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/aarch64/cpu.h"
53 #include "libavutil/ppc/cpu.h"
54 #include "libavutil/x86/asm.h"
55 #include "libavutil/x86/cpu.h"
58 #include "swscale_internal.h"
60 static void handle_formats(SwsContext *c);
62 unsigned swscale_version(void)
64 av_assert0(LIBSWSCALE_VERSION_MICRO >= 100);
65 return LIBSWSCALE_VERSION_INT;
68 const char *swscale_configuration(void)
70 return FFMPEG_CONFIGURATION;
73 const char *swscale_license(void)
75 #define LICENSE_PREFIX "libswscale license: "
76 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
79 typedef struct FormatEntry {
80 uint8_t is_supported_in :1;
81 uint8_t is_supported_out :1;
82 uint8_t is_supported_endianness :1;
85 static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
86 [AV_PIX_FMT_YUV420P] = { 1, 1 },
87 [AV_PIX_FMT_YUYV422] = { 1, 1 },
88 [AV_PIX_FMT_RGB24] = { 1, 1 },
89 [AV_PIX_FMT_BGR24] = { 1, 1 },
90 [AV_PIX_FMT_YUV422P] = { 1, 1 },
91 [AV_PIX_FMT_YUV444P] = { 1, 1 },
92 [AV_PIX_FMT_YUV410P] = { 1, 1 },
93 [AV_PIX_FMT_YUV411P] = { 1, 1 },
94 [AV_PIX_FMT_GRAY8] = { 1, 1 },
95 [AV_PIX_FMT_MONOWHITE] = { 1, 1 },
96 [AV_PIX_FMT_MONOBLACK] = { 1, 1 },
97 [AV_PIX_FMT_PAL8] = { 1, 0 },
98 [AV_PIX_FMT_YUVJ420P] = { 1, 1 },
99 [AV_PIX_FMT_YUVJ411P] = { 1, 1 },
100 [AV_PIX_FMT_YUVJ422P] = { 1, 1 },
101 [AV_PIX_FMT_YUVJ444P] = { 1, 1 },
102 [AV_PIX_FMT_YVYU422] = { 1, 1 },
103 [AV_PIX_FMT_UYVY422] = { 1, 1 },
104 [AV_PIX_FMT_UYYVYY411] = { 0, 0 },
105 [AV_PIX_FMT_BGR8] = { 1, 1 },
106 [AV_PIX_FMT_BGR4] = { 0, 1 },
107 [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
108 [AV_PIX_FMT_RGB8] = { 1, 1 },
109 [AV_PIX_FMT_RGB4] = { 0, 1 },
110 [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
111 [AV_PIX_FMT_NV12] = { 1, 1 },
112 [AV_PIX_FMT_NV21] = { 1, 1 },
113 [AV_PIX_FMT_ARGB] = { 1, 1 },
114 [AV_PIX_FMT_RGBA] = { 1, 1 },
115 [AV_PIX_FMT_ABGR] = { 1, 1 },
116 [AV_PIX_FMT_BGRA] = { 1, 1 },
117 [AV_PIX_FMT_0RGB] = { 1, 1 },
118 [AV_PIX_FMT_RGB0] = { 1, 1 },
119 [AV_PIX_FMT_0BGR] = { 1, 1 },
120 [AV_PIX_FMT_BGR0] = { 1, 1 },
121 [AV_PIX_FMT_GRAY16BE] = { 1, 1 },
122 [AV_PIX_FMT_GRAY16LE] = { 1, 1 },
123 [AV_PIX_FMT_YUV440P] = { 1, 1 },
124 [AV_PIX_FMT_YUVJ440P] = { 1, 1 },
125 [AV_PIX_FMT_YUV440P10LE] = { 1, 1 },
126 [AV_PIX_FMT_YUV440P10BE] = { 1, 1 },
127 [AV_PIX_FMT_YUV440P12LE] = { 1, 1 },
128 [AV_PIX_FMT_YUV440P12BE] = { 1, 1 },
129 [AV_PIX_FMT_YUVA420P] = { 1, 1 },
130 [AV_PIX_FMT_YUVA422P] = { 1, 1 },
131 [AV_PIX_FMT_YUVA444P] = { 1, 1 },
132 [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
133 [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
134 [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
135 [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
136 [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
137 [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
138 [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
139 [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
140 [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
141 [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
142 [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
143 [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
144 [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
145 [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
146 [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
147 [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
148 [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
149 [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
150 [AV_PIX_FMT_RGB48BE] = { 1, 1 },
151 [AV_PIX_FMT_RGB48LE] = { 1, 1 },
152 [AV_PIX_FMT_RGBA64BE] = { 1, 1, 1 },
153 [AV_PIX_FMT_RGBA64LE] = { 1, 1, 1 },
154 [AV_PIX_FMT_RGB565BE] = { 1, 1 },
155 [AV_PIX_FMT_RGB565LE] = { 1, 1 },
156 [AV_PIX_FMT_RGB555BE] = { 1, 1 },
157 [AV_PIX_FMT_RGB555LE] = { 1, 1 },
158 [AV_PIX_FMT_BGR565BE] = { 1, 1 },
159 [AV_PIX_FMT_BGR565LE] = { 1, 1 },
160 [AV_PIX_FMT_BGR555BE] = { 1, 1 },
161 [AV_PIX_FMT_BGR555LE] = { 1, 1 },
162 [AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
163 [AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
164 [AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
165 [AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
166 [AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
167 [AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
168 [AV_PIX_FMT_RGB444LE] = { 1, 1 },
169 [AV_PIX_FMT_RGB444BE] = { 1, 1 },
170 [AV_PIX_FMT_BGR444LE] = { 1, 1 },
171 [AV_PIX_FMT_BGR444BE] = { 1, 1 },
172 [AV_PIX_FMT_YA8] = { 1, 1 },
173 [AV_PIX_FMT_YA16BE] = { 1, 0 },
174 [AV_PIX_FMT_YA16LE] = { 1, 0 },
175 [AV_PIX_FMT_BGR48BE] = { 1, 1 },
176 [AV_PIX_FMT_BGR48LE] = { 1, 1 },
177 [AV_PIX_FMT_BGRA64BE] = { 1, 1, 1 },
178 [AV_PIX_FMT_BGRA64LE] = { 1, 1, 1 },
179 [AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
180 [AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
181 [AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
182 [AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
183 [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
184 [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
185 [AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
186 [AV_PIX_FMT_YUV420P14LE] = { 1, 1 },
187 [AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
188 [AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
189 [AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
190 [AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
191 [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
192 [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
193 [AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
194 [AV_PIX_FMT_YUV422P14LE] = { 1, 1 },
195 [AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
196 [AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
197 [AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
198 [AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
199 [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
200 [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
201 [AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
202 [AV_PIX_FMT_YUV444P14LE] = { 1, 1 },
203 [AV_PIX_FMT_GBRP] = { 1, 1 },
204 [AV_PIX_FMT_GBRP9LE] = { 1, 1 },
205 [AV_PIX_FMT_GBRP9BE] = { 1, 1 },
206 [AV_PIX_FMT_GBRP10LE] = { 1, 1 },
207 [AV_PIX_FMT_GBRP10BE] = { 1, 1 },
208 [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
209 [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
210 [AV_PIX_FMT_GBRAP12LE] = { 1, 0 },
211 [AV_PIX_FMT_GBRAP12BE] = { 1, 0 },
212 [AV_PIX_FMT_GBRP14LE] = { 1, 1 },
213 [AV_PIX_FMT_GBRP14BE] = { 1, 1 },
214 [AV_PIX_FMT_GBRP16LE] = { 1, 0 },
215 [AV_PIX_FMT_GBRP16BE] = { 1, 0 },
216 [AV_PIX_FMT_GBRAP] = { 1, 1 },
217 [AV_PIX_FMT_GBRAP16LE] = { 1, 0 },
218 [AV_PIX_FMT_GBRAP16BE] = { 1, 0 },
219 [AV_PIX_FMT_BAYER_BGGR8] = { 1, 0 },
220 [AV_PIX_FMT_BAYER_RGGB8] = { 1, 0 },
221 [AV_PIX_FMT_BAYER_GBRG8] = { 1, 0 },
222 [AV_PIX_FMT_BAYER_GRBG8] = { 1, 0 },
223 [AV_PIX_FMT_BAYER_BGGR16LE] = { 1, 0 },
224 [AV_PIX_FMT_BAYER_BGGR16BE] = { 1, 0 },
225 [AV_PIX_FMT_BAYER_RGGB16LE] = { 1, 0 },
226 [AV_PIX_FMT_BAYER_RGGB16BE] = { 1, 0 },
227 [AV_PIX_FMT_BAYER_GBRG16LE] = { 1, 0 },
228 [AV_PIX_FMT_BAYER_GBRG16BE] = { 1, 0 },
229 [AV_PIX_FMT_BAYER_GRBG16LE] = { 1, 0 },
230 [AV_PIX_FMT_BAYER_GRBG16BE] = { 1, 0 },
231 [AV_PIX_FMT_XYZ12BE] = { 1, 1, 1 },
232 [AV_PIX_FMT_XYZ12LE] = { 1, 1, 1 },
233 [AV_PIX_FMT_AYUV64LE] = { 1, 1},
234 [AV_PIX_FMT_P010LE] = { 1, 0 },
235 [AV_PIX_FMT_P010BE] = { 1, 0 },
238 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
240 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
241 format_entries[pix_fmt].is_supported_in : 0;
244 int sws_isSupportedOutput(enum AVPixelFormat pix_fmt)
246 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
247 format_entries[pix_fmt].is_supported_out : 0;
250 int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
252 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
253 format_entries[pix_fmt].is_supported_endianness : 0;
256 static double getSplineCoeff(double a, double b, double c, double d,
260 return ((d * dist + c) * dist + b) * dist + a;
262 return getSplineCoeff(0.0,
263 b + 2.0 * c + 3.0 * d,
265 -b - 3.0 * c - 6.0 * d,
269 static av_cold int get_local_pos(SwsContext *s, int chr_subsample, int pos, int dir)
271 if (pos == -1 || pos <= -513) {
272 pos = (128 << chr_subsample) - 128;
274 pos += 128; // relative to ideal left edge
275 return pos >> chr_subsample;
279 int flag; ///< flag associated to the algorithm
280 const char *description; ///< human-readable description
281 int size_factor; ///< size factor used when initing the filters
284 static const ScaleAlgorithm scale_algorithms[] = {
285 { SWS_AREA, "area averaging", 1 /* downscale only, for upscale it is bilinear */ },
286 { SWS_BICUBIC, "bicubic", 4 },
287 { SWS_BICUBLIN, "luma bicubic / chroma bilinear", -1 },
288 { SWS_BILINEAR, "bilinear", 2 },
289 { SWS_FAST_BILINEAR, "fast bilinear", -1 },
290 { SWS_GAUSS, "Gaussian", 8 /* infinite ;) */ },
291 { SWS_LANCZOS, "Lanczos", -1 /* custom */ },
292 { SWS_POINT, "nearest neighbor / point", -1 },
293 { SWS_SINC, "sinc", 20 /* infinite ;) */ },
294 { SWS_SPLINE, "bicubic spline", 20 /* infinite :)*/ },
295 { SWS_X, "experimental", 8 },
298 static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
299 int *outFilterSize, int xInc, int srcW,
300 int dstW, int filterAlign, int one,
301 int flags, int cpu_flags,
302 SwsVector *srcFilter, SwsVector *dstFilter,
303 double param[2], int srcPos, int dstPos)
309 int64_t *filter = NULL;
310 int64_t *filter2 = NULL;
311 const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
314 emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
316 // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
317 FF_ALLOC_ARRAY_OR_GOTO(NULL, *filterPos, (dstW + 3), sizeof(**filterPos), fail);
319 if (FFABS(xInc - 0x10000) < 10 && srcPos == dstPos) { // unscaled
322 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter,
323 dstW, sizeof(*filter) * filterSize, fail);
325 for (i = 0; i < dstW; i++) {
326 filter[i * filterSize] = fone;
329 } else if (flags & SWS_POINT) { // lame looking point sampling mode
333 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
334 dstW, sizeof(*filter) * filterSize, fail);
336 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
337 for (i = 0; i < dstW; i++) {
338 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
340 (*filterPos)[i] = xx;
344 } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
345 (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
349 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
350 dstW, sizeof(*filter) * filterSize, fail);
352 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
353 for (i = 0; i < dstW; i++) {
354 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
357 (*filterPos)[i] = xx;
358 // bilinear upscale / linear interpolate / area averaging
359 for (j = 0; j < filterSize; j++) {
360 int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
363 filter[i * filterSize + j] = coeff;
372 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
373 if (flags & scale_algorithms[i].flag && scale_algorithms[i].size_factor > 0) {
374 sizeFactor = scale_algorithms[i].size_factor;
378 if (flags & SWS_LANCZOS)
379 sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
380 av_assert0(sizeFactor > 0);
383 filterSize = 1 + sizeFactor; // upscale
385 filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
387 filterSize = FFMIN(filterSize, srcW - 2);
388 filterSize = FFMAX(filterSize, 1);
390 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
391 dstW, sizeof(*filter) * filterSize, fail);
393 xDstInSrc = ((dstPos*(int64_t)xInc)>>7) - ((srcPos*0x10000LL)>>7);
394 for (i = 0; i < dstW; i++) {
395 int xx = (xDstInSrc - (filterSize - 2) * (1LL<<16)) / (1 << 17);
397 (*filterPos)[i] = xx;
398 for (j = 0; j < filterSize; j++) {
399 int64_t d = (FFABS(((int64_t)xx * (1 << 17)) - xDstInSrc)) << 13;
405 floatd = d * (1.0 / (1 << 30));
407 if (flags & SWS_BICUBIC) {
408 int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
409 int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
411 if (d >= 1LL << 31) {
414 int64_t dd = (d * d) >> 30;
415 int64_t ddd = (dd * d) >> 30;
418 coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
419 (-18 * (1 << 24) + 12 * B + 6 * C) * dd +
420 (6 * (1 << 24) - 2 * B) * (1 << 30);
422 coeff = (-B - 6 * C) * ddd +
423 (6 * B + 30 * C) * dd +
424 (-12 * B - 48 * C) * d +
425 (8 * B + 24 * C) * (1 << 30);
427 coeff /= (1LL<<54)/fone;
430 else if (flags & SWS_X) {
431 double p = param ? param * 0.01 : 0.3;
432 coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
433 coeff *= pow(2.0, -p * d * d);
436 else if (flags & SWS_X) {
437 double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
441 c = cos(floatd * M_PI);
448 coeff = (c * 0.5 + 0.5) * fone;
449 } else if (flags & SWS_AREA) {
450 int64_t d2 = d - (1 << 29);
451 if (d2 * xInc < -(1LL << (29 + 16)))
452 coeff = 1.0 * (1LL << (30 + 16));
453 else if (d2 * xInc < (1LL << (29 + 16)))
454 coeff = -d2 * xInc + (1LL << (29 + 16));
457 coeff *= fone >> (30 + 16);
458 } else if (flags & SWS_GAUSS) {
459 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
460 coeff = exp2(-p * floatd * floatd) * fone;
461 } else if (flags & SWS_SINC) {
462 coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
463 } else if (flags & SWS_LANCZOS) {
464 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
465 coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
466 (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
469 } else if (flags & SWS_BILINEAR) {
470 coeff = (1 << 30) - d;
474 } else if (flags & SWS_SPLINE) {
475 double p = -2.196152422706632;
476 coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
481 filter[i * filterSize + j] = coeff;
484 xDstInSrc += 2 * xInc;
488 /* apply src & dst Filter to filter -> filter2
491 av_assert0(filterSize > 0);
492 filter2Size = filterSize;
494 filter2Size += srcFilter->length - 1;
496 filter2Size += dstFilter->length - 1;
497 av_assert0(filter2Size > 0);
498 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter2, dstW, filter2Size * sizeof(*filter2), fail);
500 for (i = 0; i < dstW; i++) {
504 for (k = 0; k < srcFilter->length; k++) {
505 for (j = 0; j < filterSize; j++)
506 filter2[i * filter2Size + k + j] +=
507 srcFilter->coeff[k] * filter[i * filterSize + j];
510 for (j = 0; j < filterSize; j++)
511 filter2[i * filter2Size + j] = filter[i * filterSize + j];
515 (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
519 /* try to reduce the filter-size (step1 find size and shift left) */
520 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
522 for (i = dstW - 1; i >= 0; i--) {
523 int min = filter2Size;
525 int64_t cutOff = 0.0;
527 /* get rid of near zero elements on the left by shifting left */
528 for (j = 0; j < filter2Size; j++) {
530 cutOff += FFABS(filter2[i * filter2Size]);
532 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
535 /* preserve monotonicity because the core can't handle the
536 * filter otherwise */
537 if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
540 // move filter coefficients left
541 for (k = 1; k < filter2Size; k++)
542 filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
543 filter2[i * filter2Size + k - 1] = 0;
548 /* count near zeros on the right */
549 for (j = filter2Size - 1; j > 0; j--) {
550 cutOff += FFABS(filter2[i * filter2Size + j]);
552 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
557 if (min > minFilterSize)
561 if (PPC_ALTIVEC(cpu_flags)) {
562 // we can handle the special case 4, so we don't want to go the full 8
563 if (minFilterSize < 5)
566 /* We really don't want to waste our time doing useless computation, so
567 * fall back on the scalar C code for very small filters.
568 * Vectorizing is worth it only if you have a decent-sized vector. */
569 if (minFilterSize < 3)
573 if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
574 // special case for unscaled vertical filtering
575 if (minFilterSize == 1 && filterAlign == 2)
579 av_assert0(minFilterSize > 0);
580 filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
581 av_assert0(filterSize > 0);
582 filter = av_malloc_array(dstW, filterSize * sizeof(*filter));
585 if (filterSize >= MAX_FILTER_SIZE * 16 /
586 ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16)) {
587 ret = RETCODE_USE_CASCADE;
590 *outFilterSize = filterSize;
592 if (flags & SWS_PRINT_INFO)
593 av_log(NULL, AV_LOG_VERBOSE,
594 "SwScaler: reducing / aligning filtersize %d -> %d\n",
595 filter2Size, filterSize);
596 /* try to reduce the filter-size (step2 reduce it) */
597 for (i = 0; i < dstW; i++) {
600 for (j = 0; j < filterSize; j++) {
601 if (j >= filter2Size)
602 filter[i * filterSize + j] = 0;
604 filter[i * filterSize + j] = filter2[i * filter2Size + j];
605 if ((flags & SWS_BITEXACT) && j >= minFilterSize)
606 filter[i * filterSize + j] = 0;
610 // FIXME try to align filterPos if possible
613 for (i = 0; i < dstW; i++) {
615 if ((*filterPos)[i] < 0) {
616 // move filter coefficients left to compensate for filterPos
617 for (j = 1; j < filterSize; j++) {
618 int left = FFMAX(j + (*filterPos)[i], 0);
619 filter[i * filterSize + left] += filter[i * filterSize + j];
620 filter[i * filterSize + j] = 0;
625 if ((*filterPos)[i] + filterSize > srcW) {
626 int shift = (*filterPos)[i] + FFMIN(filterSize - srcW, 0);
629 for (j = filterSize - 1; j >= 0; j--) {
630 if ((*filterPos)[i] + j >= srcW) {
631 acc += filter[i * filterSize + j];
632 filter[i * filterSize + j] = 0;
635 for (j = filterSize - 1; j >= 0; j--) {
637 filter[i * filterSize + j] = 0;
639 filter[i * filterSize + j] = filter[i * filterSize + j - shift];
643 (*filterPos)[i]-= shift;
644 filter[i * filterSize + srcW - 1 - (*filterPos)[i]] += acc;
646 av_assert0((*filterPos)[i] >= 0);
647 av_assert0((*filterPos)[i] < srcW);
648 if ((*filterPos)[i] + filterSize > srcW) {
649 for (j = 0; j < filterSize; j++) {
650 av_assert0((*filterPos)[i] + j < srcW || !filter[i * filterSize + j]);
655 // Note the +1 is for the MMX scaler which reads over the end
656 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
657 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, *outFilter,
658 (dstW + 3), *outFilterSize * sizeof(int16_t), fail);
660 /* normalize & store in outFilter */
661 for (i = 0; i < dstW; i++) {
666 for (j = 0; j < filterSize; j++) {
667 sum += filter[i * filterSize + j];
669 sum = (sum + one / 2) / one;
671 av_log(NULL, AV_LOG_WARNING, "SwScaler: zero vector in scaling\n");
674 for (j = 0; j < *outFilterSize; j++) {
675 int64_t v = filter[i * filterSize + j] + error;
676 int intV = ROUNDED_DIV(v, sum);
677 (*outFilter)[i * (*outFilterSize) + j] = intV;
678 error = v - intV * sum;
682 (*filterPos)[dstW + 0] =
683 (*filterPos)[dstW + 1] =
684 (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
685 * read over the end */
686 for (i = 0; i < *outFilterSize; i++) {
687 int k = (dstW - 1) * (*outFilterSize) + i;
688 (*outFilter)[k + 1 * (*outFilterSize)] =
689 (*outFilter)[k + 2 * (*outFilterSize)] =
690 (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
697 av_log(NULL, ret == RETCODE_USE_CASCADE ? AV_LOG_DEBUG : AV_LOG_ERROR, "sws: initFilter failed\n");
703 static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange)
705 int64_t W, V, Z, Cy, Cu, Cv;
706 int64_t vr = table[0];
707 int64_t ub = table[1];
708 int64_t ug = -table[2];
709 int64_t vg = -table[3];
712 uint8_t *p = (uint8_t*)c->input_rgb2yuv_table;
714 static const int8_t map[] = {
715 BY_IDX, GY_IDX, -1 , BY_IDX, BY_IDX, GY_IDX, -1 , BY_IDX,
716 RY_IDX, -1 , GY_IDX, RY_IDX, RY_IDX, -1 , GY_IDX, RY_IDX,
717 RY_IDX, GY_IDX, -1 , RY_IDX, RY_IDX, GY_IDX, -1 , RY_IDX,
718 BY_IDX, -1 , GY_IDX, BY_IDX, BY_IDX, -1 , GY_IDX, BY_IDX,
719 BU_IDX, GU_IDX, -1 , BU_IDX, BU_IDX, GU_IDX, -1 , BU_IDX,
720 RU_IDX, -1 , GU_IDX, RU_IDX, RU_IDX, -1 , GU_IDX, RU_IDX,
721 RU_IDX, GU_IDX, -1 , RU_IDX, RU_IDX, GU_IDX, -1 , RU_IDX,
722 BU_IDX, -1 , GU_IDX, BU_IDX, BU_IDX, -1 , GU_IDX, BU_IDX,
723 BV_IDX, GV_IDX, -1 , BV_IDX, BV_IDX, GV_IDX, -1 , BV_IDX,
724 RV_IDX, -1 , GV_IDX, RV_IDX, RV_IDX, -1 , GV_IDX, RV_IDX,
725 RV_IDX, GV_IDX, -1 , RV_IDX, RV_IDX, GV_IDX, -1 , RV_IDX,
726 BV_IDX, -1 , GV_IDX, BV_IDX, BV_IDX, -1 , GV_IDX, BV_IDX,
727 RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX,
728 BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX,
729 GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 ,
730 -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX,
731 RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX,
732 BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX,
733 GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 ,
734 -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX,
735 RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX,
736 BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX,
737 GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 ,
738 -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, //23
739 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //24
740 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //25
741 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //26
742 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //27
743 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //28
744 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //29
745 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //30
746 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //31
747 BY_IDX, GY_IDX, RY_IDX, -1 , -1 , -1 , -1 , -1 , //32
748 BU_IDX, GU_IDX, RU_IDX, -1 , -1 , -1 , -1 , -1 , //33
749 BV_IDX, GV_IDX, RV_IDX, -1 , -1 , -1 , -1 , -1 , //34
752 dstRange = 0; //FIXME range = 1 is handled elsewhere
762 W = ROUNDED_DIV(ONE*ONE*ug, ub);
763 V = ROUNDED_DIV(ONE*ONE*vg, vr);
766 Cy = ROUNDED_DIV(cy*Z, ONE);
767 Cu = ROUNDED_DIV(ub*Z, ONE);
768 Cv = ROUNDED_DIV(vr*Z, ONE);
770 c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cy);
771 c->input_rgb2yuv_table[GY_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cy);
772 c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cy);
774 c->input_rgb2yuv_table[RU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cu);
775 c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cu);
776 c->input_rgb2yuv_table[BU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W) , Cu);
778 c->input_rgb2yuv_table[RV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z) , Cv);
779 c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cv);
780 c->input_rgb2yuv_table[BV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cv);
782 if(/*!dstRange && */!memcmp(table, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sizeof(ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]))) {
783 c->input_rgb2yuv_table[BY_IDX] = ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
784 c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
785 c->input_rgb2yuv_table[BU_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
786 c->input_rgb2yuv_table[GY_IDX] = ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
787 c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
788 c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
789 c->input_rgb2yuv_table[RY_IDX] = ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
790 c->input_rgb2yuv_table[RV_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
791 c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
793 for(i=0; i<FF_ARRAY_ELEMS(map); i++)
794 AV_WL16(p + 16*4 + 2*i, map[i] >= 0 ? c->input_rgb2yuv_table[map[i]] : 0);
797 static void fill_xyztables(struct SwsContext *c)
800 double xyzgamma = XYZ_GAMMA;
801 double rgbgamma = 1.0 / RGB_GAMMA;
802 double xyzgammainv = 1.0 / XYZ_GAMMA;
803 double rgbgammainv = RGB_GAMMA;
804 static const int16_t xyz2rgb_matrix[3][4] = {
805 {13270, -6295, -2041},
807 { 228, -835, 4329} };
808 static const int16_t rgb2xyz_matrix[3][4] = {
812 static int16_t xyzgamma_tab[4096], rgbgamma_tab[4096], xyzgammainv_tab[4096], rgbgammainv_tab[4096];
814 memcpy(c->xyz2rgb_matrix, xyz2rgb_matrix, sizeof(c->xyz2rgb_matrix));
815 memcpy(c->rgb2xyz_matrix, rgb2xyz_matrix, sizeof(c->rgb2xyz_matrix));
816 c->xyzgamma = xyzgamma_tab;
817 c->rgbgamma = rgbgamma_tab;
818 c->xyzgammainv = xyzgammainv_tab;
819 c->rgbgammainv = rgbgammainv_tab;
821 if (rgbgamma_tab[4095])
824 /* set gamma vectors */
825 for (i = 0; i < 4096; i++) {
826 xyzgamma_tab[i] = lrint(pow(i / 4095.0, xyzgamma) * 4095.0);
827 rgbgamma_tab[i] = lrint(pow(i / 4095.0, rgbgamma) * 4095.0);
828 xyzgammainv_tab[i] = lrint(pow(i / 4095.0, xyzgammainv) * 4095.0);
829 rgbgammainv_tab[i] = lrint(pow(i / 4095.0, rgbgammainv) * 4095.0);
833 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
834 int srcRange, const int table[4], int dstRange,
835 int brightness, int contrast, int saturation)
837 const AVPixFmtDescriptor *desc_dst;
838 const AVPixFmtDescriptor *desc_src;
842 desc_dst = av_pix_fmt_desc_get(c->dstFormat);
843 desc_src = av_pix_fmt_desc_get(c->srcFormat);
845 if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
847 if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
850 if (c->srcRange != srcRange ||
851 c->dstRange != dstRange ||
852 c->brightness != brightness ||
853 c->contrast != contrast ||
854 c->saturation != saturation ||
855 memcmp(c->srcColorspaceTable, inv_table, sizeof(int) * 4) ||
856 memcmp(c->dstColorspaceTable, table, sizeof(int) * 4)
860 memmove(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
861 memmove(c->dstColorspaceTable, table, sizeof(int) * 4);
865 c->brightness = brightness;
866 c->contrast = contrast;
867 c->saturation = saturation;
868 c->srcRange = srcRange;
869 c->dstRange = dstRange;
871 //The srcBpc check is possibly wrong but we seem to lack a definitive reference to test this
872 //and what we have in ticket 2939 looks better with this check
873 if (need_reinit && (c->srcBpc == 8 || !isYUV(c->srcFormat)))
874 ff_sws_init_range_convert(c);
876 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
877 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
879 if (c->cascaded_context[c->cascaded_mainindex])
880 return sws_setColorspaceDetails(c->cascaded_context[c->cascaded_mainindex],inv_table, srcRange,table, dstRange, brightness, contrast, saturation);
885 if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat))) {
886 if (!c->cascaded_context[0] &&
887 memcmp(c->dstColorspaceTable, c->srcColorspaceTable, sizeof(int) * 4) &&
888 c->srcW && c->srcH && c->dstW && c->dstH) {
889 enum AVPixelFormat tmp_format;
890 int tmp_width, tmp_height;
896 av_log(c, AV_LOG_VERBOSE, "YUV color matrix differs for YUV->YUV, using intermediate RGB to convert\n");
898 if (isNBPS(c->dstFormat) || is16BPS(c->dstFormat)) {
899 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
900 tmp_format = AV_PIX_FMT_BGRA64;
902 tmp_format = AV_PIX_FMT_BGR48;
905 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
906 tmp_format = AV_PIX_FMT_BGRA;
908 tmp_format = AV_PIX_FMT_BGR24;
912 if (srcW*srcH > dstW*dstH) {
920 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
921 tmp_width, tmp_height, tmp_format, 64);
925 c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, c->srcFormat,
926 tmp_width, tmp_height, tmp_format,
928 if (!c->cascaded_context[0])
931 c->cascaded_context[0]->alphablend = c->alphablend;
932 ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
935 //we set both src and dst depending on that the RGB side will be ignored
936 sws_setColorspaceDetails(c->cascaded_context[0], inv_table,
937 srcRange, table, dstRange,
938 brightness, contrast, saturation);
940 c->cascaded_context[1] = sws_getContext(tmp_width, tmp_height, tmp_format,
941 dstW, dstH, c->dstFormat,
942 c->flags, NULL, NULL, c->param);
943 if (!c->cascaded_context[1])
945 sws_setColorspaceDetails(c->cascaded_context[1], inv_table,
946 srcRange, table, dstRange,
947 0, 1 << 16, 1 << 16);
953 if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) {
954 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
955 contrast, saturation);
959 ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
960 contrast, saturation);
963 fill_rgb2yuv_table(c, table, dstRange);
968 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
969 int *srcRange, int **table, int *dstRange,
970 int *brightness, int *contrast, int *saturation)
975 *inv_table = c->srcColorspaceTable;
976 *table = c->dstColorspaceTable;
977 *srcRange = c->srcRange;
978 *dstRange = c->dstRange;
979 *brightness = c->brightness;
980 *contrast = c->contrast;
981 *saturation = c->saturation;
986 static int handle_jpeg(enum AVPixelFormat *format)
989 case AV_PIX_FMT_YUVJ420P:
990 *format = AV_PIX_FMT_YUV420P;
992 case AV_PIX_FMT_YUVJ411P:
993 *format = AV_PIX_FMT_YUV411P;
995 case AV_PIX_FMT_YUVJ422P:
996 *format = AV_PIX_FMT_YUV422P;
998 case AV_PIX_FMT_YUVJ444P:
999 *format = AV_PIX_FMT_YUV444P;
1001 case AV_PIX_FMT_YUVJ440P:
1002 *format = AV_PIX_FMT_YUV440P;
1004 case AV_PIX_FMT_GRAY8:
1005 case AV_PIX_FMT_YA8:
1006 case AV_PIX_FMT_GRAY16LE:
1007 case AV_PIX_FMT_GRAY16BE:
1008 case AV_PIX_FMT_YA16BE:
1009 case AV_PIX_FMT_YA16LE:
1016 static int handle_0alpha(enum AVPixelFormat *format)
1019 case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
1020 case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
1021 case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
1022 case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4;
1027 static int handle_xyz(enum AVPixelFormat *format)
1030 case AV_PIX_FMT_XYZ12BE : *format = AV_PIX_FMT_RGB48BE; return 1;
1031 case AV_PIX_FMT_XYZ12LE : *format = AV_PIX_FMT_RGB48LE; return 1;
1036 static void handle_formats(SwsContext *c)
1038 c->src0Alpha |= handle_0alpha(&c->srcFormat);
1039 c->dst0Alpha |= handle_0alpha(&c->dstFormat);
1040 c->srcXYZ |= handle_xyz(&c->srcFormat);
1041 c->dstXYZ |= handle_xyz(&c->dstFormat);
1042 if (c->srcXYZ || c->dstXYZ)
1046 SwsContext *sws_alloc_context(void)
1048 SwsContext *c = av_mallocz(sizeof(SwsContext));
1050 av_assert0(offsetof(SwsContext, redDither) + DITHER32_INT == offsetof(SwsContext, dither32));
1053 c->av_class = &ff_sws_context_class;
1054 av_opt_set_defaults(c);
1060 static uint16_t * alloc_gamma_tbl(double e)
1064 tbl = (uint16_t*)av_malloc(sizeof(uint16_t) * 1 << 16);
1068 for (i = 0; i < 65536; ++i) {
1069 tbl[i] = pow(i / 65535.0, e) * 65535.0;
1074 static enum AVPixelFormat alphaless_fmt(enum AVPixelFormat fmt)
1077 case AV_PIX_FMT_ARGB: return AV_PIX_FMT_RGB24;
1078 case AV_PIX_FMT_RGBA: return AV_PIX_FMT_RGB24;
1079 case AV_PIX_FMT_ABGR: return AV_PIX_FMT_BGR24;
1080 case AV_PIX_FMT_BGRA: return AV_PIX_FMT_BGR24;
1081 case AV_PIX_FMT_YA8: return AV_PIX_FMT_GRAY8;
1083 case AV_PIX_FMT_YUVA420P: return AV_PIX_FMT_YUV420P;
1084 case AV_PIX_FMT_YUVA422P: return AV_PIX_FMT_YUV422P;
1085 case AV_PIX_FMT_YUVA444P: return AV_PIX_FMT_YUV444P;
1087 case AV_PIX_FMT_GBRAP: return AV_PIX_FMT_GBRP;
1089 case AV_PIX_FMT_GBRAP12LE: return AV_PIX_FMT_GBRP12;
1090 case AV_PIX_FMT_GBRAP12BE: return AV_PIX_FMT_GBRP12;
1092 case AV_PIX_FMT_GBRAP16LE: return AV_PIX_FMT_GBRP16;
1093 case AV_PIX_FMT_GBRAP16BE: return AV_PIX_FMT_GBRP16;
1095 case AV_PIX_FMT_RGBA64LE: return AV_PIX_FMT_RGB48;
1096 case AV_PIX_FMT_RGBA64BE: return AV_PIX_FMT_RGB48;
1097 case AV_PIX_FMT_BGRA64LE: return AV_PIX_FMT_BGR48;
1098 case AV_PIX_FMT_BGRA64BE: return AV_PIX_FMT_BGR48;
1100 case AV_PIX_FMT_YA16BE: return AV_PIX_FMT_GRAY16;
1101 case AV_PIX_FMT_YA16LE: return AV_PIX_FMT_GRAY16;
1103 case AV_PIX_FMT_YUVA420P9BE: return AV_PIX_FMT_YUV420P9;
1104 case AV_PIX_FMT_YUVA422P9BE: return AV_PIX_FMT_YUV422P9;
1105 case AV_PIX_FMT_YUVA444P9BE: return AV_PIX_FMT_YUV444P9;
1106 case AV_PIX_FMT_YUVA420P9LE: return AV_PIX_FMT_YUV420P9;
1107 case AV_PIX_FMT_YUVA422P9LE: return AV_PIX_FMT_YUV422P9;
1108 case AV_PIX_FMT_YUVA444P9LE: return AV_PIX_FMT_YUV444P9;
1109 case AV_PIX_FMT_YUVA420P10BE: return AV_PIX_FMT_YUV420P10;
1110 case AV_PIX_FMT_YUVA422P10BE: return AV_PIX_FMT_YUV422P10;
1111 case AV_PIX_FMT_YUVA444P10BE: return AV_PIX_FMT_YUV444P10;
1112 case AV_PIX_FMT_YUVA420P10LE: return AV_PIX_FMT_YUV420P10;
1113 case AV_PIX_FMT_YUVA422P10LE: return AV_PIX_FMT_YUV422P10;
1114 case AV_PIX_FMT_YUVA444P10LE: return AV_PIX_FMT_YUV444P10;
1115 case AV_PIX_FMT_YUVA420P16BE: return AV_PIX_FMT_YUV420P16;
1116 case AV_PIX_FMT_YUVA422P16BE: return AV_PIX_FMT_YUV422P16;
1117 case AV_PIX_FMT_YUVA444P16BE: return AV_PIX_FMT_YUV444P16;
1118 case AV_PIX_FMT_YUVA420P16LE: return AV_PIX_FMT_YUV420P16;
1119 case AV_PIX_FMT_YUVA422P16LE: return AV_PIX_FMT_YUV422P16;
1120 case AV_PIX_FMT_YUVA444P16LE: return AV_PIX_FMT_YUV444P16;
1122 // case AV_PIX_FMT_AYUV64LE:
1123 // case AV_PIX_FMT_AYUV64BE:
1124 // case AV_PIX_FMT_PAL8:
1125 default: return AV_PIX_FMT_NONE;
1129 av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
1130 SwsFilter *dstFilter)
1133 int usesVFilter, usesHFilter;
1135 SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
1140 int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
1141 int flags, cpu_flags;
1142 enum AVPixelFormat srcFormat = c->srcFormat;
1143 enum AVPixelFormat dstFormat = c->dstFormat;
1144 const AVPixFmtDescriptor *desc_src;
1145 const AVPixFmtDescriptor *desc_dst;
1147 enum AVPixelFormat tmpFmt;
1149 cpu_flags = av_get_cpu_flags();
1153 ff_sws_rgb2rgb_init();
1155 unscaled = (srcW == dstW && srcH == dstH);
1157 c->srcRange |= handle_jpeg(&c->srcFormat);
1158 c->dstRange |= handle_jpeg(&c->dstFormat);
1160 if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
1161 av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
1163 if (!c->contrast && !c->saturation && !c->dstFormatBpp)
1164 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
1165 ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
1166 c->dstRange, 0, 1 << 16, 1 << 16);
1169 srcFormat = c->srcFormat;
1170 dstFormat = c->dstFormat;
1171 desc_src = av_pix_fmt_desc_get(srcFormat);
1172 desc_dst = av_pix_fmt_desc_get(dstFormat);
1174 // If the source has no alpha then disable alpha blendaway
1176 c->alphablend = SWS_ALPHA_BLEND_NONE;
1178 if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
1179 av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
1180 if (!sws_isSupportedInput(srcFormat)) {
1181 av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
1182 av_get_pix_fmt_name(srcFormat));
1183 return AVERROR(EINVAL);
1185 if (!sws_isSupportedOutput(dstFormat)) {
1186 av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
1187 av_get_pix_fmt_name(dstFormat));
1188 return AVERROR(EINVAL);
1191 av_assert2(desc_src && desc_dst);
1193 i = flags & (SWS_POINT |
1205 /* provide a default scaler if not set by caller */
1207 if (dstW < srcW && dstH < srcH)
1208 flags |= SWS_BICUBIC;
1209 else if (dstW > srcW && dstH > srcH)
1210 flags |= SWS_BICUBIC;
1212 flags |= SWS_BICUBIC;
1214 } else if (i & (i - 1)) {
1215 av_log(c, AV_LOG_ERROR,
1216 "Exactly one scaler algorithm must be chosen, got %X\n", i);
1217 return AVERROR(EINVAL);
1220 if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
1221 /* FIXME check if these are enough and try to lower them after
1222 * fixing the relevant parts of the code */
1223 av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
1224 srcW, srcH, dstW, dstH);
1225 return AVERROR(EINVAL);
1227 if (flags & SWS_FAST_BILINEAR) {
1228 if (srcW < 8 || dstW < 8) {
1229 flags ^= SWS_FAST_BILINEAR | SWS_BILINEAR;
1235 dstFilter = &dummyFilter;
1237 srcFilter = &dummyFilter;
1239 c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
1240 c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
1241 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
1242 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
1243 c->vRounder = 4 * 0x0001000100010001ULL;
1245 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
1246 (srcFilter->chrV && srcFilter->chrV->length > 1) ||
1247 (dstFilter->lumV && dstFilter->lumV->length > 1) ||
1248 (dstFilter->chrV && dstFilter->chrV->length > 1);
1249 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
1250 (srcFilter->chrH && srcFilter->chrH->length > 1) ||
1251 (dstFilter->lumH && dstFilter->lumH->length > 1) ||
1252 (dstFilter->chrH && dstFilter->chrH->length > 1);
1254 av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
1255 av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);
1257 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
1259 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
1260 flags |= SWS_FULL_CHR_H_INT;
1264 if ( c->chrSrcHSubSample == 0
1265 && c->chrSrcVSubSample == 0
1266 && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
1267 && !(c->flags & SWS_FAST_BILINEAR)
1269 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
1270 flags |= SWS_FULL_CHR_H_INT;
1275 if (c->dither == SWS_DITHER_AUTO) {
1276 if (flags & SWS_ERROR_DIFFUSION)
1277 c->dither = SWS_DITHER_ED;
1280 if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1281 dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1282 dstFormat == AV_PIX_FMT_BGR8 ||
1283 dstFormat == AV_PIX_FMT_RGB8) {
1284 if (c->dither == SWS_DITHER_AUTO)
1285 c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
1286 if (!(flags & SWS_FULL_CHR_H_INT)) {
1287 if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER) {
1288 av_log(c, AV_LOG_DEBUG,
1289 "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
1290 av_get_pix_fmt_name(dstFormat));
1291 flags |= SWS_FULL_CHR_H_INT;
1295 if (flags & SWS_FULL_CHR_H_INT) {
1296 if (c->dither == SWS_DITHER_BAYER) {
1297 av_log(c, AV_LOG_DEBUG,
1298 "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1299 av_get_pix_fmt_name(dstFormat));
1300 c->dither = SWS_DITHER_ED;
1304 if (isPlanarRGB(dstFormat)) {
1305 if (!(flags & SWS_FULL_CHR_H_INT)) {
1306 av_log(c, AV_LOG_DEBUG,
1307 "%s output is not supported with half chroma resolution, switching to full\n",
1308 av_get_pix_fmt_name(dstFormat));
1309 flags |= SWS_FULL_CHR_H_INT;
1314 /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1315 * chroma interpolation */
1316 if (flags & SWS_FULL_CHR_H_INT &&
1317 isAnyRGB(dstFormat) &&
1318 !isPlanarRGB(dstFormat) &&
1319 dstFormat != AV_PIX_FMT_RGBA64LE &&
1320 dstFormat != AV_PIX_FMT_RGBA64BE &&
1321 dstFormat != AV_PIX_FMT_BGRA64LE &&
1322 dstFormat != AV_PIX_FMT_BGRA64BE &&
1323 dstFormat != AV_PIX_FMT_RGB48LE &&
1324 dstFormat != AV_PIX_FMT_RGB48BE &&
1325 dstFormat != AV_PIX_FMT_BGR48LE &&
1326 dstFormat != AV_PIX_FMT_BGR48BE &&
1327 dstFormat != AV_PIX_FMT_RGBA &&
1328 dstFormat != AV_PIX_FMT_ARGB &&
1329 dstFormat != AV_PIX_FMT_BGRA &&
1330 dstFormat != AV_PIX_FMT_ABGR &&
1331 dstFormat != AV_PIX_FMT_RGB24 &&
1332 dstFormat != AV_PIX_FMT_BGR24 &&
1333 dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1334 dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1335 dstFormat != AV_PIX_FMT_BGR8 &&
1336 dstFormat != AV_PIX_FMT_RGB8
1338 av_log(c, AV_LOG_WARNING,
1339 "full chroma interpolation for destination format '%s' not yet implemented\n",
1340 av_get_pix_fmt_name(dstFormat));
1341 flags &= ~SWS_FULL_CHR_H_INT;
1344 if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1345 c->chrDstHSubSample = 1;
1347 // drop some chroma lines if the user wants it
1348 c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1349 SWS_SRC_V_CHR_DROP_SHIFT;
1350 c->chrSrcVSubSample += c->vChrDrop;
1352 /* drop every other pixel for chroma calculation unless user
1353 * wants full chroma */
1354 if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) &&
1355 srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1356 srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1357 srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1358 srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
1359 srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
1360 srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
1361 srcFormat != AV_PIX_FMT_GBRAP12BE && srcFormat != AV_PIX_FMT_GBRAP12LE &&
1362 srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE &&
1363 srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
1364 srcFormat != AV_PIX_FMT_GBRAP16BE && srcFormat != AV_PIX_FMT_GBRAP16LE &&
1365 ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1366 (flags & SWS_FAST_BILINEAR)))
1367 c->chrSrcHSubSample = 1;
1369 // Note the AV_CEIL_RSHIFT is so that we always round toward +inf.
1370 c->chrSrcW = AV_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
1371 c->chrSrcH = AV_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
1372 c->chrDstW = AV_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
1373 c->chrDstH = AV_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
1375 FF_ALLOCZ_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
1377 c->srcBpc = desc_src->comp[0].depth;
1380 c->dstBpc = desc_dst->comp[0].depth;
1383 if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1385 if (c->dstBpc == 16)
1388 if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1389 c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1390 c->chrDstW >= c->chrSrcW &&
1392 if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1394 && (flags & SWS_FAST_BILINEAR)) {
1395 if (flags & SWS_PRINT_INFO)
1396 av_log(c, AV_LOG_INFO,
1397 "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1399 if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1400 c->canMMXEXTBeUsed = 0;
1402 c->canMMXEXTBeUsed = 0;
1404 c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1405 c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1407 /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1408 * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1410 * n-2 is the last chrominance sample available.
1411 * This is not perfect, but no one should notice the difference, the more
1412 * correct variant would be like the vertical one, but that would require
1413 * some special code for the first and last pixel */
1414 if (flags & SWS_FAST_BILINEAR) {
1415 if (c->canMMXEXTBeUsed) {
1419 // we don't use the x86 asm scaler if MMX is available
1420 else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1421 c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
1422 c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1426 // hardcoded for now
1427 c->gamma_value = 2.2;
1428 tmpFmt = AV_PIX_FMT_RGBA64LE;
1431 if (!unscaled && c->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
1433 c->cascaded_context[0] = NULL;
1435 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1436 srcW, srcH, tmpFmt, 64);
1440 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1442 flags, NULL, NULL, c->param);
1443 if (!c->cascaded_context[0]) {
1447 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
1449 flags, srcFilter, dstFilter, c->param);
1451 if (!c->cascaded_context[1])
1454 c2 = c->cascaded_context[1];
1455 c2->is_internal_gamma = 1;
1456 c2->gamma = alloc_gamma_tbl( c->gamma_value);
1457 c2->inv_gamma = alloc_gamma_tbl(1.f/c->gamma_value);
1458 if (!c2->gamma || !c2->inv_gamma)
1459 return AVERROR(ENOMEM);
1461 // is_internal_flag is set after creating the context
1462 // to properly create the gamma convert FilterDescriptor
1463 // we have to re-initialize it
1464 ff_free_filters(c2);
1465 if (ff_init_filters(c2) < 0) {
1466 sws_freeContext(c2);
1470 c->cascaded_context[2] = NULL;
1471 if (dstFormat != tmpFmt) {
1472 ret = av_image_alloc(c->cascaded1_tmp, c->cascaded1_tmpStride,
1473 dstW, dstH, tmpFmt, 64);
1477 c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
1478 dstW, dstH, dstFormat,
1479 flags, NULL, NULL, c->param);
1480 if (!c->cascaded_context[2])
1486 if (isBayer(srcFormat)) {
1488 (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P)) {
1489 enum AVPixelFormat tmpFormat = AV_PIX_FMT_RGB24;
1491 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1492 srcW, srcH, tmpFormat, 64);
1496 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1497 srcW, srcH, tmpFormat,
1498 flags, srcFilter, NULL, c->param);
1499 if (!c->cascaded_context[0])
1502 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
1503 dstW, dstH, dstFormat,
1504 flags, NULL, dstFilter, c->param);
1505 if (!c->cascaded_context[1])
1511 if (CONFIG_SWSCALE_ALPHA && isALPHA(srcFormat) && !isALPHA(dstFormat)) {
1512 enum AVPixelFormat tmpFormat = alphaless_fmt(srcFormat);
1514 if (tmpFormat != AV_PIX_FMT_NONE && c->alphablend != SWS_ALPHA_BLEND_NONE)
1516 dstFormat != tmpFormat ||
1517 usesHFilter || usesVFilter ||
1518 c->srcRange != c->dstRange
1520 c->cascaded_mainindex = 1;
1521 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1522 srcW, srcH, tmpFormat, 64);
1526 c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, srcFormat,
1527 srcW, srcH, tmpFormat,
1529 if (!c->cascaded_context[0])
1531 c->cascaded_context[0]->alphablend = c->alphablend;
1532 ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
1536 c->cascaded_context[1] = sws_alloc_set_opts(srcW, srcH, tmpFormat,
1537 dstW, dstH, dstFormat,
1539 if (!c->cascaded_context[1])
1542 c->cascaded_context[1]->srcRange = c->srcRange;
1543 c->cascaded_context[1]->dstRange = c->dstRange;
1544 ret = sws_init_context(c->cascaded_context[1], srcFilter , dstFilter);
1552 #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
1554 /* precalculate horizontal scaler filter coefficients */
1556 #if HAVE_MMXEXT_INLINE
1557 // can't downscale !!!
1558 if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1559 c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1561 c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1562 NULL, NULL, NULL, 4);
1565 c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
1566 PROT_READ | PROT_WRITE,
1567 MAP_PRIVATE | MAP_ANONYMOUS,
1569 c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
1570 PROT_READ | PROT_WRITE,
1571 MAP_PRIVATE | MAP_ANONYMOUS,
1573 #elif HAVE_VIRTUALALLOC
1574 c->lumMmxextFilterCode = VirtualAlloc(NULL,
1575 c->lumMmxextFilterCodeSize,
1577 PAGE_EXECUTE_READWRITE);
1578 c->chrMmxextFilterCode = VirtualAlloc(NULL,
1579 c->chrMmxextFilterCodeSize,
1581 PAGE_EXECUTE_READWRITE);
1583 c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
1584 c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
1587 #ifdef MAP_ANONYMOUS
1588 if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1590 if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
1593 av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1594 return AVERROR(ENOMEM);
1597 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
1598 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
1599 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
1600 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);
1602 ff_init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
1603 c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1604 ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1605 c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1608 if ( mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
1609 || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
1610 av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
1615 #endif /* HAVE_MMXEXT_INLINE */
1617 const int filterAlign = X86_MMX(cpu_flags) ? 4 :
1618 PPC_ALTIVEC(cpu_flags) ? 8 :
1619 have_neon(cpu_flags) ? 4 : 1;
1621 if ((ret = initFilter(&c->hLumFilter, &c->hLumFilterPos,
1622 &c->hLumFilterSize, c->lumXInc,
1623 srcW, dstW, filterAlign, 1 << 14,
1624 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1625 cpu_flags, srcFilter->lumH, dstFilter->lumH,
1627 get_local_pos(c, 0, 0, 0),
1628 get_local_pos(c, 0, 0, 0))) < 0)
1630 if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
1631 &c->hChrFilterSize, c->chrXInc,
1632 c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1633 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1634 cpu_flags, srcFilter->chrH, dstFilter->chrH,
1636 get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
1637 get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
1640 } // initialize horizontal stuff
1642 /* precalculate vertical scaler filter coefficients */
1644 const int filterAlign = X86_MMX(cpu_flags) ? 2 :
1645 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1647 if ((ret = initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
1648 c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1649 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1650 cpu_flags, srcFilter->lumV, dstFilter->lumV,
1652 get_local_pos(c, 0, 0, 1),
1653 get_local_pos(c, 0, 0, 1))) < 0)
1655 if ((ret = initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
1656 c->chrYInc, c->chrSrcH, c->chrDstH,
1657 filterAlign, (1 << 12),
1658 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1659 cpu_flags, srcFilter->chrV, dstFilter->chrV,
1661 get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
1662 get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1))) < 0)
1667 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
1668 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);
1670 for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1672 short *p = (short *)&c->vYCoeffsBank[i];
1673 for (j = 0; j < 8; j++)
1674 p[j] = c->vLumFilter[i];
1677 for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1679 short *p = (short *)&c->vCCoeffsBank[i];
1680 for (j = 0; j < 8; j++)
1681 p[j] = c->vChrFilter[i];
1686 // calculate buffer sizes so that they won't run out while handling these damn slices
1687 c->vLumBufSize = c->vLumFilterSize;
1688 c->vChrBufSize = c->vChrFilterSize;
1689 for (i = 0; i < dstH; i++) {
1690 int chrI = (int64_t)i * c->chrDstH / dstH;
1691 int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
1692 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
1693 << c->chrSrcVSubSample));
1695 nextSlice >>= c->chrSrcVSubSample;
1696 nextSlice <<= c->chrSrcVSubSample;
1697 if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
1698 c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
1699 if (c->vChrFilterPos[chrI] + c->vChrBufSize <
1700 (nextSlice >> c->chrSrcVSubSample))
1701 c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
1702 c->vChrFilterPos[chrI];
1705 for (i = 0; i < 4; i++)
1706 FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);
1708 /* Allocate pixbufs (we use dynamic allocation because otherwise we would
1709 * need to allocate several megabytes to handle all possible cases) */
1710 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1711 FF_ALLOCZ_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1712 FF_ALLOCZ_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1713 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
1714 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1715 /* Note we need at least one pixel more at the end because of the MMX code
1716 * (just in case someone wants to replace the 4000/8000). */
1717 /* align at 16 bytes for AltiVec */
1718 for (i = 0; i < c->vLumBufSize; i++) {
1719 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
1720 dst_stride + 16, fail);
1721 c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
1723 // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1724 c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1725 c->uv_offx2 = dst_stride + 16;
1726 for (i = 0; i < c->vChrBufSize; i++) {
1727 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
1728 dst_stride * 2 + 32, fail);
1729 c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
1730 c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
1731 = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
1733 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1734 for (i = 0; i < c->vLumBufSize; i++) {
1735 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
1736 dst_stride + 16, fail);
1737 c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
1740 // try to avoid drawing green stuff between the right end and the stride end
1741 for (i = 0; i < c->vChrBufSize; i++)
1742 if(desc_dst->comp[0].depth == 16){
1743 av_assert0(c->dstBpc > 14);
1744 for(j=0; j<dst_stride/2+1; j++)
1745 ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
1747 for(j=0; j<dst_stride+1; j++)
1748 ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
1750 av_assert0(c->chrDstH <= dstH);
1752 if (flags & SWS_PRINT_INFO) {
1753 const char *scaler = NULL, *cpucaps;
1755 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
1756 if (flags & scale_algorithms[i].flag) {
1757 scaler = scale_algorithms[i].description;
1762 scaler = "ehh flags invalid?!";
1763 av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
1765 av_get_pix_fmt_name(srcFormat),
1767 dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
1768 dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1769 dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1774 av_get_pix_fmt_name(dstFormat));
1776 if (INLINE_MMXEXT(cpu_flags))
1778 else if (INLINE_AMD3DNOW(cpu_flags))
1780 else if (INLINE_MMX(cpu_flags))
1782 else if (PPC_ALTIVEC(cpu_flags))
1783 cpucaps = "AltiVec";
1787 av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);
1789 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1790 av_log(c, AV_LOG_DEBUG,
1791 "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1792 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1793 av_log(c, AV_LOG_DEBUG,
1794 "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1795 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1796 c->chrXInc, c->chrYInc);
1799 /* alpha blend special case, note this has been split via cascaded contexts if its scaled */
1800 if (unscaled && !usesHFilter && !usesVFilter &&
1801 c->alphablend != SWS_ALPHA_BLEND_NONE &&
1802 isALPHA(srcFormat) &&
1803 (c->srcRange == c->dstRange || isAnyRGB(dstFormat)) &&
1804 alphaless_fmt(srcFormat) == dstFormat
1806 c->swscale = ff_sws_alphablendaway;
1808 if (flags & SWS_PRINT_INFO)
1809 av_log(c, AV_LOG_INFO,
1810 "using alpha blendaway %s -> %s special converter\n",
1811 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1815 /* unscaled special cases */
1816 if (unscaled && !usesHFilter && !usesVFilter &&
1817 (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
1818 ff_get_unscaled_swscale(c);
1821 if (flags & SWS_PRINT_INFO)
1822 av_log(c, AV_LOG_INFO,
1823 "using unscaled %s -> %s special converter\n",
1824 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1829 c->swscale = ff_getSwsFunc(c);
1830 return ff_init_filters(c);
1831 fail: // FIXME replace things by appropriate error codes
1832 if (ret == RETCODE_USE_CASCADE) {
1833 int tmpW = sqrt(srcW * (int64_t)dstW);
1834 int tmpH = sqrt(srcH * (int64_t)dstH);
1835 enum AVPixelFormat tmpFormat = AV_PIX_FMT_YUV420P;
1837 if (isALPHA(srcFormat))
1838 tmpFormat = AV_PIX_FMT_YUVA420P;
1840 if (srcW*(int64_t)srcH <= 4LL*dstW*dstH)
1841 return AVERROR(EINVAL);
1843 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1844 tmpW, tmpH, tmpFormat, 64);
1848 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1849 tmpW, tmpH, tmpFormat,
1850 flags, srcFilter, NULL, c->param);
1851 if (!c->cascaded_context[0])
1854 c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
1855 dstW, dstH, dstFormat,
1856 flags, NULL, dstFilter, c->param);
1857 if (!c->cascaded_context[1])
1864 SwsContext *sws_alloc_set_opts(int srcW, int srcH, enum AVPixelFormat srcFormat,
1865 int dstW, int dstH, enum AVPixelFormat dstFormat,
1866 int flags, const double *param)
1870 if (!(c = sws_alloc_context()))
1878 c->srcFormat = srcFormat;
1879 c->dstFormat = dstFormat;
1882 c->param[0] = param[0];
1883 c->param[1] = param[1];
1889 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
1890 int dstW, int dstH, enum AVPixelFormat dstFormat,
1891 int flags, SwsFilter *srcFilter,
1892 SwsFilter *dstFilter, const double *param)
1896 c = sws_alloc_set_opts(srcW, srcH, srcFormat,
1897 dstW, dstH, dstFormat,
1902 if (sws_init_context(c, srcFilter, dstFilter) < 0) {
1910 static int isnan_vec(SwsVector *a)
1913 for (i=0; i<a->length; i++)
1914 if (isnan(a->coeff[i]))
1919 static void makenan_vec(SwsVector *a)
1922 for (i=0; i<a->length; i++)
1926 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
1927 float lumaSharpen, float chromaSharpen,
1928 float chromaHShift, float chromaVShift,
1931 SwsFilter *filter = av_malloc(sizeof(SwsFilter));
1935 if (lumaGBlur != 0.0) {
1936 filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
1937 filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
1939 filter->lumH = sws_getIdentityVec();
1940 filter->lumV = sws_getIdentityVec();
1943 if (chromaGBlur != 0.0) {
1944 filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
1945 filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
1947 filter->chrH = sws_getIdentityVec();
1948 filter->chrV = sws_getIdentityVec();
1951 if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
1954 if (chromaSharpen != 0.0) {
1955 SwsVector *id = sws_getIdentityVec();
1958 sws_scaleVec(filter->chrH, -chromaSharpen);
1959 sws_scaleVec(filter->chrV, -chromaSharpen);
1960 sws_addVec(filter->chrH, id);
1961 sws_addVec(filter->chrV, id);
1965 if (lumaSharpen != 0.0) {
1966 SwsVector *id = sws_getIdentityVec();
1969 sws_scaleVec(filter->lumH, -lumaSharpen);
1970 sws_scaleVec(filter->lumV, -lumaSharpen);
1971 sws_addVec(filter->lumH, id);
1972 sws_addVec(filter->lumV, id);
1976 if (chromaHShift != 0.0)
1977 sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
1979 if (chromaVShift != 0.0)
1980 sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
1982 sws_normalizeVec(filter->chrH, 1.0);
1983 sws_normalizeVec(filter->chrV, 1.0);
1984 sws_normalizeVec(filter->lumH, 1.0);
1985 sws_normalizeVec(filter->lumV, 1.0);
1987 if (isnan_vec(filter->chrH) ||
1988 isnan_vec(filter->chrV) ||
1989 isnan_vec(filter->lumH) ||
1990 isnan_vec(filter->lumV))
1994 sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
1996 sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
2001 sws_freeVec(filter->lumH);
2002 sws_freeVec(filter->lumV);
2003 sws_freeVec(filter->chrH);
2004 sws_freeVec(filter->chrV);
2009 SwsVector *sws_allocVec(int length)
2013 if(length <= 0 || length > INT_MAX/ sizeof(double))
2016 vec = av_malloc(sizeof(SwsVector));
2019 vec->length = length;
2020 vec->coeff = av_malloc(sizeof(double) * length);
2026 SwsVector *sws_getGaussianVec(double variance, double quality)
2028 const int length = (int)(variance * quality + 0.5) | 1;
2030 double middle = (length - 1) * 0.5;
2033 if(variance < 0 || quality < 0)
2036 vec = sws_allocVec(length);
2041 for (i = 0; i < length; i++) {
2042 double dist = i - middle;
2043 vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
2044 sqrt(2 * variance * M_PI);
2047 sws_normalizeVec(vec, 1.0);
2052 SwsVector *sws_getConstVec(double c, int length)
2055 SwsVector *vec = sws_allocVec(length);
2060 for (i = 0; i < length; i++)
2066 SwsVector *sws_getIdentityVec(void)
2068 return sws_getConstVec(1.0, 1);
2071 static double sws_dcVec(SwsVector *a)
2076 for (i = 0; i < a->length; i++)
2082 void sws_scaleVec(SwsVector *a, double scalar)
2086 for (i = 0; i < a->length; i++)
2087 a->coeff[i] *= scalar;
2090 void sws_normalizeVec(SwsVector *a, double height)
2092 sws_scaleVec(a, height / sws_dcVec(a));
2095 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
2097 int length = a->length + b->length - 1;
2099 SwsVector *vec = sws_getConstVec(0.0, length);
2104 for (i = 0; i < a->length; i++) {
2105 for (j = 0; j < b->length; j++) {
2106 vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
2113 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
2115 int length = FFMAX(a->length, b->length);
2117 SwsVector *vec = sws_getConstVec(0.0, length);
2122 for (i = 0; i < a->length; i++)
2123 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
2124 for (i = 0; i < b->length; i++)
2125 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
2130 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
2132 int length = FFMAX(a->length, b->length);
2134 SwsVector *vec = sws_getConstVec(0.0, length);
2139 for (i = 0; i < a->length; i++)
2140 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
2141 for (i = 0; i < b->length; i++)
2142 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];
2147 /* shift left / or right if "shift" is negative */
2148 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
2150 int length = a->length + FFABS(shift) * 2;
2152 SwsVector *vec = sws_getConstVec(0.0, length);
2157 for (i = 0; i < a->length; i++) {
2158 vec->coeff[i + (length - 1) / 2 -
2159 (a->length - 1) / 2 - shift] = a->coeff[i];
2165 void sws_shiftVec(SwsVector *a, int shift)
2167 SwsVector *shifted = sws_getShiftedVec(a, shift);
2173 a->coeff = shifted->coeff;
2174 a->length = shifted->length;
2178 void sws_addVec(SwsVector *a, SwsVector *b)
2180 SwsVector *sum = sws_sumVec(a, b);
2186 a->coeff = sum->coeff;
2187 a->length = sum->length;
2191 void sws_subVec(SwsVector *a, SwsVector *b)
2193 SwsVector *diff = sws_diffVec(a, b);
2199 a->coeff = diff->coeff;
2200 a->length = diff->length;
2204 void sws_convVec(SwsVector *a, SwsVector *b)
2206 SwsVector *conv = sws_getConvVec(a, b);
2212 a->coeff = conv->coeff;
2213 a->length = conv->length;
2217 SwsVector *sws_cloneVec(SwsVector *a)
2219 SwsVector *vec = sws_allocVec(a->length);
2224 memcpy(vec->coeff, a->coeff, a->length * sizeof(*a->coeff));
2229 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
2236 for (i = 0; i < a->length; i++)
2237 if (a->coeff[i] > max)
2240 for (i = 0; i < a->length; i++)
2241 if (a->coeff[i] < min)
2246 for (i = 0; i < a->length; i++) {
2247 int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
2248 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
2250 av_log(log_ctx, log_level, " ");
2251 av_log(log_ctx, log_level, "|\n");
2255 void sws_freeVec(SwsVector *a)
2259 av_freep(&a->coeff);
2264 void sws_freeFilter(SwsFilter *filter)
2269 sws_freeVec(filter->lumH);
2270 sws_freeVec(filter->lumV);
2271 sws_freeVec(filter->chrH);
2272 sws_freeVec(filter->chrV);
2276 void sws_freeContext(SwsContext *c)
2283 for (i = 0; i < c->vLumBufSize; i++)
2284 av_freep(&c->lumPixBuf[i]);
2285 av_freep(&c->lumPixBuf);
2288 if (c->chrUPixBuf) {
2289 for (i = 0; i < c->vChrBufSize; i++)
2290 av_freep(&c->chrUPixBuf[i]);
2291 av_freep(&c->chrUPixBuf);
2292 av_freep(&c->chrVPixBuf);
2295 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
2296 for (i = 0; i < c->vLumBufSize; i++)
2297 av_freep(&c->alpPixBuf[i]);
2298 av_freep(&c->alpPixBuf);
2301 for (i = 0; i < 4; i++)
2302 av_freep(&c->dither_error[i]);
2304 av_freep(&c->vLumFilter);
2305 av_freep(&c->vChrFilter);
2306 av_freep(&c->hLumFilter);
2307 av_freep(&c->hChrFilter);
2309 av_freep(&c->vYCoeffsBank);
2310 av_freep(&c->vCCoeffsBank);
2313 av_freep(&c->vLumFilterPos);
2314 av_freep(&c->vChrFilterPos);
2315 av_freep(&c->hLumFilterPos);
2316 av_freep(&c->hChrFilterPos);
2320 if (c->lumMmxextFilterCode)
2321 munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
2322 if (c->chrMmxextFilterCode)
2323 munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize);
2324 #elif HAVE_VIRTUALALLOC
2325 if (c->lumMmxextFilterCode)
2326 VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
2327 if (c->chrMmxextFilterCode)
2328 VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
2330 av_free(c->lumMmxextFilterCode);
2331 av_free(c->chrMmxextFilterCode);
2333 c->lumMmxextFilterCode = NULL;
2334 c->chrMmxextFilterCode = NULL;
2335 #endif /* HAVE_MMX_INLINE */
2337 av_freep(&c->yuvTable);
2338 av_freep(&c->formatConvBuffer);
2340 sws_freeContext(c->cascaded_context[0]);
2341 sws_freeContext(c->cascaded_context[1]);
2342 sws_freeContext(c->cascaded_context[2]);
2343 memset(c->cascaded_context, 0, sizeof(c->cascaded_context));
2344 av_freep(&c->cascaded_tmp[0]);
2345 av_freep(&c->cascaded1_tmp[0]);
2347 av_freep(&c->gamma);
2348 av_freep(&c->inv_gamma);
2355 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
2356 int srcH, enum AVPixelFormat srcFormat,
2358 enum AVPixelFormat dstFormat, int flags,
2359 SwsFilter *srcFilter,
2360 SwsFilter *dstFilter,
2361 const double *param)
2363 static const double default_param[2] = { SWS_PARAM_DEFAULT,
2364 SWS_PARAM_DEFAULT };
2365 int64_t src_h_chr_pos = -513, dst_h_chr_pos = -513,
2366 src_v_chr_pos = -513, dst_v_chr_pos = -513;
2369 param = default_param;
2372 (context->srcW != srcW ||
2373 context->srcH != srcH ||
2374 context->srcFormat != srcFormat ||
2375 context->dstW != dstW ||
2376 context->dstH != dstH ||
2377 context->dstFormat != dstFormat ||
2378 context->flags != flags ||
2379 context->param[0] != param[0] ||
2380 context->param[1] != param[1])) {
2382 av_opt_get_int(context, "src_h_chr_pos", 0, &src_h_chr_pos);
2383 av_opt_get_int(context, "src_v_chr_pos", 0, &src_v_chr_pos);
2384 av_opt_get_int(context, "dst_h_chr_pos", 0, &dst_h_chr_pos);
2385 av_opt_get_int(context, "dst_v_chr_pos", 0, &dst_v_chr_pos);
2386 sws_freeContext(context);
2391 if (!(context = sws_alloc_context()))
2393 context->srcW = srcW;
2394 context->srcH = srcH;
2395 context->srcFormat = srcFormat;
2396 context->dstW = dstW;
2397 context->dstH = dstH;
2398 context->dstFormat = dstFormat;
2399 context->flags = flags;
2400 context->param[0] = param[0];
2401 context->param[1] = param[1];
2403 av_opt_set_int(context, "src_h_chr_pos", src_h_chr_pos, 0);
2404 av_opt_set_int(context, "src_v_chr_pos", src_v_chr_pos, 0);
2405 av_opt_set_int(context, "dst_h_chr_pos", dst_h_chr_pos, 0);
2406 av_opt_set_int(context, "dst_v_chr_pos", dst_v_chr_pos, 0);
2408 if (sws_init_context(context, srcFilter, dstFilter) < 0) {
2409 sws_freeContext(context);