext(context);