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 _SVID_SOURCE // needed for MAP_ANONYMOUS
24 #define _DARWIN_C_SOURCE // needed for MAP_ANON
31 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
32 #define MAP_ANONYMOUS MAP_ANON
36 #define WIN32_LEAN_AND_MEAN
40 #include "libavutil/attributes.h"
41 #include "libavutil/avassert.h"
42 #include "libavutil/avutil.h"
43 #include "libavutil/bswap.h"
44 #include "libavutil/cpu.h"
45 #include "libavutil/imgutils.h"
46 #include "libavutil/intreadwrite.h"
47 #include "libavutil/mathematics.h"
48 #include "libavutil/opt.h"
49 #include "libavutil/pixdesc.h"
50 #include "libavutil/ppc/cpu.h"
51 #include "libavutil/x86/asm.h"
52 #include "libavutil/x86/cpu.h"
55 #include "swscale_internal.h"
57 static void handle_formats(SwsContext *c);
59 unsigned swscale_version(void)
61 av_assert0(LIBSWSCALE_VERSION_MICRO >= 100);
62 return LIBSWSCALE_VERSION_INT;
65 const char *swscale_configuration(void)
67 return FFMPEG_CONFIGURATION;
70 const char *swscale_license(void)
72 #define LICENSE_PREFIX "libswscale license: "
73 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
76 typedef struct FormatEntry {
77 uint8_t is_supported_in :1;
78 uint8_t is_supported_out :1;
79 uint8_t is_supported_endianness :1;
82 static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
83 [AV_PIX_FMT_YUV420P] = { 1, 1 },
84 [AV_PIX_FMT_YUYV422] = { 1, 1 },
85 [AV_PIX_FMT_RGB24] = { 1, 1 },
86 [AV_PIX_FMT_BGR24] = { 1, 1 },
87 [AV_PIX_FMT_YUV422P] = { 1, 1 },
88 [AV_PIX_FMT_YUV444P] = { 1, 1 },
89 [AV_PIX_FMT_YUV410P] = { 1, 1 },
90 [AV_PIX_FMT_YUV411P] = { 1, 1 },
91 [AV_PIX_FMT_GRAY8] = { 1, 1 },
92 [AV_PIX_FMT_MONOWHITE] = { 1, 1 },
93 [AV_PIX_FMT_MONOBLACK] = { 1, 1 },
94 [AV_PIX_FMT_PAL8] = { 1, 0 },
95 [AV_PIX_FMT_YUVJ420P] = { 1, 1 },
96 [AV_PIX_FMT_YUVJ411P] = { 1, 1 },
97 [AV_PIX_FMT_YUVJ422P] = { 1, 1 },
98 [AV_PIX_FMT_YUVJ444P] = { 1, 1 },
99 [AV_PIX_FMT_YVYU422] = { 1, 1 },
100 [AV_PIX_FMT_UYVY422] = { 1, 1 },
101 [AV_PIX_FMT_UYYVYY411] = { 0, 0 },
102 [AV_PIX_FMT_BGR8] = { 1, 1 },
103 [AV_PIX_FMT_BGR4] = { 0, 1 },
104 [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
105 [AV_PIX_FMT_RGB8] = { 1, 1 },
106 [AV_PIX_FMT_RGB4] = { 0, 1 },
107 [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
108 [AV_PIX_FMT_NV12] = { 1, 1 },
109 [AV_PIX_FMT_NV21] = { 1, 1 },
110 [AV_PIX_FMT_ARGB] = { 1, 1 },
111 [AV_PIX_FMT_RGBA] = { 1, 1 },
112 [AV_PIX_FMT_ABGR] = { 1, 1 },
113 [AV_PIX_FMT_BGRA] = { 1, 1 },
114 [AV_PIX_FMT_0RGB] = { 1, 1 },
115 [AV_PIX_FMT_RGB0] = { 1, 1 },
116 [AV_PIX_FMT_0BGR] = { 1, 1 },
117 [AV_PIX_FMT_BGR0] = { 1, 1 },
118 [AV_PIX_FMT_GRAY16BE] = { 1, 1 },
119 [AV_PIX_FMT_GRAY16LE] = { 1, 1 },
120 [AV_PIX_FMT_YUV440P] = { 1, 1 },
121 [AV_PIX_FMT_YUVJ440P] = { 1, 1 },
122 [AV_PIX_FMT_YUV440P10LE] = { 1, 1 },
123 [AV_PIX_FMT_YUV440P10BE] = { 1, 1 },
124 [AV_PIX_FMT_YUV440P12LE] = { 1, 1 },
125 [AV_PIX_FMT_YUV440P12BE] = { 1, 1 },
126 [AV_PIX_FMT_YUVA420P] = { 1, 1 },
127 [AV_PIX_FMT_YUVA422P] = { 1, 1 },
128 [AV_PIX_FMT_YUVA444P] = { 1, 1 },
129 [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
130 [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
131 [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
132 [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
133 [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
134 [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
135 [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
136 [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
137 [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
138 [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
139 [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
140 [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
141 [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
142 [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
143 [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
144 [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
145 [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
146 [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
147 [AV_PIX_FMT_RGB48BE] = { 1, 1 },
148 [AV_PIX_FMT_RGB48LE] = { 1, 1 },
149 [AV_PIX_FMT_RGBA64BE] = { 1, 1, 1 },
150 [AV_PIX_FMT_RGBA64LE] = { 1, 1, 1 },
151 [AV_PIX_FMT_RGB565BE] = { 1, 1 },
152 [AV_PIX_FMT_RGB565LE] = { 1, 1 },
153 [AV_PIX_FMT_RGB555BE] = { 1, 1 },
154 [AV_PIX_FMT_RGB555LE] = { 1, 1 },
155 [AV_PIX_FMT_BGR565BE] = { 1, 1 },
156 [AV_PIX_FMT_BGR565LE] = { 1, 1 },
157 [AV_PIX_FMT_BGR555BE] = { 1, 1 },
158 [AV_PIX_FMT_BGR555LE] = { 1, 1 },
159 [AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
160 [AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
161 [AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
162 [AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
163 [AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
164 [AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
165 [AV_PIX_FMT_RGB444LE] = { 1, 1 },
166 [AV_PIX_FMT_RGB444BE] = { 1, 1 },
167 [AV_PIX_FMT_BGR444LE] = { 1, 1 },
168 [AV_PIX_FMT_BGR444BE] = { 1, 1 },
169 [AV_PIX_FMT_YA8] = { 1, 0 },
170 [AV_PIX_FMT_YA16BE] = { 1, 0 },
171 [AV_PIX_FMT_YA16LE] = { 1, 0 },
172 [AV_PIX_FMT_BGR48BE] = { 1, 1 },
173 [AV_PIX_FMT_BGR48LE] = { 1, 1 },
174 [AV_PIX_FMT_BGRA64BE] = { 1, 1, 1 },
175 [AV_PIX_FMT_BGRA64LE] = { 1, 1, 1 },
176 [AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
177 [AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
178 [AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
179 [AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
180 [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
181 [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
182 [AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
183 [AV_PIX_FMT_YUV420P14LE] = { 1, 1 },
184 [AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
185 [AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
186 [AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
187 [AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
188 [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
189 [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
190 [AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
191 [AV_PIX_FMT_YUV422P14LE] = { 1, 1 },
192 [AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
193 [AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
194 [AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
195 [AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
196 [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
197 [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
198 [AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
199 [AV_PIX_FMT_YUV444P14LE] = { 1, 1 },
200 [AV_PIX_FMT_GBRP] = { 1, 1 },
201 [AV_PIX_FMT_GBRP9LE] = { 1, 1 },
202 [AV_PIX_FMT_GBRP9BE] = { 1, 1 },
203 [AV_PIX_FMT_GBRP10LE] = { 1, 1 },
204 [AV_PIX_FMT_GBRP10BE] = { 1, 1 },
205 [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
206 [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
207 [AV_PIX_FMT_GBRP14LE] = { 1, 1 },
208 [AV_PIX_FMT_GBRP14BE] = { 1, 1 },
209 [AV_PIX_FMT_GBRP16LE] = { 1, 0 },
210 [AV_PIX_FMT_GBRP16BE] = { 1, 0 },
211 [AV_PIX_FMT_GBRAP] = { 1, 1 },
212 [AV_PIX_FMT_GBRAP16LE] = { 1, 0 },
213 [AV_PIX_FMT_GBRAP16BE] = { 1, 0 },
214 [AV_PIX_FMT_BAYER_BGGR8] = { 1, 0 },
215 [AV_PIX_FMT_BAYER_RGGB8] = { 1, 0 },
216 [AV_PIX_FMT_BAYER_GBRG8] = { 1, 0 },
217 [AV_PIX_FMT_BAYER_GRBG8] = { 1, 0 },
218 [AV_PIX_FMT_BAYER_BGGR16LE] = { 1, 0 },
219 [AV_PIX_FMT_BAYER_BGGR16BE] = { 1, 0 },
220 [AV_PIX_FMT_BAYER_RGGB16LE] = { 1, 0 },
221 [AV_PIX_FMT_BAYER_RGGB16BE] = { 1, 0 },
222 [AV_PIX_FMT_BAYER_GBRG16LE] = { 1, 0 },
223 [AV_PIX_FMT_BAYER_GBRG16BE] = { 1, 0 },
224 [AV_PIX_FMT_BAYER_GRBG16LE] = { 1, 0 },
225 [AV_PIX_FMT_BAYER_GRBG16BE] = { 1, 0 },
226 [AV_PIX_FMT_XYZ12BE] = { 1, 1, 1 },
227 [AV_PIX_FMT_XYZ12LE] = { 1, 1, 1 },
230 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
232 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
233 format_entries[pix_fmt].is_supported_in : 0;
236 int sws_isSupportedOutput(enum AVPixelFormat pix_fmt)
238 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
239 format_entries[pix_fmt].is_supported_out : 0;
242 int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
244 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
245 format_entries[pix_fmt].is_supported_endianness : 0;
248 static double getSplineCoeff(double a, double b, double c, double d,
252 return ((d * dist + c) * dist + b) * dist + a;
254 return getSplineCoeff(0.0,
255 b + 2.0 * c + 3.0 * d,
257 -b - 3.0 * c - 6.0 * d,
261 static av_cold int get_local_pos(SwsContext *s, int chr_subsample, int pos, int dir)
263 if (pos == -1 || pos <= -513) {
264 pos = (128 << chr_subsample) - 128;
266 pos += 128; // relative to ideal left edge
267 return pos >> chr_subsample;
271 int flag; ///< flag associated to the algorithm
272 const char *description; ///< human-readable description
273 int size_factor; ///< size factor used when initing the filters
276 static const ScaleAlgorithm scale_algorithms[] = {
277 { SWS_AREA, "area averaging", 1 /* downscale only, for upscale it is bilinear */ },
278 { SWS_BICUBIC, "bicubic", 4 },
279 { SWS_BICUBLIN, "luma bicubic / chroma bilinear", -1 },
280 { SWS_BILINEAR, "bilinear", 2 },
281 { SWS_FAST_BILINEAR, "fast bilinear", -1 },
282 { SWS_GAUSS, "Gaussian", 8 /* infinite ;) */ },
283 { SWS_LANCZOS, "Lanczos", -1 /* custom */ },
284 { SWS_POINT, "nearest neighbor / point", -1 },
285 { SWS_SINC, "sinc", 20 /* infinite ;) */ },
286 { SWS_SPLINE, "bicubic spline", 20 /* infinite :)*/ },
287 { SWS_X, "experimental", 8 },
290 static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
291 int *outFilterSize, int xInc, int srcW,
292 int dstW, int filterAlign, int one,
293 int flags, int cpu_flags,
294 SwsVector *srcFilter, SwsVector *dstFilter,
295 double param[2], int srcPos, int dstPos)
301 int64_t *filter = NULL;
302 int64_t *filter2 = NULL;
303 const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
306 emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
308 // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
309 FF_ALLOC_ARRAY_OR_GOTO(NULL, *filterPos, (dstW + 3), sizeof(**filterPos), fail);
311 if (FFABS(xInc - 0x10000) < 10 && srcPos == dstPos) { // unscaled
314 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter,
315 dstW, sizeof(*filter) * filterSize, fail);
317 for (i = 0; i < dstW; i++) {
318 filter[i * filterSize] = fone;
321 } else if (flags & SWS_POINT) { // lame looking point sampling mode
325 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
326 dstW, sizeof(*filter) * filterSize, fail);
328 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
329 for (i = 0; i < dstW; i++) {
330 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
332 (*filterPos)[i] = xx;
336 } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
337 (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
341 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
342 dstW, sizeof(*filter) * filterSize, fail);
344 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
345 for (i = 0; i < dstW; i++) {
346 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
349 (*filterPos)[i] = xx;
350 // bilinear upscale / linear interpolate / area averaging
351 for (j = 0; j < filterSize; j++) {
352 int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
355 filter[i * filterSize + j] = coeff;
364 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
365 if (flags & scale_algorithms[i].flag && scale_algorithms[i].size_factor > 0) {
366 sizeFactor = scale_algorithms[i].size_factor;
370 if (flags & SWS_LANCZOS)
371 sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
372 av_assert0(sizeFactor > 0);
375 filterSize = 1 + sizeFactor; // upscale
377 filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
379 filterSize = FFMIN(filterSize, srcW - 2);
380 filterSize = FFMAX(filterSize, 1);
382 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
383 dstW, sizeof(*filter) * filterSize, fail);
385 xDstInSrc = ((dstPos*(int64_t)xInc)>>7) - ((srcPos*0x10000LL)>>7);
386 for (i = 0; i < dstW; i++) {
387 int xx = (xDstInSrc - ((int64_t)(filterSize - 2) << 16)) / (1 << 17);
389 (*filterPos)[i] = xx;
390 for (j = 0; j < filterSize; j++) {
391 int64_t d = (FFABS(((int64_t)xx * (1 << 17)) - xDstInSrc)) << 13;
397 floatd = d * (1.0 / (1 << 30));
399 if (flags & SWS_BICUBIC) {
400 int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
401 int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
403 if (d >= 1LL << 31) {
406 int64_t dd = (d * d) >> 30;
407 int64_t ddd = (dd * d) >> 30;
410 coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
411 (-18 * (1 << 24) + 12 * B + 6 * C) * dd +
412 (6 * (1 << 24) - 2 * B) * (1 << 30);
414 coeff = (-B - 6 * C) * ddd +
415 (6 * B + 30 * C) * dd +
416 (-12 * B - 48 * C) * d +
417 (8 * B + 24 * C) * (1 << 30);
419 coeff /= (1LL<<54)/fone;
422 else if (flags & SWS_X) {
423 double p = param ? param * 0.01 : 0.3;
424 coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
425 coeff *= pow(2.0, -p * d * d);
428 else if (flags & SWS_X) {
429 double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
433 c = cos(floatd * M_PI);
440 coeff = (c * 0.5 + 0.5) * fone;
441 } else if (flags & SWS_AREA) {
442 int64_t d2 = d - (1 << 29);
443 if (d2 * xInc < -(1LL << (29 + 16)))
444 coeff = 1.0 * (1LL << (30 + 16));
445 else if (d2 * xInc < (1LL << (29 + 16)))
446 coeff = -d2 * xInc + (1LL << (29 + 16));
449 coeff *= fone >> (30 + 16);
450 } else if (flags & SWS_GAUSS) {
451 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
452 coeff = (pow(2.0, -p * floatd * floatd)) * fone;
453 } else if (flags & SWS_SINC) {
454 coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
455 } else if (flags & SWS_LANCZOS) {
456 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
457 coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
458 (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
461 } else if (flags & SWS_BILINEAR) {
462 coeff = (1 << 30) - d;
466 } else if (flags & SWS_SPLINE) {
467 double p = -2.196152422706632;
468 coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
473 filter[i * filterSize + j] = coeff;
476 xDstInSrc += 2 * xInc;
480 /* apply src & dst Filter to filter -> filter2
483 av_assert0(filterSize > 0);
484 filter2Size = filterSize;
486 filter2Size += srcFilter->length - 1;
488 filter2Size += dstFilter->length - 1;
489 av_assert0(filter2Size > 0);
490 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter2, dstW, filter2Size * sizeof(*filter2), fail);
492 for (i = 0; i < dstW; i++) {
496 for (k = 0; k < srcFilter->length; k++) {
497 for (j = 0; j < filterSize; j++)
498 filter2[i * filter2Size + k + j] +=
499 srcFilter->coeff[k] * filter[i * filterSize + j];
502 for (j = 0; j < filterSize; j++)
503 filter2[i * filter2Size + j] = filter[i * filterSize + j];
507 (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
511 /* try to reduce the filter-size (step1 find size and shift left) */
512 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
514 for (i = dstW - 1; i >= 0; i--) {
515 int min = filter2Size;
517 int64_t cutOff = 0.0;
519 /* get rid of near zero elements on the left by shifting left */
520 for (j = 0; j < filter2Size; j++) {
522 cutOff += FFABS(filter2[i * filter2Size]);
524 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
527 /* preserve monotonicity because the core can't handle the
528 * filter otherwise */
529 if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
532 // move filter coefficients left
533 for (k = 1; k < filter2Size; k++)
534 filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
535 filter2[i * filter2Size + k - 1] = 0;
540 /* count near zeros on the right */
541 for (j = filter2Size - 1; j > 0; j--) {
542 cutOff += FFABS(filter2[i * filter2Size + j]);
544 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
549 if (min > minFilterSize)
553 if (PPC_ALTIVEC(cpu_flags)) {
554 // we can handle the special case 4, so we don't want to go the full 8
555 if (minFilterSize < 5)
558 /* We really don't want to waste our time doing useless computation, so
559 * fall back on the scalar C code for very small filters.
560 * Vectorizing is worth it only if you have a decent-sized vector. */
561 if (minFilterSize < 3)
565 if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
566 // special case for unscaled vertical filtering
567 if (minFilterSize == 1 && filterAlign == 2)
571 av_assert0(minFilterSize > 0);
572 filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
573 av_assert0(filterSize > 0);
574 filter = av_malloc_array(dstW, filterSize * sizeof(*filter));
577 if (filterSize >= MAX_FILTER_SIZE * 16 /
578 ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16)) {
579 ret = RETCODE_USE_CASCADE;
582 *outFilterSize = filterSize;
584 if (flags & SWS_PRINT_INFO)
585 av_log(NULL, AV_LOG_VERBOSE,
586 "SwScaler: reducing / aligning filtersize %d -> %d\n",
587 filter2Size, filterSize);
588 /* try to reduce the filter-size (step2 reduce it) */
589 for (i = 0; i < dstW; i++) {
592 for (j = 0; j < filterSize; j++) {
593 if (j >= filter2Size)
594 filter[i * filterSize + j] = 0;
596 filter[i * filterSize + j] = filter2[i * filter2Size + j];
597 if ((flags & SWS_BITEXACT) && j >= minFilterSize)
598 filter[i * filterSize + j] = 0;
602 // FIXME try to align filterPos if possible
605 for (i = 0; i < dstW; i++) {
607 if ((*filterPos)[i] < 0) {
608 // move filter coefficients left to compensate for filterPos
609 for (j = 1; j < filterSize; j++) {
610 int left = FFMAX(j + (*filterPos)[i], 0);
611 filter[i * filterSize + left] += filter[i * filterSize + j];
612 filter[i * filterSize + j] = 0;
617 if ((*filterPos)[i] + filterSize > srcW) {
618 int shift = (*filterPos)[i] + FFMIN(filterSize - srcW, 0);
621 for (j = filterSize - 1; j >= 0; j--) {
622 if ((*filterPos)[i] + j >= srcW) {
623 acc += filter[i * filterSize + j];
624 filter[i * filterSize + j] = 0;
627 for (j = filterSize - 1; j >= 0; j--) {
629 filter[i * filterSize + j] = 0;
631 filter[i * filterSize + j] = filter[i * filterSize + j - shift];
635 (*filterPos)[i]-= shift;
636 filter[i * filterSize + srcW - 1 - (*filterPos)[i]] += acc;
638 av_assert0((*filterPos)[i] >= 0);
639 av_assert0((*filterPos)[i] < srcW);
640 if ((*filterPos)[i] + filterSize > srcW) {
641 for (j = 0; j < filterSize; j++) {
642 av_assert0((*filterPos)[i] + j < srcW || !filter[i * filterSize + j]);
647 // Note the +1 is for the MMX scaler which reads over the end
648 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
649 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, *outFilter,
650 (dstW + 3), *outFilterSize * sizeof(int16_t), fail);
652 /* normalize & store in outFilter */
653 for (i = 0; i < dstW; i++) {
658 for (j = 0; j < filterSize; j++) {
659 sum += filter[i * filterSize + j];
661 sum = (sum + one / 2) / one;
663 av_log(NULL, AV_LOG_WARNING, "SwScaler: zero vector in scaling\n");
666 for (j = 0; j < *outFilterSize; j++) {
667 int64_t v = filter[i * filterSize + j] + error;
668 int intV = ROUNDED_DIV(v, sum);
669 (*outFilter)[i * (*outFilterSize) + j] = intV;
670 error = v - intV * sum;
674 (*filterPos)[dstW + 0] =
675 (*filterPos)[dstW + 1] =
676 (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
677 * read over the end */
678 for (i = 0; i < *outFilterSize; i++) {
679 int k = (dstW - 1) * (*outFilterSize) + i;
680 (*outFilter)[k + 1 * (*outFilterSize)] =
681 (*outFilter)[k + 2 * (*outFilterSize)] =
682 (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
689 av_log(NULL, ret == RETCODE_USE_CASCADE ? AV_LOG_DEBUG : AV_LOG_ERROR, "sws: initFilter failed\n");
695 static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange)
697 int64_t W, V, Z, Cy, Cu, Cv;
698 int64_t vr = table[0];
699 int64_t ub = table[1];
700 int64_t ug = -table[2];
701 int64_t vg = -table[3];
704 uint8_t *p = (uint8_t*)c->input_rgb2yuv_table;
706 static const int8_t map[] = {
707 BY_IDX, GY_IDX, -1 , BY_IDX, BY_IDX, GY_IDX, -1 , BY_IDX,
708 RY_IDX, -1 , GY_IDX, RY_IDX, RY_IDX, -1 , GY_IDX, RY_IDX,
709 RY_IDX, GY_IDX, -1 , RY_IDX, RY_IDX, GY_IDX, -1 , RY_IDX,
710 BY_IDX, -1 , GY_IDX, BY_IDX, BY_IDX, -1 , GY_IDX, BY_IDX,
711 BU_IDX, GU_IDX, -1 , BU_IDX, BU_IDX, GU_IDX, -1 , BU_IDX,
712 RU_IDX, -1 , GU_IDX, RU_IDX, RU_IDX, -1 , GU_IDX, RU_IDX,
713 RU_IDX, GU_IDX, -1 , RU_IDX, RU_IDX, GU_IDX, -1 , RU_IDX,
714 BU_IDX, -1 , GU_IDX, BU_IDX, BU_IDX, -1 , GU_IDX, BU_IDX,
715 BV_IDX, GV_IDX, -1 , BV_IDX, BV_IDX, GV_IDX, -1 , BV_IDX,
716 RV_IDX, -1 , GV_IDX, RV_IDX, RV_IDX, -1 , GV_IDX, RV_IDX,
717 RV_IDX, GV_IDX, -1 , RV_IDX, RV_IDX, GV_IDX, -1 , RV_IDX,
718 BV_IDX, -1 , GV_IDX, BV_IDX, BV_IDX, -1 , GV_IDX, BV_IDX,
719 RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX,
720 BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX,
721 GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 ,
722 -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX,
723 RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX,
724 BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX,
725 GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 ,
726 -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX,
727 RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX,
728 BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX,
729 GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 ,
730 -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, //23
731 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //24
732 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //25
733 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //26
734 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //27
735 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //28
736 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //29
737 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //30
738 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //31
739 BY_IDX, GY_IDX, RY_IDX, -1 , -1 , -1 , -1 , -1 , //32
740 BU_IDX, GU_IDX, RU_IDX, -1 , -1 , -1 , -1 , -1 , //33
741 BV_IDX, GV_IDX, RV_IDX, -1 , -1 , -1 , -1 , -1 , //34
744 dstRange = 0; //FIXME range = 1 is handled elsewhere
754 W = ROUNDED_DIV(ONE*ONE*ug, ub);
755 V = ROUNDED_DIV(ONE*ONE*vg, vr);
758 Cy = ROUNDED_DIV(cy*Z, ONE);
759 Cu = ROUNDED_DIV(ub*Z, ONE);
760 Cv = ROUNDED_DIV(vr*Z, ONE);
762 c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cy);
763 c->input_rgb2yuv_table[GY_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cy);
764 c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cy);
766 c->input_rgb2yuv_table[RU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cu);
767 c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cu);
768 c->input_rgb2yuv_table[BU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W) , Cu);
770 c->input_rgb2yuv_table[RV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z) , Cv);
771 c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cv);
772 c->input_rgb2yuv_table[BV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cv);
774 if(/*!dstRange && */!memcmp(table, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sizeof(ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]))) {
775 c->input_rgb2yuv_table[BY_IDX] = ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
776 c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
777 c->input_rgb2yuv_table[BU_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
778 c->input_rgb2yuv_table[GY_IDX] = ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
779 c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
780 c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
781 c->input_rgb2yuv_table[RY_IDX] = ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
782 c->input_rgb2yuv_table[RV_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
783 c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
785 for(i=0; i<FF_ARRAY_ELEMS(map); i++)
786 AV_WL16(p + 16*4 + 2*i, map[i] >= 0 ? c->input_rgb2yuv_table[map[i]] : 0);
789 static void fill_xyztables(struct SwsContext *c)
792 double xyzgamma = XYZ_GAMMA;
793 double rgbgamma = 1.0 / RGB_GAMMA;
794 double xyzgammainv = 1.0 / XYZ_GAMMA;
795 double rgbgammainv = RGB_GAMMA;
796 static const int16_t xyz2rgb_matrix[3][4] = {
797 {13270, -6295, -2041},
799 { 228, -835, 4329} };
800 static const int16_t rgb2xyz_matrix[3][4] = {
804 static int16_t xyzgamma_tab[4096], rgbgamma_tab[4096], xyzgammainv_tab[4096], rgbgammainv_tab[4096];
806 memcpy(c->xyz2rgb_matrix, xyz2rgb_matrix, sizeof(c->xyz2rgb_matrix));
807 memcpy(c->rgb2xyz_matrix, rgb2xyz_matrix, sizeof(c->rgb2xyz_matrix));
808 c->xyzgamma = xyzgamma_tab;
809 c->rgbgamma = rgbgamma_tab;
810 c->xyzgammainv = xyzgammainv_tab;
811 c->rgbgammainv = rgbgammainv_tab;
813 if (rgbgamma_tab[4095])
816 /* set gamma vectors */
817 for (i = 0; i < 4096; i++) {
818 xyzgamma_tab[i] = lrint(pow(i / 4095.0, xyzgamma) * 4095.0);
819 rgbgamma_tab[i] = lrint(pow(i / 4095.0, rgbgamma) * 4095.0);
820 xyzgammainv_tab[i] = lrint(pow(i / 4095.0, xyzgammainv) * 4095.0);
821 rgbgammainv_tab[i] = lrint(pow(i / 4095.0, rgbgammainv) * 4095.0);
825 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
826 int srcRange, const int table[4], int dstRange,
827 int brightness, int contrast, int saturation)
829 const AVPixFmtDescriptor *desc_dst;
830 const AVPixFmtDescriptor *desc_src;
832 memmove(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
833 memmove(c->dstColorspaceTable, table, sizeof(int) * 4);
836 desc_dst = av_pix_fmt_desc_get(c->dstFormat);
837 desc_src = av_pix_fmt_desc_get(c->srcFormat);
839 if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
841 if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
844 c->brightness = brightness;
845 c->contrast = contrast;
846 c->saturation = saturation;
847 if (c->srcRange != srcRange || c->dstRange != dstRange)
849 c->srcRange = srcRange;
850 c->dstRange = dstRange;
852 //The srcBpc check is possibly wrong but we seem to lack a definitive reference to test this
853 //and what we have in ticket 2939 looks better with this check
854 if (need_reinit && (c->srcBpc == 8 || !isYUV(c->srcFormat)))
855 ff_sws_init_range_convert(c);
857 if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat)))
860 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
861 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
863 if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) {
864 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
865 contrast, saturation);
869 ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
870 contrast, saturation);
873 fill_rgb2yuv_table(c, table, dstRange);
878 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
879 int *srcRange, int **table, int *dstRange,
880 int *brightness, int *contrast, int *saturation)
885 *inv_table = c->srcColorspaceTable;
886 *table = c->dstColorspaceTable;
887 *srcRange = c->srcRange;
888 *dstRange = c->dstRange;
889 *brightness = c->brightness;
890 *contrast = c->contrast;
891 *saturation = c->saturation;
896 static int handle_jpeg(enum AVPixelFormat *format)
899 case AV_PIX_FMT_YUVJ420P:
900 *format = AV_PIX_FMT_YUV420P;
902 case AV_PIX_FMT_YUVJ411P:
903 *format = AV_PIX_FMT_YUV411P;
905 case AV_PIX_FMT_YUVJ422P:
906 *format = AV_PIX_FMT_YUV422P;
908 case AV_PIX_FMT_YUVJ444P:
909 *format = AV_PIX_FMT_YUV444P;
911 case AV_PIX_FMT_YUVJ440P:
912 *format = AV_PIX_FMT_YUV440P;
914 case AV_PIX_FMT_GRAY8:
915 case AV_PIX_FMT_GRAY16LE:
916 case AV_PIX_FMT_GRAY16BE:
923 static int handle_0alpha(enum AVPixelFormat *format)
926 case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
927 case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
928 case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
929 case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4;
934 static int handle_xyz(enum AVPixelFormat *format)
937 case AV_PIX_FMT_XYZ12BE : *format = AV_PIX_FMT_RGB48BE; return 1;
938 case AV_PIX_FMT_XYZ12LE : *format = AV_PIX_FMT_RGB48LE; return 1;
943 static void handle_formats(SwsContext *c)
945 c->src0Alpha |= handle_0alpha(&c->srcFormat);
946 c->dst0Alpha |= handle_0alpha(&c->dstFormat);
947 c->srcXYZ |= handle_xyz(&c->srcFormat);
948 c->dstXYZ |= handle_xyz(&c->dstFormat);
949 if (c->srcXYZ || c->dstXYZ)
953 SwsContext *sws_alloc_context(void)
955 SwsContext *c = av_mallocz(sizeof(SwsContext));
957 av_assert0(offsetof(SwsContext, redDither) + DITHER32_INT == offsetof(SwsContext, dither32));
960 c->av_class = &sws_context_class;
961 av_opt_set_defaults(c);
967 static uint16_t * alloc_gamma_tbl(double e)
971 tbl = (uint16_t*)av_malloc(sizeof(uint16_t) * 1 << 16);
975 for (i = 0; i < 65536; ++i) {
976 tbl[i] = pow(i / 65535.0, e) * 65535.0;
981 av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
982 SwsFilter *dstFilter)
985 int usesVFilter, usesHFilter;
987 SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
992 int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
993 int flags, cpu_flags;
994 enum AVPixelFormat srcFormat = c->srcFormat;
995 enum AVPixelFormat dstFormat = c->dstFormat;
996 const AVPixFmtDescriptor *desc_src;
997 const AVPixFmtDescriptor *desc_dst;
999 enum AVPixelFormat tmpFmt;
1001 cpu_flags = av_get_cpu_flags();
1007 unscaled = (srcW == dstW && srcH == dstH);
1009 c->srcRange |= handle_jpeg(&c->srcFormat);
1010 c->dstRange |= handle_jpeg(&c->dstFormat);
1012 if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
1013 av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
1015 if (!c->contrast && !c->saturation && !c->dstFormatBpp)
1016 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
1017 ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
1018 c->dstRange, 0, 1 << 16, 1 << 16);
1021 srcFormat = c->srcFormat;
1022 dstFormat = c->dstFormat;
1023 desc_src = av_pix_fmt_desc_get(srcFormat);
1024 desc_dst = av_pix_fmt_desc_get(dstFormat);
1026 if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
1027 av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
1028 if (!sws_isSupportedInput(srcFormat)) {
1029 av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
1030 av_get_pix_fmt_name(srcFormat));
1031 return AVERROR(EINVAL);
1033 if (!sws_isSupportedOutput(dstFormat)) {
1034 av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
1035 av_get_pix_fmt_name(dstFormat));
1036 return AVERROR(EINVAL);
1040 i = flags & (SWS_POINT |
1052 /* provide a default scaler if not set by caller */
1054 if (dstW < srcW && dstH < srcH)
1055 flags |= SWS_BICUBIC;
1056 else if (dstW > srcW && dstH > srcH)
1057 flags |= SWS_BICUBIC;
1059 flags |= SWS_BICUBIC;
1061 } else if (i & (i - 1)) {
1062 av_log(c, AV_LOG_ERROR,
1063 "Exactly one scaler algorithm must be chosen, got %X\n", i);
1064 return AVERROR(EINVAL);
1067 if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
1068 /* FIXME check if these are enough and try to lower them after
1069 * fixing the relevant parts of the code */
1070 av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
1071 srcW, srcH, dstW, dstH);
1072 return AVERROR(EINVAL);
1076 dstFilter = &dummyFilter;
1078 srcFilter = &dummyFilter;
1080 c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
1081 c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
1082 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
1083 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
1084 c->vRounder = 4 * 0x0001000100010001ULL;
1086 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
1087 (srcFilter->chrV && srcFilter->chrV->length > 1) ||
1088 (dstFilter->lumV && dstFilter->lumV->length > 1) ||
1089 (dstFilter->chrV && dstFilter->chrV->length > 1);
1090 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
1091 (srcFilter->chrH && srcFilter->chrH->length > 1) ||
1092 (dstFilter->lumH && dstFilter->lumH->length > 1) ||
1093 (dstFilter->chrH && dstFilter->chrH->length > 1);
1095 av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
1096 av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);
1098 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
1100 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
1101 flags |= SWS_FULL_CHR_H_INT;
1105 if ( c->chrSrcHSubSample == 0
1106 && c->chrSrcVSubSample == 0
1107 && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
1108 && !(c->flags & SWS_FAST_BILINEAR)
1110 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
1111 flags |= SWS_FULL_CHR_H_INT;
1116 if (c->dither == SWS_DITHER_AUTO) {
1117 if (flags & SWS_ERROR_DIFFUSION)
1118 c->dither = SWS_DITHER_ED;
1121 if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1122 dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1123 dstFormat == AV_PIX_FMT_BGR8 ||
1124 dstFormat == AV_PIX_FMT_RGB8) {
1125 if (c->dither == SWS_DITHER_AUTO)
1126 c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
1127 if (!(flags & SWS_FULL_CHR_H_INT)) {
1128 if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER) {
1129 av_log(c, AV_LOG_DEBUG,
1130 "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
1131 av_get_pix_fmt_name(dstFormat));
1132 flags |= SWS_FULL_CHR_H_INT;
1136 if (flags & SWS_FULL_CHR_H_INT) {
1137 if (c->dither == SWS_DITHER_BAYER) {
1138 av_log(c, AV_LOG_DEBUG,
1139 "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1140 av_get_pix_fmt_name(dstFormat));
1141 c->dither = SWS_DITHER_ED;
1145 if (isPlanarRGB(dstFormat)) {
1146 if (!(flags & SWS_FULL_CHR_H_INT)) {
1147 av_log(c, AV_LOG_DEBUG,
1148 "%s output is not supported with half chroma resolution, switching to full\n",
1149 av_get_pix_fmt_name(dstFormat));
1150 flags |= SWS_FULL_CHR_H_INT;
1155 /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1156 * chroma interpolation */
1157 if (flags & SWS_FULL_CHR_H_INT &&
1158 isAnyRGB(dstFormat) &&
1159 !isPlanarRGB(dstFormat) &&
1160 dstFormat != AV_PIX_FMT_RGBA &&
1161 dstFormat != AV_PIX_FMT_ARGB &&
1162 dstFormat != AV_PIX_FMT_BGRA &&
1163 dstFormat != AV_PIX_FMT_ABGR &&
1164 dstFormat != AV_PIX_FMT_RGB24 &&
1165 dstFormat != AV_PIX_FMT_BGR24 &&
1166 dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1167 dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1168 dstFormat != AV_PIX_FMT_BGR8 &&
1169 dstFormat != AV_PIX_FMT_RGB8
1171 av_log(c, AV_LOG_WARNING,
1172 "full chroma interpolation for destination format '%s' not yet implemented\n",
1173 av_get_pix_fmt_name(dstFormat));
1174 flags &= ~SWS_FULL_CHR_H_INT;
1177 if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1178 c->chrDstHSubSample = 1;
1180 // drop some chroma lines if the user wants it
1181 c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1182 SWS_SRC_V_CHR_DROP_SHIFT;
1183 c->chrSrcVSubSample += c->vChrDrop;
1185 /* drop every other pixel for chroma calculation unless user
1186 * wants full chroma */
1187 if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) &&
1188 srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1189 srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1190 srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1191 srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
1192 srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
1193 srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
1194 srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE &&
1195 srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
1196 ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1197 (flags & SWS_FAST_BILINEAR)))
1198 c->chrSrcHSubSample = 1;
1200 // Note the FF_CEIL_RSHIFT is so that we always round toward +inf.
1201 c->chrSrcW = FF_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
1202 c->chrSrcH = FF_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
1203 c->chrDstW = FF_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
1204 c->chrDstH = FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
1206 FF_ALLOCZ_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
1208 c->srcBpc = 1 + desc_src->comp[0].depth_minus1;
1211 c->dstBpc = 1 + desc_dst->comp[0].depth_minus1;
1214 if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1216 if (c->dstBpc == 16)
1219 if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1220 c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1221 c->chrDstW >= c->chrSrcW &&
1223 if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1225 && (flags & SWS_FAST_BILINEAR)) {
1226 if (flags & SWS_PRINT_INFO)
1227 av_log(c, AV_LOG_INFO,
1228 "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1230 if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1231 c->canMMXEXTBeUsed = 0;
1233 c->canMMXEXTBeUsed = 0;
1235 c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1236 c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1238 /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1239 * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1241 * n-2 is the last chrominance sample available.
1242 * This is not perfect, but no one should notice the difference, the more
1243 * correct variant would be like the vertical one, but that would require
1244 * some special code for the first and last pixel */
1245 if (flags & SWS_FAST_BILINEAR) {
1246 if (c->canMMXEXTBeUsed) {
1250 // we don't use the x86 asm scaler if MMX is available
1251 else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1252 c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
1253 c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1257 // hardcoded for now
1258 c->gamma_value = 2.2;
1259 tmpFmt = AV_PIX_FMT_RGBA64LE;
1262 if (!unscaled && c->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
1264 c->cascaded_context[0] = NULL;
1266 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1267 srcW, srcH, tmpFmt, 64);
1271 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1273 flags, NULL, NULL, c->param);
1274 if (!c->cascaded_context[0]) {
1278 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
1280 flags, srcFilter, dstFilter, c->param);
1282 if (!c->cascaded_context[1])
1285 c2 = c->cascaded_context[1];
1286 c2->is_internal_gamma = 1;
1287 c2->gamma = alloc_gamma_tbl( c->gamma_value);
1288 c2->inv_gamma = alloc_gamma_tbl(1.f/c->gamma_value);
1289 if (!c2->gamma || !c2->inv_gamma)
1290 return AVERROR(ENOMEM);
1292 c->cascaded_context[2] = NULL;
1293 if (dstFormat != tmpFmt) {
1294 ret = av_image_alloc(c->cascaded1_tmp, c->cascaded1_tmpStride,
1295 dstW, dstH, tmpFmt, 64);
1299 c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
1300 dstW, dstH, dstFormat,
1301 flags, NULL, NULL, c->param);
1302 if (!c->cascaded_context[2])
1308 if (isBayer(srcFormat)) {
1310 (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P)) {
1311 enum AVPixelFormat tmpFormat = AV_PIX_FMT_RGB24;
1313 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1314 srcW, srcH, tmpFormat, 64);
1318 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1319 srcW, srcH, tmpFormat,
1320 flags, srcFilter, NULL, c->param);
1321 if (!c->cascaded_context[0])
1324 c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
1325 dstW, dstH, dstFormat,
1326 flags, NULL, dstFilter, c->param);
1327 if (!c->cascaded_context[1])
1333 #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
1335 /* precalculate horizontal scaler filter coefficients */
1337 #if HAVE_MMXEXT_INLINE
1338 // can't downscale !!!
1339 if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1340 c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1342 c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1343 NULL, NULL, NULL, 4);
1346 c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
1347 PROT_READ | PROT_WRITE,
1348 MAP_PRIVATE | MAP_ANONYMOUS,
1350 c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
1351 PROT_READ | PROT_WRITE,
1352 MAP_PRIVATE | MAP_ANONYMOUS,
1354 #elif HAVE_VIRTUALALLOC
1355 c->lumMmxextFilterCode = VirtualAlloc(NULL,
1356 c->lumMmxextFilterCodeSize,
1358 PAGE_EXECUTE_READWRITE);
1359 c->chrMmxextFilterCode = VirtualAlloc(NULL,
1360 c->chrMmxextFilterCodeSize,
1362 PAGE_EXECUTE_READWRITE);
1364 c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
1365 c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
1368 #ifdef MAP_ANONYMOUS
1369 if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1371 if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
1374 av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1375 return AVERROR(ENOMEM);
1378 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
1379 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
1380 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
1381 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);
1383 ff_init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
1384 c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1385 ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1386 c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1389 if ( mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
1390 || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
1391 av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
1396 #endif /* HAVE_MMXEXT_INLINE */
1398 const int filterAlign = X86_MMX(cpu_flags) ? 4 :
1399 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1401 if ((ret = initFilter(&c->hLumFilter, &c->hLumFilterPos,
1402 &c->hLumFilterSize, c->lumXInc,
1403 srcW, dstW, filterAlign, 1 << 14,
1404 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1405 cpu_flags, srcFilter->lumH, dstFilter->lumH,
1407 get_local_pos(c, 0, 0, 0),
1408 get_local_pos(c, 0, 0, 0))) < 0)
1410 if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
1411 &c->hChrFilterSize, c->chrXInc,
1412 c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1413 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1414 cpu_flags, srcFilter->chrH, dstFilter->chrH,
1416 get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
1417 get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
1420 } // initialize horizontal stuff
1422 /* precalculate vertical scaler filter coefficients */
1424 const int filterAlign = X86_MMX(cpu_flags) ? 2 :
1425 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1427 if ((ret = initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
1428 c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1429 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1430 cpu_flags, srcFilter->lumV, dstFilter->lumV,
1432 get_local_pos(c, 0, 0, 1),
1433 get_local_pos(c, 0, 0, 1))) < 0)
1435 if ((ret = initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
1436 c->chrYInc, c->chrSrcH, c->chrDstH,
1437 filterAlign, (1 << 12),
1438 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1439 cpu_flags, srcFilter->chrV, dstFilter->chrV,
1441 get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
1442 get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1))) < 0)
1447 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
1448 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);
1450 for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1452 short *p = (short *)&c->vYCoeffsBank[i];
1453 for (j = 0; j < 8; j++)
1454 p[j] = c->vLumFilter[i];
1457 for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1459 short *p = (short *)&c->vCCoeffsBank[i];
1460 for (j = 0; j < 8; j++)
1461 p[j] = c->vChrFilter[i];
1466 // calculate buffer sizes so that they won't run out while handling these damn slices
1467 c->vLumBufSize = c->vLumFilterSize;
1468 c->vChrBufSize = c->vChrFilterSize;
1469 for (i = 0; i < dstH; i++) {
1470 int chrI = (int64_t)i * c->chrDstH / dstH;
1471 int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
1472 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
1473 << c->chrSrcVSubSample));
1475 nextSlice >>= c->chrSrcVSubSample;
1476 nextSlice <<= c->chrSrcVSubSample;
1477 if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
1478 c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
1479 if (c->vChrFilterPos[chrI] + c->vChrBufSize <
1480 (nextSlice >> c->chrSrcVSubSample))
1481 c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
1482 c->vChrFilterPos[chrI];
1485 for (i = 0; i < 4; i++)
1486 FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);
1488 /* Allocate pixbufs (we use dynamic allocation because otherwise we would
1489 * need to allocate several megabytes to handle all possible cases) */
1490 FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1491 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1492 FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1493 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
1494 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1495 /* Note we need at least one pixel more at the end because of the MMX code
1496 * (just in case someone wants to replace the 4000/8000). */
1497 /* align at 16 bytes for AltiVec */
1498 for (i = 0; i < c->vLumBufSize; i++) {
1499 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
1500 dst_stride + 16, fail);
1501 c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
1503 // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1504 c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1505 c->uv_offx2 = dst_stride + 16;
1506 for (i = 0; i < c->vChrBufSize; i++) {
1507 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
1508 dst_stride * 2 + 32, fail);
1509 c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
1510 c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
1511 = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
1513 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1514 for (i = 0; i < c->vLumBufSize; i++) {
1515 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
1516 dst_stride + 16, fail);
1517 c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
1520 // try to avoid drawing green stuff between the right end and the stride end
1521 for (i = 0; i < c->vChrBufSize; i++)
1522 if(desc_dst->comp[0].depth_minus1 == 15){
1523 av_assert0(c->dstBpc > 14);
1524 for(j=0; j<dst_stride/2+1; j++)
1525 ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
1527 for(j=0; j<dst_stride+1; j++)
1528 ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
1530 av_assert0(c->chrDstH <= dstH);
1532 if (flags & SWS_PRINT_INFO) {
1533 const char *scaler = NULL, *cpucaps;
1535 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
1536 if (flags & scale_algorithms[i].flag) {
1537 scaler = scale_algorithms[i].description;
1542 scaler = "ehh flags invalid?!";
1543 av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
1545 av_get_pix_fmt_name(srcFormat),
1547 dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
1548 dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1549 dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1554 av_get_pix_fmt_name(dstFormat));
1556 if (INLINE_MMXEXT(cpu_flags))
1558 else if (INLINE_AMD3DNOW(cpu_flags))
1560 else if (INLINE_MMX(cpu_flags))
1562 else if (PPC_ALTIVEC(cpu_flags))
1563 cpucaps = "AltiVec";
1567 av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);
1569 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1570 av_log(c, AV_LOG_DEBUG,
1571 "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1572 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1573 av_log(c, AV_LOG_DEBUG,
1574 "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1575 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1576 c->chrXInc, c->chrYInc);
1579 /* unscaled special cases */
1580 if (unscaled && !usesHFilter && !usesVFilter &&
1581 (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
1582 ff_get_unscaled_swscale(c);
1585 if (flags & SWS_PRINT_INFO)
1586 av_log(c, AV_LOG_INFO,
1587 "using unscaled %s -> %s special converter\n",
1588 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1593 c->swscale = ff_getSwsFunc(c);
1595 fail: // FIXME replace things by appropriate error codes
1596 if (ret == RETCODE_USE_CASCADE) {
1597 int tmpW = sqrt(srcW * (int64_t)dstW);
1598 int tmpH = sqrt(srcH * (int64_t)dstH);
1599 enum AVPixelFormat tmpFormat = AV_PIX_FMT_YUV420P;
1601 if (srcW*(int64_t)srcH <= 4LL*dstW*dstH)
1602 return AVERROR(EINVAL);
1604 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1605 tmpW, tmpH, tmpFormat, 64);
1609 c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1610 tmpW, tmpH, tmpFormat,
1611 flags, srcFilter, NULL, c->param);
1612 if (!c->cascaded_context[0])
1615 c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
1616 dstW, dstH, dstFormat,
1617 flags, NULL, dstFilter, c->param);
1618 if (!c->cascaded_context[1])
1625 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
1626 int dstW, int dstH, enum AVPixelFormat dstFormat,
1627 int flags, SwsFilter *srcFilter,
1628 SwsFilter *dstFilter, const double *param)
1632 if (!(c = sws_alloc_context()))
1640 c->srcFormat = srcFormat;
1641 c->dstFormat = dstFormat;
1644 c->param[0] = param[0];
1645 c->param[1] = param[1];
1648 if (sws_init_context(c, srcFilter, dstFilter) < 0) {
1656 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
1657 float lumaSharpen, float chromaSharpen,
1658 float chromaHShift, float chromaVShift,
1661 SwsFilter *filter = av_malloc(sizeof(SwsFilter));
1665 if (lumaGBlur != 0.0) {
1666 filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
1667 filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
1669 filter->lumH = sws_getIdentityVec();
1670 filter->lumV = sws_getIdentityVec();
1673 if (chromaGBlur != 0.0) {
1674 filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
1675 filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
1677 filter->chrH = sws_getIdentityVec();
1678 filter->chrV = sws_getIdentityVec();
1681 if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
1684 if (chromaSharpen != 0.0) {
1685 SwsVector *id = sws_getIdentityVec();
1688 sws_scaleVec(filter->chrH, -chromaSharpen);
1689 sws_scaleVec(filter->chrV, -chromaSharpen);
1690 sws_addVec(filter->chrH, id);
1691 sws_addVec(filter->chrV, id);
1695 if (lumaSharpen != 0.0) {
1696 SwsVector *id = sws_getIdentityVec();
1699 sws_scaleVec(filter->lumH, -lumaSharpen);
1700 sws_scaleVec(filter->lumV, -lumaSharpen);
1701 sws_addVec(filter->lumH, id);
1702 sws_addVec(filter->lumV, id);
1706 if (chromaHShift != 0.0)
1707 sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
1709 if (chromaVShift != 0.0)
1710 sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
1712 sws_normalizeVec(filter->chrH, 1.0);
1713 sws_normalizeVec(filter->chrV, 1.0);
1714 sws_normalizeVec(filter->lumH, 1.0);
1715 sws_normalizeVec(filter->lumV, 1.0);
1718 sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
1720 sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
1725 sws_freeVec(filter->lumH);
1726 sws_freeVec(filter->lumV);
1727 sws_freeVec(filter->chrH);
1728 sws_freeVec(filter->chrV);
1733 SwsVector *sws_allocVec(int length)
1737 if(length <= 0 || length > INT_MAX/ sizeof(double))
1740 vec = av_malloc(sizeof(SwsVector));
1743 vec->length = length;
1744 vec->coeff = av_malloc(sizeof(double) * length);
1750 SwsVector *sws_getGaussianVec(double variance, double quality)
1752 const int length = (int)(variance * quality + 0.5) | 1;
1754 double middle = (length - 1) * 0.5;
1757 if(variance < 0 || quality < 0)
1760 vec = sws_allocVec(length);
1765 for (i = 0; i < length; i++) {
1766 double dist = i - middle;
1767 vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
1768 sqrt(2 * variance * M_PI);
1771 sws_normalizeVec(vec, 1.0);
1776 SwsVector *sws_getConstVec(double c, int length)
1779 SwsVector *vec = sws_allocVec(length);
1784 for (i = 0; i < length; i++)
1790 SwsVector *sws_getIdentityVec(void)
1792 return sws_getConstVec(1.0, 1);
1795 static double sws_dcVec(SwsVector *a)
1800 for (i = 0; i < a->length; i++)
1806 void sws_scaleVec(SwsVector *a, double scalar)
1810 for (i = 0; i < a->length; i++)
1811 a->coeff[i] *= scalar;
1814 void sws_normalizeVec(SwsVector *a, double height)
1816 sws_scaleVec(a, height / sws_dcVec(a));
1819 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
1821 int length = a->length + b->length - 1;
1823 SwsVector *vec = sws_getConstVec(0.0, length);
1828 for (i = 0; i < a->length; i++) {
1829 for (j = 0; j < b->length; j++) {
1830 vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
1837 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
1839 int length = FFMAX(a->length, b->length);
1841 SwsVector *vec = sws_getConstVec(0.0, length);
1846 for (i = 0; i < a->length; i++)
1847 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1848 for (i = 0; i < b->length; i++)
1849 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
1854 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
1856 int length = FFMAX(a->length, b->length);
1858 SwsVector *vec = sws_getConstVec(0.0, length);
1863 for (i = 0; i < a->length; i++)
1864 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1865 for (i = 0; i < b->length; i++)
1866 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];
1871 /* shift left / or right if "shift" is negative */
1872 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
1874 int length = a->length + FFABS(shift) * 2;
1876 SwsVector *vec = sws_getConstVec(0.0, length);
1881 for (i = 0; i < a->length; i++) {
1882 vec->coeff[i + (length - 1) / 2 -
1883 (a->length - 1) / 2 - shift] = a->coeff[i];
1889 void sws_shiftVec(SwsVector *a, int shift)
1891 SwsVector *shifted = sws_getShiftedVec(a, shift);
1893 a->coeff = shifted->coeff;
1894 a->length = shifted->length;
1898 void sws_addVec(SwsVector *a, SwsVector *b)
1900 SwsVector *sum = sws_sumVec(a, b);
1902 a->coeff = sum->coeff;
1903 a->length = sum->length;
1907 void sws_subVec(SwsVector *a, SwsVector *b)
1909 SwsVector *diff = sws_diffVec(a, b);
1911 a->coeff = diff->coeff;
1912 a->length = diff->length;
1916 void sws_convVec(SwsVector *a, SwsVector *b)
1918 SwsVector *conv = sws_getConvVec(a, b);
1920 a->coeff = conv->coeff;
1921 a->length = conv->length;
1925 SwsVector *sws_cloneVec(SwsVector *a)
1927 SwsVector *vec = sws_allocVec(a->length);
1932 memcpy(vec->coeff, a->coeff, a->length * sizeof(*a->coeff));
1937 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
1944 for (i = 0; i < a->length; i++)
1945 if (a->coeff[i] > max)
1948 for (i = 0; i < a->length; i++)
1949 if (a->coeff[i] < min)
1954 for (i = 0; i < a->length; i++) {
1955 int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
1956 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
1958 av_log(log_ctx, log_level, " ");
1959 av_log(log_ctx, log_level, "|\n");
1963 void sws_freeVec(SwsVector *a)
1967 av_freep(&a->coeff);
1972 void sws_freeFilter(SwsFilter *filter)
1977 sws_freeVec(filter->lumH);
1978 sws_freeVec(filter->lumV);
1979 sws_freeVec(filter->chrH);
1980 sws_freeVec(filter->chrV);
1984 void sws_freeContext(SwsContext *c)
1991 for (i = 0; i < c->vLumBufSize; i++)
1992 av_freep(&c->lumPixBuf[i]);
1993 av_freep(&c->lumPixBuf);
1996 if (c->chrUPixBuf) {
1997 for (i = 0; i < c->vChrBufSize; i++)
1998 av_freep(&c->chrUPixBuf[i]);
1999 av_freep(&c->chrUPixBuf);
2000 av_freep(&c->chrVPixBuf);
2003 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
2004 for (i = 0; i < c->vLumBufSize; i++)
2005 av_freep(&c->alpPixBuf[i]);
2006 av_freep(&c->alpPixBuf);
2009 for (i = 0; i < 4; i++)
2010 av_freep(&c->dither_error[i]);
2012 av_freep(&c->vLumFilter);
2013 av_freep(&c->vChrFilter);
2014 av_freep(&c->hLumFilter);
2015 av_freep(&c->hChrFilter);
2017 av_freep(&c->vYCoeffsBank);
2018 av_freep(&c->vCCoeffsBank);
2021 av_freep(&c->vLumFilterPos);
2022 av_freep(&c->vChrFilterPos);
2023 av_freep(&c->hLumFilterPos);
2024 av_freep(&c->hChrFilterPos);
2028 if (c->lumMmxextFilterCode)
2029 munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
2030 if (c->chrMmxextFilterCode)
2031 munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize);
2032 #elif HAVE_VIRTUALALLOC
2033 if (c->lumMmxextFilterCode)
2034 VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
2035 if (c->chrMmxextFilterCode)
2036 VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
2038 av_free(c->lumMmxextFilterCode);
2039 av_free(c->chrMmxextFilterCode);
2041 c->lumMmxextFilterCode = NULL;
2042 c->chrMmxextFilterCode = NULL;
2043 #endif /* HAVE_MMX_INLINE */
2045 av_freep(&c->yuvTable);
2046 av_freep(&c->formatConvBuffer);
2048 sws_freeContext(c->cascaded_context[0]);
2049 sws_freeContext(c->cascaded_context[1]);
2050 sws_freeContext(c->cascaded_context[2]);
2051 memset(c->cascaded_context, 0, sizeof(c->cascaded_context));
2052 av_freep(&c->cascaded_tmp[0]);
2053 av_freep(&c->cascaded1_tmp[0]);
2055 av_freep(&c->gamma);
2056 av_freep(&c->inv_gamma);
2062 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
2063 int srcH, enum AVPixelFormat srcFormat,
2065 enum AVPixelFormat dstFormat, int flags,
2066 SwsFilter *srcFilter,
2067 SwsFilter *dstFilter,
2068 const double *param)
2070 static const double default_param[2] = { SWS_PARAM_DEFAULT,
2071 SWS_PARAM_DEFAULT };
2072 int64_t src_h_chr_pos = -513, dst_h_chr_pos = -513,
2073 src_v_chr_pos = -513, dst_v_chr_pos = -513;
2076 param = default_param;
2079 (context->srcW != srcW ||
2080 context->srcH != srcH ||
2081 context->srcFormat != srcFormat ||
2082 context->dstW != dstW ||
2083 context->dstH != dstH ||
2084 context->dstFormat != dstFormat ||
2085 context->flags != flags ||
2086 context->param[0] != param[0] ||
2087 context->param[1] != param[1])) {
2089 av_opt_get_int(context, "src_h_chr_pos", 0, &src_h_chr_pos);
2090 av_opt_get_int(context, "src_v_chr_pos", 0, &src_v_chr_pos);
2091 av_opt_get_int(context, "dst_h_chr_pos", 0, &dst_h_chr_pos);
2092 av_opt_get_int(context, "dst_v_chr_pos", 0, &dst_v_chr_pos);
2093 sws_freeContext(context);
2098 if (!(context = sws_alloc_context()))
2100 context->srcW = srcW;
2101 context->srcH = srcH;
2102 context->srcFormat = srcFormat;
2103 context->dstW = dstW;
2104 context->dstH = dstH;
2105 context->dstFormat = dstFormat;
2106 context->flags = flags;
2107 context->param[0] = param[0];
2108 context->param[1] = param[1];
2110 av_opt_set_int(context, "src_h_chr_pos", src_h_chr_pos, 0);
2111 av_opt_set_int(context, "src_v_chr_pos", src_v_chr_pos, 0);
2112 av_opt_set_int(context, "dst_h_chr_pos", dst_h_chr_pos, 0);
2113 av_opt_set_int(context, "dst_v_chr_pos", dst_v_chr_pos, 0);
2115 if (sws_init_context(context, srcFilter, dstFilter) < 0) {
2116 sws_freeContext(context);