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/intreadwrite.h"
46 #include "libavutil/mathematics.h"
47 #include "libavutil/opt.h"
48 #include "libavutil/pixdesc.h"
49 #include "libavutil/ppc/cpu.h"
50 #include "libavutil/x86/asm.h"
51 #include "libavutil/x86/cpu.h"
54 #include "swscale_internal.h"
56 static void handle_formats(SwsContext *c);
58 unsigned swscale_version(void)
60 av_assert0(LIBSWSCALE_VERSION_MICRO >= 100);
61 return LIBSWSCALE_VERSION_INT;
64 const char *swscale_configuration(void)
66 return FFMPEG_CONFIGURATION;
69 const char *swscale_license(void)
71 #define LICENSE_PREFIX "libswscale license: "
72 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
75 typedef struct FormatEntry {
76 uint8_t is_supported_in :1;
77 uint8_t is_supported_out :1;
78 uint8_t is_supported_endianness :1;
81 static const FormatEntry format_entries[AV_PIX_FMT_NB] = {
82 [AV_PIX_FMT_YUV420P] = { 1, 1 },
83 [AV_PIX_FMT_YUYV422] = { 1, 1 },
84 [AV_PIX_FMT_RGB24] = { 1, 1 },
85 [AV_PIX_FMT_BGR24] = { 1, 1 },
86 [AV_PIX_FMT_YUV422P] = { 1, 1 },
87 [AV_PIX_FMT_YUV444P] = { 1, 1 },
88 [AV_PIX_FMT_YUV410P] = { 1, 1 },
89 [AV_PIX_FMT_YUV411P] = { 1, 1 },
90 [AV_PIX_FMT_GRAY8] = { 1, 1 },
91 [AV_PIX_FMT_MONOWHITE] = { 1, 1 },
92 [AV_PIX_FMT_MONOBLACK] = { 1, 1 },
93 [AV_PIX_FMT_PAL8] = { 1, 0 },
94 [AV_PIX_FMT_YUVJ420P] = { 1, 1 },
95 [AV_PIX_FMT_YUVJ411P] = { 1, 1 },
96 [AV_PIX_FMT_YUVJ422P] = { 1, 1 },
97 [AV_PIX_FMT_YUVJ444P] = { 1, 1 },
98 [AV_PIX_FMT_YVYU422] = { 1, 1 },
99 [AV_PIX_FMT_UYVY422] = { 1, 1 },
100 [AV_PIX_FMT_UYYVYY411] = { 0, 0 },
101 [AV_PIX_FMT_BGR8] = { 1, 1 },
102 [AV_PIX_FMT_BGR4] = { 0, 1 },
103 [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
104 [AV_PIX_FMT_RGB8] = { 1, 1 },
105 [AV_PIX_FMT_RGB4] = { 0, 1 },
106 [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
107 [AV_PIX_FMT_NV12] = { 1, 1 },
108 [AV_PIX_FMT_NV21] = { 1, 1 },
109 [AV_PIX_FMT_ARGB] = { 1, 1 },
110 [AV_PIX_FMT_RGBA] = { 1, 1 },
111 [AV_PIX_FMT_ABGR] = { 1, 1 },
112 [AV_PIX_FMT_BGRA] = { 1, 1 },
113 [AV_PIX_FMT_0RGB] = { 1, 1 },
114 [AV_PIX_FMT_RGB0] = { 1, 1 },
115 [AV_PIX_FMT_0BGR] = { 1, 1 },
116 [AV_PIX_FMT_BGR0] = { 1, 1 },
117 [AV_PIX_FMT_GRAY16BE] = { 1, 1 },
118 [AV_PIX_FMT_GRAY16LE] = { 1, 1 },
119 [AV_PIX_FMT_YUV440P] = { 1, 1 },
120 [AV_PIX_FMT_YUVJ440P] = { 1, 1 },
121 [AV_PIX_FMT_YUVA420P] = { 1, 1 },
122 [AV_PIX_FMT_YUVA422P] = { 1, 1 },
123 [AV_PIX_FMT_YUVA444P] = { 1, 1 },
124 [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
125 [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
126 [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
127 [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
128 [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
129 [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
130 [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
131 [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
132 [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
133 [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
134 [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
135 [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
136 [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
137 [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
138 [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
139 [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
140 [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
141 [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
142 [AV_PIX_FMT_RGB48BE] = { 1, 1 },
143 [AV_PIX_FMT_RGB48LE] = { 1, 1 },
144 [AV_PIX_FMT_RGBA64BE] = { 1, 1, 1 },
145 [AV_PIX_FMT_RGBA64LE] = { 1, 1, 1 },
146 [AV_PIX_FMT_RGB565BE] = { 1, 1 },
147 [AV_PIX_FMT_RGB565LE] = { 1, 1 },
148 [AV_PIX_FMT_RGB555BE] = { 1, 1 },
149 [AV_PIX_FMT_RGB555LE] = { 1, 1 },
150 [AV_PIX_FMT_BGR565BE] = { 1, 1 },
151 [AV_PIX_FMT_BGR565LE] = { 1, 1 },
152 [AV_PIX_FMT_BGR555BE] = { 1, 1 },
153 [AV_PIX_FMT_BGR555LE] = { 1, 1 },
154 [AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
155 [AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
156 [AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
157 [AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
158 [AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
159 [AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
160 [AV_PIX_FMT_RGB444LE] = { 1, 1 },
161 [AV_PIX_FMT_RGB444BE] = { 1, 1 },
162 [AV_PIX_FMT_BGR444LE] = { 1, 1 },
163 [AV_PIX_FMT_BGR444BE] = { 1, 1 },
164 [AV_PIX_FMT_Y400A] = { 1, 0 },
165 [AV_PIX_FMT_BGR48BE] = { 1, 1 },
166 [AV_PIX_FMT_BGR48LE] = { 1, 1 },
167 [AV_PIX_FMT_BGRA64BE] = { 1, 1, 1 },
168 [AV_PIX_FMT_BGRA64LE] = { 1, 1, 1 },
169 [AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
170 [AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
171 [AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
172 [AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
173 [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
174 [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
175 [AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
176 [AV_PIX_FMT_YUV420P14LE] = { 1, 1 },
177 [AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
178 [AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
179 [AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
180 [AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
181 [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
182 [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
183 [AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
184 [AV_PIX_FMT_YUV422P14LE] = { 1, 1 },
185 [AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
186 [AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
187 [AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
188 [AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
189 [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
190 [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
191 [AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
192 [AV_PIX_FMT_YUV444P14LE] = { 1, 1 },
193 [AV_PIX_FMT_GBRP] = { 1, 1 },
194 [AV_PIX_FMT_GBRP9LE] = { 1, 1 },
195 [AV_PIX_FMT_GBRP9BE] = { 1, 1 },
196 [AV_PIX_FMT_GBRP10LE] = { 1, 1 },
197 [AV_PIX_FMT_GBRP10BE] = { 1, 1 },
198 [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
199 [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
200 [AV_PIX_FMT_GBRP14LE] = { 1, 1 },
201 [AV_PIX_FMT_GBRP14BE] = { 1, 1 },
202 [AV_PIX_FMT_GBRP16LE] = { 1, 0 },
203 [AV_PIX_FMT_GBRP16BE] = { 1, 0 },
204 [AV_PIX_FMT_XYZ12BE] = { 1, 1, 1 },
205 [AV_PIX_FMT_XYZ12LE] = { 1, 1, 1 },
206 [AV_PIX_FMT_GBRAP] = { 1, 1 },
207 [AV_PIX_FMT_GBRAP16LE] = { 1, 0 },
208 [AV_PIX_FMT_GBRAP16BE] = { 1, 0 },
209 [AV_PIX_FMT_BAYER_BGGR8] = { 1, 0 },
210 [AV_PIX_FMT_BAYER_RGGB8] = { 1, 0 },
211 [AV_PIX_FMT_BAYER_GBRG8] = { 1, 0 },
212 [AV_PIX_FMT_BAYER_GRBG8] = { 1, 0 },
213 [AV_PIX_FMT_BAYER_BGGR16LE] = { 1, 0 },
214 [AV_PIX_FMT_BAYER_BGGR16BE] = { 1, 0 },
215 [AV_PIX_FMT_BAYER_RGGB16LE] = { 1, 0 },
216 [AV_PIX_FMT_BAYER_RGGB16BE] = { 1, 0 },
217 [AV_PIX_FMT_BAYER_GBRG16LE] = { 1, 0 },
218 [AV_PIX_FMT_BAYER_GBRG16BE] = { 1, 0 },
219 [AV_PIX_FMT_BAYER_GRBG16LE] = { 1, 0 },
220 [AV_PIX_FMT_BAYER_GRBG16BE] = { 1, 0 },
223 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
225 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
226 format_entries[pix_fmt].is_supported_in : 0;
229 int sws_isSupportedOutput(enum AVPixelFormat pix_fmt)
231 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
232 format_entries[pix_fmt].is_supported_out : 0;
235 int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
237 return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
238 format_entries[pix_fmt].is_supported_endianness : 0;
241 #if FF_API_SWS_FORMAT_NAME
242 const char *sws_format_name(enum AVPixelFormat format)
244 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);
248 return "Unknown format";
252 static double getSplineCoeff(double a, double b, double c, double d,
256 return ((d * dist + c) * dist + b) * dist + a;
258 return getSplineCoeff(0.0,
259 b + 2.0 * c + 3.0 * d,
261 -b - 3.0 * c - 6.0 * d,
265 static av_cold int get_local_pos(SwsContext *s, int chr_subsample, int pos, int dir)
268 pos = (128 << chr_subsample) - 128;
270 pos += 128; // relative to ideal left edge
271 return pos >> chr_subsample;
275 int flag; ///< flag associated to the algorithm
276 const char *description; ///< human-readable description
277 int size_factor; ///< size factor used when initing the filters
280 static const ScaleAlgorithm scale_algorithms[] = {
281 { SWS_AREA, "area averaging", 1 /* downscale only, for upscale it is bilinear */ },
282 { SWS_BICUBIC, "bicubic", 4 },
283 { SWS_BICUBLIN, "luma bicubic / chroma bilinear", -1 },
284 { SWS_BILINEAR, "bilinear", 2 },
285 { SWS_FAST_BILINEAR, "fast bilinear", -1 },
286 { SWS_GAUSS, "Gaussian", 8 /* infinite ;) */ },
287 { SWS_LANCZOS, "Lanczos", -1 /* custom */ },
288 { SWS_POINT, "nearest neighbor / point", -1 },
289 { SWS_SINC, "sinc", 20 /* infinite ;) */ },
290 { SWS_SPLINE, "bicubic spline", 20 /* infinite :)*/ },
291 { SWS_X, "experimental", 8 },
294 static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
295 int *outFilterSize, int xInc, int srcW,
296 int dstW, int filterAlign, int one,
297 int flags, int cpu_flags,
298 SwsVector *srcFilter, SwsVector *dstFilter,
299 double param[2], int srcPos, int dstPos)
305 int64_t *filter = NULL;
306 int64_t *filter2 = NULL;
307 const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
310 emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
312 // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
313 FF_ALLOC_ARRAY_OR_GOTO(NULL, *filterPos, (dstW + 3), sizeof(**filterPos), fail);
315 if (FFABS(xInc - 0x10000) < 10 && srcPos == dstPos) { // unscaled
318 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter,
319 dstW, sizeof(*filter) * filterSize, fail);
321 for (i = 0; i < dstW; i++) {
322 filter[i * filterSize] = fone;
325 } else if (flags & SWS_POINT) { // lame looking point sampling mode
329 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
330 dstW, sizeof(*filter) * filterSize, fail);
332 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
333 for (i = 0; i < dstW; i++) {
334 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
336 (*filterPos)[i] = xx;
340 } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
341 (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
345 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
346 dstW, sizeof(*filter) * filterSize, fail);
348 xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
349 for (i = 0; i < dstW; i++) {
350 int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
353 (*filterPos)[i] = xx;
354 // bilinear upscale / linear interpolate / area averaging
355 for (j = 0; j < filterSize; j++) {
356 int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
359 filter[i * filterSize + j] = coeff;
368 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
369 if (flags & scale_algorithms[i].flag && scale_algorithms[i].size_factor > 0) {
370 sizeFactor = scale_algorithms[i].size_factor;
374 if (flags & SWS_LANCZOS)
375 sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
376 av_assert0(sizeFactor > 0);
379 filterSize = 1 + sizeFactor; // upscale
381 filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
383 filterSize = FFMIN(filterSize, srcW - 2);
384 filterSize = FFMAX(filterSize, 1);
386 FF_ALLOC_ARRAY_OR_GOTO(NULL, filter,
387 dstW, sizeof(*filter) * filterSize, fail);
389 xDstInSrc = ((dstPos*(int64_t)xInc)>>7) - ((srcPos*0x10000LL)>>7);
390 for (i = 0; i < dstW; i++) {
391 int xx = (xDstInSrc - ((int64_t)(filterSize - 2) << 16)) / (1 << 17);
393 (*filterPos)[i] = xx;
394 for (j = 0; j < filterSize; j++) {
395 int64_t d = (FFABS(((int64_t)xx << 17) - xDstInSrc)) << 13;
401 floatd = d * (1.0 / (1 << 30));
403 if (flags & SWS_BICUBIC) {
404 int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
405 int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
407 if (d >= 1LL << 31) {
410 int64_t dd = (d * d) >> 30;
411 int64_t ddd = (dd * d) >> 30;
414 coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
415 (-18 * (1 << 24) + 12 * B + 6 * C) * dd +
416 (6 * (1 << 24) - 2 * B) * (1 << 30);
418 coeff = (-B - 6 * C) * ddd +
419 (6 * B + 30 * C) * dd +
420 (-12 * B - 48 * C) * d +
421 (8 * B + 24 * C) * (1 << 30);
423 coeff /= (1LL<<54)/fone;
426 else if (flags & SWS_X) {
427 double p = param ? param * 0.01 : 0.3;
428 coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
429 coeff *= pow(2.0, -p * d * d);
432 else if (flags & SWS_X) {
433 double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
437 c = cos(floatd * M_PI);
444 coeff = (c * 0.5 + 0.5) * fone;
445 } else if (flags & SWS_AREA) {
446 int64_t d2 = d - (1 << 29);
447 if (d2 * xInc < -(1LL << (29 + 16)))
448 coeff = 1.0 * (1LL << (30 + 16));
449 else if (d2 * xInc < (1LL << (29 + 16)))
450 coeff = -d2 * xInc + (1LL << (29 + 16));
453 coeff *= fone >> (30 + 16);
454 } else if (flags & SWS_GAUSS) {
455 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
456 coeff = (pow(2.0, -p * floatd * floatd)) * fone;
457 } else if (flags & SWS_SINC) {
458 coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
459 } else if (flags & SWS_LANCZOS) {
460 double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
461 coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
462 (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
465 } else if (flags & SWS_BILINEAR) {
466 coeff = (1 << 30) - d;
470 } else if (flags & SWS_SPLINE) {
471 double p = -2.196152422706632;
472 coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
477 filter[i * filterSize + j] = coeff;
480 xDstInSrc += 2 * xInc;
484 /* apply src & dst Filter to filter -> filter2
487 av_assert0(filterSize > 0);
488 filter2Size = filterSize;
490 filter2Size += srcFilter->length - 1;
492 filter2Size += dstFilter->length - 1;
493 av_assert0(filter2Size > 0);
494 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, filter2, dstW, filter2Size * sizeof(*filter2), fail);
496 for (i = 0; i < dstW; i++) {
500 for (k = 0; k < srcFilter->length; k++) {
501 for (j = 0; j < filterSize; j++)
502 filter2[i * filter2Size + k + j] +=
503 srcFilter->coeff[k] * filter[i * filterSize + j];
506 for (j = 0; j < filterSize; j++)
507 filter2[i * filter2Size + j] = filter[i * filterSize + j];
511 (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
515 /* try to reduce the filter-size (step1 find size and shift left) */
516 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
518 for (i = dstW - 1; i >= 0; i--) {
519 int min = filter2Size;
521 int64_t cutOff = 0.0;
523 /* get rid of near zero elements on the left by shifting left */
524 for (j = 0; j < filter2Size; j++) {
526 cutOff += FFABS(filter2[i * filter2Size]);
528 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
531 /* preserve monotonicity because the core can't handle the
532 * filter otherwise */
533 if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
536 // move filter coefficients left
537 for (k = 1; k < filter2Size; k++)
538 filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
539 filter2[i * filter2Size + k - 1] = 0;
544 /* count near zeros on the right */
545 for (j = filter2Size - 1; j > 0; j--) {
546 cutOff += FFABS(filter2[i * filter2Size + j]);
548 if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
553 if (min > minFilterSize)
557 if (PPC_ALTIVEC(cpu_flags)) {
558 // we can handle the special case 4, so we don't want to go the full 8
559 if (minFilterSize < 5)
562 /* We really don't want to waste our time doing useless computation, so
563 * fall back on the scalar C code for very small filters.
564 * Vectorizing is worth it only if you have a decent-sized vector. */
565 if (minFilterSize < 3)
569 if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
570 // special case for unscaled vertical filtering
571 if (minFilterSize == 1 && filterAlign == 2)
575 av_assert0(minFilterSize > 0);
576 filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
577 av_assert0(filterSize > 0);
578 filter = av_malloc_array(dstW, filterSize * sizeof(*filter));
581 if (filterSize >= MAX_FILTER_SIZE * 16 /
582 ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16)) {
583 av_log(NULL, AV_LOG_ERROR, "sws: filterSize %d is too large, try less extreme scaling or set --sws-max-filter-size and recompile\n",
584 FF_CEIL_RSHIFT((filterSize+1) * ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16), 4));
587 *outFilterSize = filterSize;
589 if (flags & SWS_PRINT_INFO)
590 av_log(NULL, AV_LOG_VERBOSE,
591 "SwScaler: reducing / aligning filtersize %d -> %d\n",
592 filter2Size, filterSize);
593 /* try to reduce the filter-size (step2 reduce it) */
594 for (i = 0; i < dstW; i++) {
597 for (j = 0; j < filterSize; j++) {
598 if (j >= filter2Size)
599 filter[i * filterSize + j] = 0;
601 filter[i * filterSize + j] = filter2[i * filter2Size + j];
602 if ((flags & SWS_BITEXACT) && j >= minFilterSize)
603 filter[i * filterSize + j] = 0;
607 // FIXME try to align filterPos if possible
610 for (i = 0; i < dstW; i++) {
612 if ((*filterPos)[i] < 0) {
613 // move filter coefficients left to compensate for filterPos
614 for (j = 1; j < filterSize; j++) {
615 int left = FFMAX(j + (*filterPos)[i], 0);
616 filter[i * filterSize + left] += filter[i * filterSize + j];
617 filter[i * filterSize + j] = 0;
622 if ((*filterPos)[i] + filterSize > srcW) {
623 int shift = (*filterPos)[i] + filterSize - srcW;
624 // move filter coefficients right to compensate for filterPos
625 for (j = filterSize - 2; j >= 0; j--) {
626 int right = FFMIN(j + shift, filterSize - 1);
627 filter[i * filterSize + right] += filter[i * filterSize + j];
628 filter[i * filterSize + j] = 0;
630 (*filterPos)[i]= srcW - filterSize;
634 // Note the +1 is for the MMX scaler which reads over the end
635 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
636 FF_ALLOCZ_ARRAY_OR_GOTO(NULL, *outFilter,
637 (dstW + 3), *outFilterSize * sizeof(int16_t), fail);
639 /* normalize & store in outFilter */
640 for (i = 0; i < dstW; i++) {
645 for (j = 0; j < filterSize; j++) {
646 sum += filter[i * filterSize + j];
648 sum = (sum + one / 2) / one;
650 av_log(NULL, AV_LOG_WARNING, "SwScaler: zero vector in scaling\n");
653 for (j = 0; j < *outFilterSize; j++) {
654 int64_t v = filter[i * filterSize + j] + error;
655 int intV = ROUNDED_DIV(v, sum);
656 (*outFilter)[i * (*outFilterSize) + j] = intV;
657 error = v - intV * sum;
661 (*filterPos)[dstW + 0] =
662 (*filterPos)[dstW + 1] =
663 (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
664 * read over the end */
665 for (i = 0; i < *outFilterSize; i++) {
666 int k = (dstW - 1) * (*outFilterSize) + i;
667 (*outFilter)[k + 1 * (*outFilterSize)] =
668 (*outFilter)[k + 2 * (*outFilterSize)] =
669 (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
676 av_log(NULL, AV_LOG_ERROR, "sws: initFilter failed\n");
682 static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange)
684 int64_t W, V, Z, Cy, Cu, Cv;
685 int64_t vr = table[0];
686 int64_t ub = table[1];
687 int64_t ug = -table[2];
688 int64_t vg = -table[3];
691 uint8_t *p = (uint8_t*)c->input_rgb2yuv_table;
693 static const int8_t map[] = {
694 BY_IDX, GY_IDX, -1 , BY_IDX, BY_IDX, GY_IDX, -1 , BY_IDX,
695 RY_IDX, -1 , GY_IDX, RY_IDX, RY_IDX, -1 , GY_IDX, RY_IDX,
696 RY_IDX, GY_IDX, -1 , RY_IDX, RY_IDX, GY_IDX, -1 , RY_IDX,
697 BY_IDX, -1 , GY_IDX, BY_IDX, BY_IDX, -1 , GY_IDX, BY_IDX,
698 BU_IDX, GU_IDX, -1 , BU_IDX, BU_IDX, GU_IDX, -1 , BU_IDX,
699 RU_IDX, -1 , GU_IDX, RU_IDX, RU_IDX, -1 , GU_IDX, RU_IDX,
700 RU_IDX, GU_IDX, -1 , RU_IDX, RU_IDX, GU_IDX, -1 , RU_IDX,
701 BU_IDX, -1 , GU_IDX, BU_IDX, BU_IDX, -1 , GU_IDX, BU_IDX,
702 BV_IDX, GV_IDX, -1 , BV_IDX, BV_IDX, GV_IDX, -1 , BV_IDX,
703 RV_IDX, -1 , GV_IDX, RV_IDX, RV_IDX, -1 , GV_IDX, RV_IDX,
704 RV_IDX, GV_IDX, -1 , RV_IDX, RV_IDX, GV_IDX, -1 , RV_IDX,
705 BV_IDX, -1 , GV_IDX, BV_IDX, BV_IDX, -1 , GV_IDX, BV_IDX,
706 RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX,
707 BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX,
708 GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 ,
709 -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX, -1 , GY_IDX,
710 RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX,
711 BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX,
712 GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 ,
713 -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX, -1 , GU_IDX,
714 RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX,
715 BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX,
716 GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 ,
717 -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, -1 , GV_IDX, //23
718 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //24
719 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //25
720 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //26
721 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //27
722 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //28
723 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //29
724 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //30
725 -1 , -1 , -1 , -1 , -1 , -1 , -1 , -1 , //31
726 BY_IDX, GY_IDX, RY_IDX, -1 , -1 , -1 , -1 , -1 , //32
727 BU_IDX, GU_IDX, RU_IDX, -1 , -1 , -1 , -1 , -1 , //33
728 BV_IDX, GV_IDX, RV_IDX, -1 , -1 , -1 , -1 , -1 , //34
731 dstRange = 0; //FIXME range = 1 is handled elsewhere
741 W = ROUNDED_DIV(ONE*ONE*ug, ub);
742 V = ROUNDED_DIV(ONE*ONE*vg, vr);
745 Cy = ROUNDED_DIV(cy*Z, ONE);
746 Cu = ROUNDED_DIV(ub*Z, ONE);
747 Cv = ROUNDED_DIV(vr*Z, ONE);
749 c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cy);
750 c->input_rgb2yuv_table[GY_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cy);
751 c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cy);
753 c->input_rgb2yuv_table[RU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V , Cu);
754 c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cu);
755 c->input_rgb2yuv_table[BU_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W) , Cu);
757 c->input_rgb2yuv_table[RV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z) , Cv);
758 c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE , Cv);
759 c->input_rgb2yuv_table[BV_IDX] = ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W , Cv);
761 if(/*!dstRange && */!memcmp(table, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sizeof(ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]))) {
762 c->input_rgb2yuv_table[BY_IDX] = ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
763 c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
764 c->input_rgb2yuv_table[BU_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
765 c->input_rgb2yuv_table[GY_IDX] = ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
766 c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
767 c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
768 c->input_rgb2yuv_table[RY_IDX] = ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
769 c->input_rgb2yuv_table[RV_IDX] = ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
770 c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
772 for(i=0; i<FF_ARRAY_ELEMS(map); i++)
773 AV_WL16(p + 16*4 + 2*i, map[i] >= 0 ? c->input_rgb2yuv_table[map[i]] : 0);
776 static void fill_xyztables(struct SwsContext *c)
779 double xyzgamma = XYZ_GAMMA;
780 double rgbgamma = 1.0 / RGB_GAMMA;
781 double xyzgammainv = 1.0 / XYZ_GAMMA;
782 double rgbgammainv = RGB_GAMMA;
783 static const int16_t xyz2rgb_matrix[3][4] = {
784 {13270, -6295, -2041},
786 { 228, -835, 4329} };
787 static const int16_t rgb2xyz_matrix[3][4] = {
791 static int16_t xyzgamma_tab[4096], rgbgamma_tab[4096], xyzgammainv_tab[4096], rgbgammainv_tab[4096];
793 memcpy(c->xyz2rgb_matrix, xyz2rgb_matrix, sizeof(c->xyz2rgb_matrix));
794 memcpy(c->rgb2xyz_matrix, rgb2xyz_matrix, sizeof(c->rgb2xyz_matrix));
795 c->xyzgamma = xyzgamma_tab;
796 c->rgbgamma = rgbgamma_tab;
797 c->xyzgammainv = xyzgammainv_tab;
798 c->rgbgammainv = rgbgammainv_tab;
800 if (rgbgamma_tab[4095])
803 /* set gamma vectors */
804 for (i = 0; i < 4096; i++) {
805 xyzgamma_tab[i] = lrint(pow(i / 4095.0, xyzgamma) * 4095.0);
806 rgbgamma_tab[i] = lrint(pow(i / 4095.0, rgbgamma) * 4095.0);
807 xyzgammainv_tab[i] = lrint(pow(i / 4095.0, xyzgammainv) * 4095.0);
808 rgbgammainv_tab[i] = lrint(pow(i / 4095.0, rgbgammainv) * 4095.0);
812 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
813 int srcRange, const int table[4], int dstRange,
814 int brightness, int contrast, int saturation)
816 const AVPixFmtDescriptor *desc_dst;
817 const AVPixFmtDescriptor *desc_src;
819 memmove(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
820 memmove(c->dstColorspaceTable, table, sizeof(int) * 4);
823 desc_dst = av_pix_fmt_desc_get(c->dstFormat);
824 desc_src = av_pix_fmt_desc_get(c->srcFormat);
826 if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
828 if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
831 c->brightness = brightness;
832 c->contrast = contrast;
833 c->saturation = saturation;
834 if (c->srcRange != srcRange || c->dstRange != dstRange)
836 c->srcRange = srcRange;
837 c->dstRange = dstRange;
839 //The srcBpc check is possibly wrong but we seem to lack a definitive reference to test this
840 //and what we have in ticket 2939 looks better with this check
841 if (need_reinit && (c->srcBpc == 8 || !isYUV(c->srcFormat)))
842 ff_sws_init_range_convert(c);
844 if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat)))
847 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
848 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
850 if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) {
851 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
852 contrast, saturation);
856 ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
857 contrast, saturation);
860 fill_rgb2yuv_table(c, table, dstRange);
865 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
866 int *srcRange, int **table, int *dstRange,
867 int *brightness, int *contrast, int *saturation)
872 *inv_table = c->srcColorspaceTable;
873 *table = c->dstColorspaceTable;
874 *srcRange = c->srcRange;
875 *dstRange = c->dstRange;
876 *brightness = c->brightness;
877 *contrast = c->contrast;
878 *saturation = c->saturation;
883 static int handle_jpeg(enum AVPixelFormat *format)
886 case AV_PIX_FMT_YUVJ420P:
887 *format = AV_PIX_FMT_YUV420P;
889 case AV_PIX_FMT_YUVJ411P:
890 *format = AV_PIX_FMT_YUV411P;
892 case AV_PIX_FMT_YUVJ422P:
893 *format = AV_PIX_FMT_YUV422P;
895 case AV_PIX_FMT_YUVJ444P:
896 *format = AV_PIX_FMT_YUV444P;
898 case AV_PIX_FMT_YUVJ440P:
899 *format = AV_PIX_FMT_YUV440P;
901 case AV_PIX_FMT_GRAY8:
902 case AV_PIX_FMT_GRAY16LE:
903 case AV_PIX_FMT_GRAY16BE:
910 static int handle_0alpha(enum AVPixelFormat *format)
913 case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
914 case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
915 case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
916 case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4;
921 static int handle_xyz(enum AVPixelFormat *format)
924 case AV_PIX_FMT_XYZ12BE : *format = AV_PIX_FMT_RGB48BE; return 1;
925 case AV_PIX_FMT_XYZ12LE : *format = AV_PIX_FMT_RGB48LE; return 1;
930 static void handle_formats(SwsContext *c)
932 c->src0Alpha |= handle_0alpha(&c->srcFormat);
933 c->dst0Alpha |= handle_0alpha(&c->dstFormat);
934 c->srcXYZ |= handle_xyz(&c->srcFormat);
935 c->dstXYZ |= handle_xyz(&c->dstFormat);
936 if (c->srcXYZ || c->dstXYZ)
940 SwsContext *sws_alloc_context(void)
942 SwsContext *c = av_mallocz(sizeof(SwsContext));
944 av_assert0(offsetof(SwsContext, redDither) + DITHER32_INT == offsetof(SwsContext, dither32));
947 c->av_class = &sws_context_class;
948 av_opt_set_defaults(c);
954 av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
955 SwsFilter *dstFilter)
958 int usesVFilter, usesHFilter;
960 SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
965 int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
966 int flags, cpu_flags;
967 enum AVPixelFormat srcFormat = c->srcFormat;
968 enum AVPixelFormat dstFormat = c->dstFormat;
969 const AVPixFmtDescriptor *desc_src;
970 const AVPixFmtDescriptor *desc_dst;
972 cpu_flags = av_get_cpu_flags();
978 unscaled = (srcW == dstW && srcH == dstH);
980 c->srcRange |= handle_jpeg(&c->srcFormat);
981 c->dstRange |= handle_jpeg(&c->dstFormat);
983 if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
984 av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
986 if (!c->contrast && !c->saturation && !c->dstFormatBpp)
987 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
988 ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
989 c->dstRange, 0, 1 << 16, 1 << 16);
992 srcFormat = c->srcFormat;
993 dstFormat = c->dstFormat;
994 desc_src = av_pix_fmt_desc_get(srcFormat);
995 desc_dst = av_pix_fmt_desc_get(dstFormat);
997 if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
998 av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
999 if (!sws_isSupportedInput(srcFormat)) {
1000 av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
1001 av_get_pix_fmt_name(srcFormat));
1002 return AVERROR(EINVAL);
1004 if (!sws_isSupportedOutput(dstFormat)) {
1005 av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
1006 av_get_pix_fmt_name(dstFormat));
1007 return AVERROR(EINVAL);
1011 i = flags & (SWS_POINT |
1023 /* provide a default scaler if not set by caller */
1025 if (dstW < srcW && dstH < srcH)
1026 flags |= SWS_BICUBIC;
1027 else if (dstW > srcW && dstH > srcH)
1028 flags |= SWS_BICUBIC;
1030 flags |= SWS_BICUBIC;
1032 } else if (i & (i - 1)) {
1033 av_log(c, AV_LOG_ERROR,
1034 "Exactly one scaler algorithm must be chosen, got %X\n", i);
1035 return AVERROR(EINVAL);
1038 if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
1039 /* FIXME check if these are enough and try to lower them after
1040 * fixing the relevant parts of the code */
1041 av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
1042 srcW, srcH, dstW, dstH);
1043 return AVERROR(EINVAL);
1047 dstFilter = &dummyFilter;
1049 srcFilter = &dummyFilter;
1051 c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
1052 c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
1053 c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
1054 c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
1055 c->vRounder = 4 * 0x0001000100010001ULL;
1057 usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
1058 (srcFilter->chrV && srcFilter->chrV->length > 1) ||
1059 (dstFilter->lumV && dstFilter->lumV->length > 1) ||
1060 (dstFilter->chrV && dstFilter->chrV->length > 1);
1061 usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
1062 (srcFilter->chrH && srcFilter->chrH->length > 1) ||
1063 (dstFilter->lumH && dstFilter->lumH->length > 1) ||
1064 (dstFilter->chrH && dstFilter->chrH->length > 1);
1066 av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
1067 av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);
1069 if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
1071 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
1072 flags |= SWS_FULL_CHR_H_INT;
1076 if ( c->chrSrcHSubSample == 0
1077 && c->chrSrcVSubSample == 0
1078 && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
1079 && !(c->flags & SWS_FAST_BILINEAR)
1081 av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
1082 flags |= SWS_FULL_CHR_H_INT;
1087 if (c->dither == SWS_DITHER_AUTO) {
1088 if (flags & SWS_ERROR_DIFFUSION)
1089 c->dither = SWS_DITHER_ED;
1092 if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1093 dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1094 dstFormat == AV_PIX_FMT_BGR8 ||
1095 dstFormat == AV_PIX_FMT_RGB8) {
1096 if (c->dither == SWS_DITHER_AUTO)
1097 c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
1098 if (!(flags & SWS_FULL_CHR_H_INT)) {
1099 if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER) {
1100 av_log(c, AV_LOG_DEBUG,
1101 "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
1102 av_get_pix_fmt_name(dstFormat));
1103 flags |= SWS_FULL_CHR_H_INT;
1107 if (flags & SWS_FULL_CHR_H_INT) {
1108 if (c->dither == SWS_DITHER_BAYER) {
1109 av_log(c, AV_LOG_DEBUG,
1110 "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1111 av_get_pix_fmt_name(dstFormat));
1112 c->dither = SWS_DITHER_ED;
1116 if (isPlanarRGB(dstFormat)) {
1117 if (!(flags & SWS_FULL_CHR_H_INT)) {
1118 av_log(c, AV_LOG_DEBUG,
1119 "%s output is not supported with half chroma resolution, switching to full\n",
1120 av_get_pix_fmt_name(dstFormat));
1121 flags |= SWS_FULL_CHR_H_INT;
1126 /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1127 * chroma interpolation */
1128 if (flags & SWS_FULL_CHR_H_INT &&
1129 isAnyRGB(dstFormat) &&
1130 !isPlanarRGB(dstFormat) &&
1131 dstFormat != AV_PIX_FMT_RGBA &&
1132 dstFormat != AV_PIX_FMT_ARGB &&
1133 dstFormat != AV_PIX_FMT_BGRA &&
1134 dstFormat != AV_PIX_FMT_ABGR &&
1135 dstFormat != AV_PIX_FMT_RGB24 &&
1136 dstFormat != AV_PIX_FMT_BGR24 &&
1137 dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1138 dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1139 dstFormat != AV_PIX_FMT_BGR8 &&
1140 dstFormat != AV_PIX_FMT_RGB8
1142 av_log(c, AV_LOG_WARNING,
1143 "full chroma interpolation for destination format '%s' not yet implemented\n",
1144 av_get_pix_fmt_name(dstFormat));
1145 flags &= ~SWS_FULL_CHR_H_INT;
1148 if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1149 c->chrDstHSubSample = 1;
1151 // drop some chroma lines if the user wants it
1152 c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1153 SWS_SRC_V_CHR_DROP_SHIFT;
1154 c->chrSrcVSubSample += c->vChrDrop;
1156 /* drop every other pixel for chroma calculation unless user
1157 * wants full chroma */
1158 if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) &&
1159 srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1160 srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1161 srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1162 srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
1163 srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
1164 srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
1165 srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE &&
1166 srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
1167 ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1168 (flags & SWS_FAST_BILINEAR)))
1169 c->chrSrcHSubSample = 1;
1171 // Note the FF_CEIL_RSHIFT is so that we always round toward +inf.
1172 c->chrSrcW = FF_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
1173 c->chrSrcH = FF_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
1174 c->chrDstW = FF_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
1175 c->chrDstH = FF_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
1177 FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
1179 c->srcBpc = 1 + desc_src->comp[0].depth_minus1;
1182 c->dstBpc = 1 + desc_dst->comp[0].depth_minus1;
1185 if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1187 if (c->dstBpc == 16)
1190 if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1191 c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1192 c->chrDstW >= c->chrSrcW &&
1194 if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1196 && (flags & SWS_FAST_BILINEAR)) {
1197 if (flags & SWS_PRINT_INFO)
1198 av_log(c, AV_LOG_INFO,
1199 "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1201 if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1202 c->canMMXEXTBeUsed = 0;
1204 c->canMMXEXTBeUsed = 0;
1206 c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1207 c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1209 /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1210 * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1212 * n-2 is the last chrominance sample available.
1213 * This is not perfect, but no one should notice the difference, the more
1214 * correct variant would be like the vertical one, but that would require
1215 * some special code for the first and last pixel */
1216 if (flags & SWS_FAST_BILINEAR) {
1217 if (c->canMMXEXTBeUsed) {
1221 // we don't use the x86 asm scaler if MMX is available
1222 else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1223 c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
1224 c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1228 #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
1230 /* precalculate horizontal scaler filter coefficients */
1232 #if HAVE_MMXEXT_INLINE
1233 // can't downscale !!!
1234 if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1235 c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1237 c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1238 NULL, NULL, NULL, 4);
1241 c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
1242 PROT_READ | PROT_WRITE,
1243 MAP_PRIVATE | MAP_ANONYMOUS,
1245 c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
1246 PROT_READ | PROT_WRITE,
1247 MAP_PRIVATE | MAP_ANONYMOUS,
1249 #elif HAVE_VIRTUALALLOC
1250 c->lumMmxextFilterCode = VirtualAlloc(NULL,
1251 c->lumMmxextFilterCodeSize,
1253 PAGE_EXECUTE_READWRITE);
1254 c->chrMmxextFilterCode = VirtualAlloc(NULL,
1255 c->chrMmxextFilterCodeSize,
1257 PAGE_EXECUTE_READWRITE);
1259 c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
1260 c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
1263 #ifdef MAP_ANONYMOUS
1264 if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1266 if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
1269 av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1270 return AVERROR(ENOMEM);
1273 FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
1274 FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
1275 FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
1276 FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);
1278 ff_init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
1279 c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1280 ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1281 c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1284 if ( mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
1285 || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
1286 av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
1291 #endif /* HAVE_MMXEXT_INLINE */
1293 const int filterAlign = X86_MMX(cpu_flags) ? 4 :
1294 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1296 if (initFilter(&c->hLumFilter, &c->hLumFilterPos,
1297 &c->hLumFilterSize, c->lumXInc,
1298 srcW, dstW, filterAlign, 1 << 14,
1299 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1300 cpu_flags, srcFilter->lumH, dstFilter->lumH,
1302 get_local_pos(c, 0, 0, 0),
1303 get_local_pos(c, 0, 0, 0)) < 0)
1305 if (initFilter(&c->hChrFilter, &c->hChrFilterPos,
1306 &c->hChrFilterSize, c->chrXInc,
1307 c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1308 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1309 cpu_flags, srcFilter->chrH, dstFilter->chrH,
1311 get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
1312 get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0)) < 0)
1315 } // initialize horizontal stuff
1317 /* precalculate vertical scaler filter coefficients */
1319 const int filterAlign = X86_MMX(cpu_flags) ? 2 :
1320 PPC_ALTIVEC(cpu_flags) ? 8 : 1;
1322 if (initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
1323 c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1324 (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1325 cpu_flags, srcFilter->lumV, dstFilter->lumV,
1327 get_local_pos(c, 0, 0, 1),
1328 get_local_pos(c, 0, 0, 1)) < 0)
1330 if (initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
1331 c->chrYInc, c->chrSrcH, c->chrDstH,
1332 filterAlign, (1 << 12),
1333 (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1334 cpu_flags, srcFilter->chrV, dstFilter->chrV,
1336 get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
1337 get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1)) < 0)
1342 FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
1343 FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);
1345 for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1347 short *p = (short *)&c->vYCoeffsBank[i];
1348 for (j = 0; j < 8; j++)
1349 p[j] = c->vLumFilter[i];
1352 for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1354 short *p = (short *)&c->vCCoeffsBank[i];
1355 for (j = 0; j < 8; j++)
1356 p[j] = c->vChrFilter[i];
1361 // calculate buffer sizes so that they won't run out while handling these damn slices
1362 c->vLumBufSize = c->vLumFilterSize;
1363 c->vChrBufSize = c->vChrFilterSize;
1364 for (i = 0; i < dstH; i++) {
1365 int chrI = (int64_t)i * c->chrDstH / dstH;
1366 int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
1367 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
1368 << c->chrSrcVSubSample));
1370 nextSlice >>= c->chrSrcVSubSample;
1371 nextSlice <<= c->chrSrcVSubSample;
1372 if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
1373 c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
1374 if (c->vChrFilterPos[chrI] + c->vChrBufSize <
1375 (nextSlice >> c->chrSrcVSubSample))
1376 c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
1377 c->vChrFilterPos[chrI];
1380 for (i = 0; i < 4; i++)
1381 FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);
1383 /* Allocate pixbufs (we use dynamic allocation because otherwise we would
1384 * need to allocate several megabytes to handle all possible cases) */
1385 FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1386 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1387 FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1388 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
1389 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1390 /* Note we need at least one pixel more at the end because of the MMX code
1391 * (just in case someone wants to replace the 4000/8000). */
1392 /* align at 16 bytes for AltiVec */
1393 for (i = 0; i < c->vLumBufSize; i++) {
1394 FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
1395 dst_stride + 16, fail);
1396 c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
1398 // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1399 c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1400 c->uv_offx2 = dst_stride + 16;
1401 for (i = 0; i < c->vChrBufSize; i++) {
1402 FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
1403 dst_stride * 2 + 32, fail);
1404 c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
1405 c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
1406 = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
1408 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1409 for (i = 0; i < c->vLumBufSize; i++) {
1410 FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
1411 dst_stride + 16, fail);
1412 c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
1415 // try to avoid drawing green stuff between the right end and the stride end
1416 for (i = 0; i < c->vChrBufSize; i++)
1417 if(desc_dst->comp[0].depth_minus1 == 15){
1418 av_assert0(c->dstBpc > 14);
1419 for(j=0; j<dst_stride/2+1; j++)
1420 ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
1422 for(j=0; j<dst_stride+1; j++)
1423 ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
1425 av_assert0(c->chrDstH <= dstH);
1427 if (flags & SWS_PRINT_INFO) {
1428 const char *scaler = NULL, *cpucaps;
1430 for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
1431 if (flags & scale_algorithms[i].flag) {
1432 scaler = scale_algorithms[i].description;
1437 scaler = "ehh flags invalid?!";
1438 av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
1440 av_get_pix_fmt_name(srcFormat),
1442 dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
1443 dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1444 dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1449 av_get_pix_fmt_name(dstFormat));
1451 if (INLINE_MMXEXT(cpu_flags))
1453 else if (INLINE_AMD3DNOW(cpu_flags))
1455 else if (INLINE_MMX(cpu_flags))
1457 else if (PPC_ALTIVEC(cpu_flags))
1458 cpucaps = "AltiVec";
1462 av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);
1464 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1465 av_log(c, AV_LOG_DEBUG,
1466 "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1467 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1468 av_log(c, AV_LOG_DEBUG,
1469 "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1470 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1471 c->chrXInc, c->chrYInc);
1474 /* unscaled special cases */
1475 if (unscaled && !usesHFilter && !usesVFilter &&
1476 (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
1477 ff_get_unscaled_swscale(c);
1480 if (flags & SWS_PRINT_INFO)
1481 av_log(c, AV_LOG_INFO,
1482 "using unscaled %s -> %s special converter\n",
1483 av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1488 c->swscale = ff_getSwsFunc(c);
1490 fail: // FIXME replace things by appropriate error codes
1494 #if FF_API_SWS_GETCONTEXT
1495 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
1496 int dstW, int dstH, enum AVPixelFormat dstFormat,
1497 int flags, SwsFilter *srcFilter,
1498 SwsFilter *dstFilter, const double *param)
1502 if (!(c = sws_alloc_context()))
1510 c->srcFormat = srcFormat;
1511 c->dstFormat = dstFormat;
1514 c->param[0] = param[0];
1515 c->param[1] = param[1];
1518 if (sws_init_context(c, srcFilter, dstFilter) < 0) {
1527 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
1528 float lumaSharpen, float chromaSharpen,
1529 float chromaHShift, float chromaVShift,
1532 SwsFilter *filter = av_malloc(sizeof(SwsFilter));
1536 if (lumaGBlur != 0.0) {
1537 filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
1538 filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
1540 filter->lumH = sws_getIdentityVec();
1541 filter->lumV = sws_getIdentityVec();
1544 if (chromaGBlur != 0.0) {
1545 filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
1546 filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
1548 filter->chrH = sws_getIdentityVec();
1549 filter->chrV = sws_getIdentityVec();
1552 if (chromaSharpen != 0.0) {
1553 SwsVector *id = sws_getIdentityVec();
1554 sws_scaleVec(filter->chrH, -chromaSharpen);
1555 sws_scaleVec(filter->chrV, -chromaSharpen);
1556 sws_addVec(filter->chrH, id);
1557 sws_addVec(filter->chrV, id);
1561 if (lumaSharpen != 0.0) {
1562 SwsVector *id = sws_getIdentityVec();
1563 sws_scaleVec(filter->lumH, -lumaSharpen);
1564 sws_scaleVec(filter->lumV, -lumaSharpen);
1565 sws_addVec(filter->lumH, id);
1566 sws_addVec(filter->lumV, id);
1570 if (chromaHShift != 0.0)
1571 sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
1573 if (chromaVShift != 0.0)
1574 sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
1576 sws_normalizeVec(filter->chrH, 1.0);
1577 sws_normalizeVec(filter->chrV, 1.0);
1578 sws_normalizeVec(filter->lumH, 1.0);
1579 sws_normalizeVec(filter->lumV, 1.0);
1582 sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
1584 sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
1589 SwsVector *sws_allocVec(int length)
1593 if(length <= 0 || length > INT_MAX/ sizeof(double))
1596 vec = av_malloc(sizeof(SwsVector));
1599 vec->length = length;
1600 vec->coeff = av_malloc(sizeof(double) * length);
1606 SwsVector *sws_getGaussianVec(double variance, double quality)
1608 const int length = (int)(variance * quality + 0.5) | 1;
1610 double middle = (length - 1) * 0.5;
1613 if(variance < 0 || quality < 0)
1616 vec = sws_allocVec(length);
1621 for (i = 0; i < length; i++) {
1622 double dist = i - middle;
1623 vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
1624 sqrt(2 * variance * M_PI);
1627 sws_normalizeVec(vec, 1.0);
1632 SwsVector *sws_getConstVec(double c, int length)
1635 SwsVector *vec = sws_allocVec(length);
1640 for (i = 0; i < length; i++)
1646 SwsVector *sws_getIdentityVec(void)
1648 return sws_getConstVec(1.0, 1);
1651 static double sws_dcVec(SwsVector *a)
1656 for (i = 0; i < a->length; i++)
1662 void sws_scaleVec(SwsVector *a, double scalar)
1666 for (i = 0; i < a->length; i++)
1667 a->coeff[i] *= scalar;
1670 void sws_normalizeVec(SwsVector *a, double height)
1672 sws_scaleVec(a, height / sws_dcVec(a));
1675 static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b)
1677 int length = a->length + b->length - 1;
1679 SwsVector *vec = sws_getConstVec(0.0, length);
1684 for (i = 0; i < a->length; i++) {
1685 for (j = 0; j < b->length; j++) {
1686 vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
1693 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
1695 int length = FFMAX(a->length, b->length);
1697 SwsVector *vec = sws_getConstVec(0.0, length);
1702 for (i = 0; i < a->length; i++)
1703 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1704 for (i = 0; i < b->length; i++)
1705 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
1710 static SwsVector *sws_diffVec(SwsVector *a, SwsVector *b)
1712 int length = FFMAX(a->length, b->length);
1714 SwsVector *vec = sws_getConstVec(0.0, length);
1719 for (i = 0; i < a->length; i++)
1720 vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1721 for (i = 0; i < b->length; i++)
1722 vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];
1727 /* shift left / or right if "shift" is negative */
1728 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
1730 int length = a->length + FFABS(shift) * 2;
1732 SwsVector *vec = sws_getConstVec(0.0, length);
1737 for (i = 0; i < a->length; i++) {
1738 vec->coeff[i + (length - 1) / 2 -
1739 (a->length - 1) / 2 - shift] = a->coeff[i];
1745 void sws_shiftVec(SwsVector *a, int shift)
1747 SwsVector *shifted = sws_getShiftedVec(a, shift);
1749 a->coeff = shifted->coeff;
1750 a->length = shifted->length;
1754 void sws_addVec(SwsVector *a, SwsVector *b)
1756 SwsVector *sum = sws_sumVec(a, b);
1758 a->coeff = sum->coeff;
1759 a->length = sum->length;
1763 void sws_subVec(SwsVector *a, SwsVector *b)
1765 SwsVector *diff = sws_diffVec(a, b);
1767 a->coeff = diff->coeff;
1768 a->length = diff->length;
1772 void sws_convVec(SwsVector *a, SwsVector *b)
1774 SwsVector *conv = sws_getConvVec(a, b);
1776 a->coeff = conv->coeff;
1777 a->length = conv->length;
1781 SwsVector *sws_cloneVec(SwsVector *a)
1783 SwsVector *vec = sws_allocVec(a->length);
1788 memcpy(vec->coeff, a->coeff, a->length * sizeof(*a->coeff));
1793 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
1800 for (i = 0; i < a->length; i++)
1801 if (a->coeff[i] > max)
1804 for (i = 0; i < a->length; i++)
1805 if (a->coeff[i] < min)
1810 for (i = 0; i < a->length; i++) {
1811 int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
1812 av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
1814 av_log(log_ctx, log_level, " ");
1815 av_log(log_ctx, log_level, "|\n");
1819 void sws_freeVec(SwsVector *a)
1823 av_freep(&a->coeff);
1828 void sws_freeFilter(SwsFilter *filter)
1833 sws_freeVec(filter->lumH);
1834 sws_freeVec(filter->lumV);
1835 sws_freeVec(filter->chrH);
1836 sws_freeVec(filter->chrV);
1840 void sws_freeContext(SwsContext *c)
1847 for (i = 0; i < c->vLumBufSize; i++)
1848 av_freep(&c->lumPixBuf[i]);
1849 av_freep(&c->lumPixBuf);
1852 if (c->chrUPixBuf) {
1853 for (i = 0; i < c->vChrBufSize; i++)
1854 av_freep(&c->chrUPixBuf[i]);
1855 av_freep(&c->chrUPixBuf);
1856 av_freep(&c->chrVPixBuf);
1859 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1860 for (i = 0; i < c->vLumBufSize; i++)
1861 av_freep(&c->alpPixBuf[i]);
1862 av_freep(&c->alpPixBuf);
1865 for (i = 0; i < 4; i++)
1866 av_freep(&c->dither_error[i]);
1868 av_freep(&c->vLumFilter);
1869 av_freep(&c->vChrFilter);
1870 av_freep(&c->hLumFilter);
1871 av_freep(&c->hChrFilter);
1873 av_freep(&c->vYCoeffsBank);
1874 av_freep(&c->vCCoeffsBank);
1877 av_freep(&c->vLumFilterPos);
1878 av_freep(&c->vChrFilterPos);
1879 av_freep(&c->hLumFilterPos);
1880 av_freep(&c->hChrFilterPos);
1884 if (c->lumMmxextFilterCode)
1885 munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
1886 if (c->chrMmxextFilterCode)
1887 munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize);
1888 #elif HAVE_VIRTUALALLOC
1889 if (c->lumMmxextFilterCode)
1890 VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
1891 if (c->chrMmxextFilterCode)
1892 VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
1894 av_free(c->lumMmxextFilterCode);
1895 av_free(c->chrMmxextFilterCode);
1897 c->lumMmxextFilterCode = NULL;
1898 c->chrMmxextFilterCode = NULL;
1899 #endif /* HAVE_MMX_INLINE */
1901 av_freep(&c->yuvTable);
1902 av_freep(&c->formatConvBuffer);
1907 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
1908 int srcH, enum AVPixelFormat srcFormat,
1910 enum AVPixelFormat dstFormat, int flags,
1911 SwsFilter *srcFilter,
1912 SwsFilter *dstFilter,
1913 const double *param)
1915 static const double default_param[2] = { SWS_PARAM_DEFAULT,
1916 SWS_PARAM_DEFAULT };
1919 param = default_param;
1922 (context->srcW != srcW ||
1923 context->srcH != srcH ||
1924 context->srcFormat != srcFormat ||
1925 context->dstW != dstW ||
1926 context->dstH != dstH ||
1927 context->dstFormat != dstFormat ||
1928 context->flags != flags ||
1929 context->param[0] != param[0] ||
1930 context->param[1] != param[1])) {
1931 sws_freeContext(context);
1936 if (!(context = sws_alloc_context()))
1938 context->srcW = srcW;
1939 context->srcH = srcH;
1940 context->srcFormat = srcFormat;
1941 context->dstW = dstW;
1942 context->dstH = dstH;
1943 context->dstFormat = dstFormat;
1944 context->flags = flags;
1945 context->param[0] = param[0];
1946 context->param[1] = param[1];
1947 if (sws_init_context(context, srcFilter, dstFilter) < 0) {
1948 sws_freeContext(context);