• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
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.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
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
19  */
20 
21 #include "config.h"
22 
23 #define _DEFAULT_SOURCE
24 #define _SVID_SOURCE // needed for MAP_ANONYMOUS
25 #define _DARWIN_C_SOURCE // needed for MAP_ANON
26 #include <inttypes.h>
27 #include <math.h>
28 #include <stdio.h>
29 #include <string.h>
30 #if HAVE_MMAP
31 #include <sys/mman.h>
32 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
33 #define MAP_ANONYMOUS MAP_ANON
34 #endif
35 #endif
36 #if HAVE_VIRTUALALLOC
37 #define WIN32_LEAN_AND_MEAN
38 #include <windows.h>
39 #endif
40 
41 #include "libavutil/attributes.h"
42 #include "libavutil/avassert.h"
43 #include "libavutil/cpu.h"
44 #include "libavutil/imgutils.h"
45 #include "libavutil/intreadwrite.h"
46 #include "libavutil/libm.h"
47 #include "libavutil/mathematics.h"
48 #include "libavutil/opt.h"
49 #include "libavutil/pixdesc.h"
50 #include "libavutil/slicethread.h"
51 #include "libavutil/thread.h"
52 #include "libavutil/aarch64/cpu.h"
53 #include "libavutil/ppc/cpu.h"
54 #include "libavutil/x86/asm.h"
55 #include "libavutil/x86/cpu.h"
56 
57 #include "rgb2rgb.h"
58 #include "swscale.h"
59 #include "swscale_internal.h"
60 
61 static SwsVector *sws_getIdentityVec(void);
62 static void sws_addVec(SwsVector *a, SwsVector *b);
63 static void sws_shiftVec(SwsVector *a, int shift);
64 static void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level);
65 
66 static void handle_formats(SwsContext *c);
67 
68 typedef struct FormatEntry {
69     uint8_t is_supported_in         :1;
70     uint8_t is_supported_out        :1;
71     uint8_t is_supported_endianness :1;
72 } FormatEntry;
73 
74 static const FormatEntry format_entries[] = {
75     [AV_PIX_FMT_YUV420P]     = { 1, 1 },
76     [AV_PIX_FMT_YUYV422]     = { 1, 1 },
77     [AV_PIX_FMT_RGB24]       = { 1, 1 },
78     [AV_PIX_FMT_BGR24]       = { 1, 1 },
79     [AV_PIX_FMT_YUV422P]     = { 1, 1 },
80     [AV_PIX_FMT_YUV444P]     = { 1, 1 },
81     [AV_PIX_FMT_YUV410P]     = { 1, 1 },
82     [AV_PIX_FMT_YUV411P]     = { 1, 1 },
83     [AV_PIX_FMT_GRAY8]       = { 1, 1 },
84     [AV_PIX_FMT_MONOWHITE]   = { 1, 1 },
85     [AV_PIX_FMT_MONOBLACK]   = { 1, 1 },
86     [AV_PIX_FMT_PAL8]        = { 1, 0 },
87     [AV_PIX_FMT_YUVJ420P]    = { 1, 1 },
88     [AV_PIX_FMT_YUVJ411P]    = { 1, 1 },
89     [AV_PIX_FMT_YUVJ422P]    = { 1, 1 },
90     [AV_PIX_FMT_YUVJ444P]    = { 1, 1 },
91     [AV_PIX_FMT_YVYU422]     = { 1, 1 },
92     [AV_PIX_FMT_UYVY422]     = { 1, 1 },
93     [AV_PIX_FMT_UYYVYY411]   = { 0, 0 },
94     [AV_PIX_FMT_BGR8]        = { 1, 1 },
95     [AV_PIX_FMT_BGR4]        = { 0, 1 },
96     [AV_PIX_FMT_BGR4_BYTE]   = { 1, 1 },
97     [AV_PIX_FMT_RGB8]        = { 1, 1 },
98     [AV_PIX_FMT_RGB4]        = { 0, 1 },
99     [AV_PIX_FMT_RGB4_BYTE]   = { 1, 1 },
100     [AV_PIX_FMT_NV12]        = { 1, 1 },
101     [AV_PIX_FMT_NV21]        = { 1, 1 },
102     [AV_PIX_FMT_ARGB]        = { 1, 1 },
103     [AV_PIX_FMT_RGBA]        = { 1, 1 },
104     [AV_PIX_FMT_ABGR]        = { 1, 1 },
105     [AV_PIX_FMT_BGRA]        = { 1, 1 },
106     [AV_PIX_FMT_0RGB]        = { 1, 1 },
107     [AV_PIX_FMT_RGB0]        = { 1, 1 },
108     [AV_PIX_FMT_0BGR]        = { 1, 1 },
109     [AV_PIX_FMT_BGR0]        = { 1, 1 },
110     [AV_PIX_FMT_GRAY9BE]     = { 1, 1 },
111     [AV_PIX_FMT_GRAY9LE]     = { 1, 1 },
112     [AV_PIX_FMT_GRAY10BE]    = { 1, 1 },
113     [AV_PIX_FMT_GRAY10LE]    = { 1, 1 },
114     [AV_PIX_FMT_GRAY12BE]    = { 1, 1 },
115     [AV_PIX_FMT_GRAY12LE]    = { 1, 1 },
116     [AV_PIX_FMT_GRAY14BE]    = { 1, 1 },
117     [AV_PIX_FMT_GRAY14LE]    = { 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, 1 },
170     [AV_PIX_FMT_YA16BE]      = { 1, 1 },
171     [AV_PIX_FMT_YA16LE]      = { 1, 1 },
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_GBRAP10LE]   = { 1, 1 },
206     [AV_PIX_FMT_GBRAP10BE]   = { 1, 1 },
207     [AV_PIX_FMT_GBRP12LE]    = { 1, 1 },
208     [AV_PIX_FMT_GBRP12BE]    = { 1, 1 },
209     [AV_PIX_FMT_GBRAP12LE]   = { 1, 1 },
210     [AV_PIX_FMT_GBRAP12BE]   = { 1, 1 },
211     [AV_PIX_FMT_GBRP14LE]    = { 1, 1 },
212     [AV_PIX_FMT_GBRP14BE]    = { 1, 1 },
213     [AV_PIX_FMT_GBRP16LE]    = { 1, 1 },
214     [AV_PIX_FMT_GBRP16BE]    = { 1, 1 },
215     [AV_PIX_FMT_GBRPF32LE]   = { 1, 1 },
216     [AV_PIX_FMT_GBRPF32BE]   = { 1, 1 },
217     [AV_PIX_FMT_GBRAPF32LE]  = { 1, 1 },
218     [AV_PIX_FMT_GBRAPF32BE]  = { 1, 1 },
219     [AV_PIX_FMT_GBRAP]       = { 1, 1 },
220     [AV_PIX_FMT_GBRAP16LE]   = { 1, 1 },
221     [AV_PIX_FMT_GBRAP16BE]   = { 1, 1 },
222     [AV_PIX_FMT_BAYER_BGGR8] = { 1, 0 },
223     [AV_PIX_FMT_BAYER_RGGB8] = { 1, 0 },
224     [AV_PIX_FMT_BAYER_GBRG8] = { 1, 0 },
225     [AV_PIX_FMT_BAYER_GRBG8] = { 1, 0 },
226     [AV_PIX_FMT_BAYER_BGGR16LE] = { 1, 0 },
227     [AV_PIX_FMT_BAYER_BGGR16BE] = { 1, 0 },
228     [AV_PIX_FMT_BAYER_RGGB16LE] = { 1, 0 },
229     [AV_PIX_FMT_BAYER_RGGB16BE] = { 1, 0 },
230     [AV_PIX_FMT_BAYER_GBRG16LE] = { 1, 0 },
231     [AV_PIX_FMT_BAYER_GBRG16BE] = { 1, 0 },
232     [AV_PIX_FMT_BAYER_GRBG16LE] = { 1, 0 },
233     [AV_PIX_FMT_BAYER_GRBG16BE] = { 1, 0 },
234     [AV_PIX_FMT_XYZ12BE]     = { 1, 1, 1 },
235     [AV_PIX_FMT_XYZ12LE]     = { 1, 1, 1 },
236     [AV_PIX_FMT_AYUV64LE]    = { 1, 1},
237     [AV_PIX_FMT_P010LE]      = { 1, 1 },
238     [AV_PIX_FMT_P010BE]      = { 1, 1 },
239     [AV_PIX_FMT_P016LE]      = { 1, 1 },
240     [AV_PIX_FMT_P016BE]      = { 1, 1 },
241     [AV_PIX_FMT_GRAYF32LE]   = { 1, 1 },
242     [AV_PIX_FMT_GRAYF32BE]   = { 1, 1 },
243     [AV_PIX_FMT_YUVA422P12BE] = { 1, 1 },
244     [AV_PIX_FMT_YUVA422P12LE] = { 1, 1 },
245     [AV_PIX_FMT_YUVA444P12BE] = { 1, 1 },
246     [AV_PIX_FMT_YUVA444P12LE] = { 1, 1 },
247     [AV_PIX_FMT_NV24]        = { 1, 1 },
248     [AV_PIX_FMT_NV42]        = { 1, 1 },
249     [AV_PIX_FMT_Y210LE]      = { 1, 0 },
250     [AV_PIX_FMT_X2RGB10LE]   = { 1, 1 },
251     [AV_PIX_FMT_X2BGR10LE]   = { 1, 1 },
252     [AV_PIX_FMT_P210BE]      = { 1, 1 },
253     [AV_PIX_FMT_P210LE]      = { 1, 1 },
254     [AV_PIX_FMT_P410BE]      = { 1, 1 },
255     [AV_PIX_FMT_P410LE]      = { 1, 1 },
256     [AV_PIX_FMT_P216BE]      = { 1, 1 },
257     [AV_PIX_FMT_P216LE]      = { 1, 1 },
258     [AV_PIX_FMT_P416BE]      = { 1, 1 },
259     [AV_PIX_FMT_P416LE]      = { 1, 1 },
260 };
261 
ff_shuffle_filter_coefficients(SwsContext * c,int * filterPos,int filterSize,int16_t * filter,int dstW)262 int ff_shuffle_filter_coefficients(SwsContext *c, int *filterPos,
263                                    int filterSize, int16_t *filter,
264                                    int dstW)
265 {
266 #if ARCH_X86_64
267     int i, j, k;
268     int cpu_flags = av_get_cpu_flags();
269     // avx2 hscale filter processes 16 pixel blocks.
270     if (!filter || dstW % 16 != 0)
271         return 0;
272     if (EXTERNAL_AVX2_FAST(cpu_flags) && !(cpu_flags & AV_CPU_FLAG_SLOW_GATHER)) {
273         if ((c->srcBpc == 8) && (c->dstBpc <= 14)) {
274            int16_t *filterCopy = NULL;
275            if (filterSize > 4) {
276                if (!FF_ALLOC_TYPED_ARRAY(filterCopy, dstW * filterSize))
277                    return AVERROR(ENOMEM);
278                memcpy(filterCopy, filter, dstW * filterSize * sizeof(int16_t));
279            }
280            // Do not swap filterPos for pixels which won't be processed by
281            // the main loop.
282            for (i = 0; i + 8 <= dstW; i += 8) {
283                FFSWAP(int, filterPos[i + 2], filterPos[i + 4]);
284                FFSWAP(int, filterPos[i + 3], filterPos[i + 5]);
285            }
286            if (filterSize > 4) {
287                // 16 pixels are processed at a time.
288                for (i = 0; i + 16 <= dstW; i += 16) {
289                    // 4 filter coeffs are processed at a time.
290                    for (k = 0; k + 4 <= filterSize; k += 4) {
291                        for (j = 0; j < 16; ++j) {
292                            int from = (i + j) * filterSize + k;
293                            int to = i * filterSize + j * 4 + k * 16;
294                            memcpy(&filter[to], &filterCopy[from], 4 * sizeof(int16_t));
295                        }
296                    }
297                }
298            }
299            av_free(filterCopy);
300         }
301     }
302 #endif
303     return 0;
304 }
305 
sws_isSupportedInput(enum AVPixelFormat pix_fmt)306 int sws_isSupportedInput(enum AVPixelFormat pix_fmt)
307 {
308     return (unsigned)pix_fmt < FF_ARRAY_ELEMS(format_entries) ?
309            format_entries[pix_fmt].is_supported_in : 0;
310 }
311 
sws_isSupportedOutput(enum AVPixelFormat pix_fmt)312 int sws_isSupportedOutput(enum AVPixelFormat pix_fmt)
313 {
314     return (unsigned)pix_fmt < FF_ARRAY_ELEMS(format_entries) ?
315            format_entries[pix_fmt].is_supported_out : 0;
316 }
317 
sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)318 int sws_isSupportedEndiannessConversion(enum AVPixelFormat pix_fmt)
319 {
320     return (unsigned)pix_fmt < FF_ARRAY_ELEMS(format_entries) ?
321            format_entries[pix_fmt].is_supported_endianness : 0;
322 }
323 
getSplineCoeff(double a,double b,double c,double d,double dist)324 static double getSplineCoeff(double a, double b, double c, double d,
325                              double dist)
326 {
327     if (dist <= 1.0)
328         return ((d * dist + c) * dist + b) * dist + a;
329     else
330         return getSplineCoeff(0.0,
331                                b + 2.0 * c + 3.0 * d,
332                                c + 3.0 * d,
333                               -b - 3.0 * c - 6.0 * d,
334                               dist - 1.0);
335 }
336 
get_local_pos(SwsContext * s,int chr_subsample,int pos,int dir)337 static av_cold int get_local_pos(SwsContext *s, int chr_subsample, int pos, int dir)
338 {
339     if (pos == -1 || pos <= -513) {
340         pos = (128 << chr_subsample) - 128;
341     }
342     pos += 128; // relative to ideal left edge
343     return pos >> chr_subsample;
344 }
345 
346 typedef struct {
347     int flag;                   ///< flag associated to the algorithm
348     const char *description;    ///< human-readable description
349     int size_factor;            ///< size factor used when initing the filters
350 } ScaleAlgorithm;
351 
352 static const ScaleAlgorithm scale_algorithms[] = {
353     { SWS_AREA,          "area averaging",                  1 /* downscale only, for upscale it is bilinear */ },
354     { SWS_BICUBIC,       "bicubic",                         4 },
355     { SWS_BICUBLIN,      "luma bicubic / chroma bilinear", -1 },
356     { SWS_BILINEAR,      "bilinear",                        2 },
357     { SWS_FAST_BILINEAR, "fast bilinear",                  -1 },
358     { SWS_GAUSS,         "Gaussian",                        8 /* infinite ;) */ },
359     { SWS_LANCZOS,       "Lanczos",                        -1 /* custom */ },
360     { SWS_POINT,         "nearest neighbor / point",       -1 },
361     { SWS_SINC,          "sinc",                           20 /* infinite ;) */ },
362     { SWS_SPLINE,        "bicubic spline",                 20 /* infinite :)*/ },
363     { SWS_X,             "experimental",                    8 },
364 };
365 
initFilter(int16_t ** outFilter,int32_t ** filterPos,int * outFilterSize,int xInc,int srcW,int dstW,int filterAlign,int one,int flags,int cpu_flags,SwsVector * srcFilter,SwsVector * dstFilter,double param[2],int srcPos,int dstPos)366 static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos,
367                               int *outFilterSize, int xInc, int srcW,
368                               int dstW, int filterAlign, int one,
369                               int flags, int cpu_flags,
370                               SwsVector *srcFilter, SwsVector *dstFilter,
371                               double param[2], int srcPos, int dstPos)
372 {
373     int i;
374     int filterSize;
375     int filter2Size;
376     int minFilterSize;
377     int64_t *filter    = NULL;
378     int64_t *filter2   = NULL;
379     const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
380     int ret            = -1;
381 
382     emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
383 
384     // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
385     if (!FF_ALLOC_TYPED_ARRAY(*filterPos, dstW + 3))
386         goto nomem;
387 
388     if (FFABS(xInc - 0x10000) < 10 && srcPos == dstPos) { // unscaled
389         int i;
390         filterSize = 1;
391         if (!FF_ALLOCZ_TYPED_ARRAY(filter, dstW * filterSize))
392             goto nomem;
393 
394         for (i = 0; i < dstW; i++) {
395             filter[i * filterSize] = fone;
396             (*filterPos)[i]        = i;
397         }
398     } else if (flags & SWS_POINT) { // lame looking point sampling mode
399         int i;
400         int64_t xDstInSrc;
401         filterSize = 1;
402         if (!FF_ALLOC_TYPED_ARRAY(filter, dstW * filterSize))
403             goto nomem;
404 
405         xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
406         for (i = 0; i < dstW; i++) {
407             int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
408 
409             (*filterPos)[i] = xx;
410             filter[i]       = fone;
411             xDstInSrc      += xInc;
412         }
413     } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
414                (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
415         int i;
416         int64_t xDstInSrc;
417         filterSize = 2;
418         if (!FF_ALLOC_TYPED_ARRAY(filter, dstW * filterSize))
419             goto nomem;
420 
421         xDstInSrc = ((dstPos*(int64_t)xInc)>>8) - ((srcPos*0x8000LL)>>7);
422         for (i = 0; i < dstW; i++) {
423             int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
424             int j;
425 
426             (*filterPos)[i] = xx;
427             // bilinear upscale / linear interpolate / area averaging
428             for (j = 0; j < filterSize; j++) {
429                 int64_t coeff = fone - FFABS((int64_t)xx * (1 << 16) - xDstInSrc) * (fone >> 16);
430                 if (coeff < 0)
431                     coeff = 0;
432                 filter[i * filterSize + j] = coeff;
433                 xx++;
434             }
435             xDstInSrc += xInc;
436         }
437     } else {
438         int64_t xDstInSrc;
439         int sizeFactor = -1;
440 
441         for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
442             if (flags & scale_algorithms[i].flag && scale_algorithms[i].size_factor > 0) {
443                 sizeFactor = scale_algorithms[i].size_factor;
444                 break;
445             }
446         }
447         if (flags & SWS_LANCZOS)
448             sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
449         av_assert0(sizeFactor > 0);
450 
451         if (xInc <= 1 << 16)
452             filterSize = 1 + sizeFactor;    // upscale
453         else
454             filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
455 
456         filterSize = FFMIN(filterSize, srcW - 2);
457         filterSize = FFMAX(filterSize, 1);
458 
459         if (!FF_ALLOC_TYPED_ARRAY(filter, dstW * filterSize))
460             goto nomem;
461         xDstInSrc = ((dstPos*(int64_t)xInc)>>7) - ((srcPos*0x10000LL)>>7);
462         for (i = 0; i < dstW; i++) {
463             int xx = (xDstInSrc - (filterSize - 2) * (1LL<<16)) / (1 << 17);
464             int j;
465             (*filterPos)[i] = xx;
466             for (j = 0; j < filterSize; j++) {
467                 int64_t d = (FFABS(((int64_t)xx * (1 << 17)) - xDstInSrc)) << 13;
468                 double floatd;
469                 int64_t coeff;
470 
471                 if (xInc > 1 << 16)
472                     d = d * dstW / srcW;
473                 floatd = d * (1.0 / (1 << 30));
474 
475                 if (flags & SWS_BICUBIC) {
476                     int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] :   0) * (1 << 24);
477                     int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
478 
479                     if (d >= 1LL << 31) {
480                         coeff = 0.0;
481                     } else {
482                         int64_t dd  = (d  * d) >> 30;
483                         int64_t ddd = (dd * d) >> 30;
484 
485                         if (d < 1LL << 30)
486                             coeff =  (12 * (1 << 24) -  9 * B - 6 * C) * ddd +
487                                     (-18 * (1 << 24) + 12 * B + 6 * C) *  dd +
488                                       (6 * (1 << 24) -  2 * B)         * (1 << 30);
489                         else
490                             coeff =      (-B -  6 * C) * ddd +
491                                       (6 * B + 30 * C) * dd  +
492                                     (-12 * B - 48 * C) * d   +
493                                       (8 * B + 24 * C) * (1 << 30);
494                     }
495                     coeff /= (1LL<<54)/fone;
496                 } else if (flags & SWS_X) {
497                     double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
498                     double c;
499 
500                     if (floatd < 1.0)
501                         c = cos(floatd * M_PI);
502                     else
503                         c = -1.0;
504                     if (c < 0.0)
505                         c = -pow(-c, A);
506                     else
507                         c = pow(c, A);
508                     coeff = (c * 0.5 + 0.5) * fone;
509                 } else if (flags & SWS_AREA) {
510                     int64_t d2 = d - (1 << 29);
511                     if (d2 * xInc < -(1LL << (29 + 16)))
512                         coeff = 1.0 * (1LL << (30 + 16));
513                     else if (d2 * xInc < (1LL << (29 + 16)))
514                         coeff = -d2 * xInc + (1LL << (29 + 16));
515                     else
516                         coeff = 0.0;
517                     coeff *= fone >> (30 + 16);
518                 } else if (flags & SWS_GAUSS) {
519                     double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
520                     coeff = exp2(-p * floatd * floatd) * fone;
521                 } else if (flags & SWS_SINC) {
522                     coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
523                 } else if (flags & SWS_LANCZOS) {
524                     double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
525                     coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
526                              (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
527                     if (floatd > p)
528                         coeff = 0;
529                 } else if (flags & SWS_BILINEAR) {
530                     coeff = (1 << 30) - d;
531                     if (coeff < 0)
532                         coeff = 0;
533                     coeff *= fone >> 30;
534                 } else if (flags & SWS_SPLINE) {
535                     double p = -2.196152422706632;
536                     coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
537                 } else {
538                     av_assert0(0);
539                 }
540 
541                 filter[i * filterSize + j] = coeff;
542                 xx++;
543             }
544             xDstInSrc += 2 * xInc;
545         }
546     }
547 
548     /* apply src & dst Filter to filter -> filter2
549      * av_free(filter);
550      */
551     av_assert0(filterSize > 0);
552     filter2Size = filterSize;
553     if (srcFilter)
554         filter2Size += srcFilter->length - 1;
555     if (dstFilter)
556         filter2Size += dstFilter->length - 1;
557     av_assert0(filter2Size > 0);
558     if (!FF_ALLOCZ_TYPED_ARRAY(filter2, dstW * filter2Size))
559         goto nomem;
560     for (i = 0; i < dstW; i++) {
561         int j, k;
562 
563         if (srcFilter) {
564             for (k = 0; k < srcFilter->length; k++) {
565                 for (j = 0; j < filterSize; j++)
566                     filter2[i * filter2Size + k + j] +=
567                         srcFilter->coeff[k] * filter[i * filterSize + j];
568             }
569         } else {
570             for (j = 0; j < filterSize; j++)
571                 filter2[i * filter2Size + j] = filter[i * filterSize + j];
572         }
573         // FIXME dstFilter
574 
575         (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
576     }
577     av_freep(&filter);
578 
579     /* try to reduce the filter-size (step1 find size and shift left) */
580     // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
581     minFilterSize = 0;
582     for (i = dstW - 1; i >= 0; i--) {
583         int min = filter2Size;
584         int j;
585         int64_t cutOff = 0.0;
586 
587         /* get rid of near zero elements on the left by shifting left */
588         for (j = 0; j < filter2Size; j++) {
589             int k;
590             cutOff += FFABS(filter2[i * filter2Size]);
591 
592             if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
593                 break;
594 
595             /* preserve monotonicity because the core can't handle the
596              * filter otherwise */
597             if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
598                 break;
599 
600             // move filter coefficients left
601             for (k = 1; k < filter2Size; k++)
602                 filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
603             filter2[i * filter2Size + k - 1] = 0;
604             (*filterPos)[i]++;
605         }
606 
607         cutOff = 0;
608         /* count near zeros on the right */
609         for (j = filter2Size - 1; j > 0; j--) {
610             cutOff += FFABS(filter2[i * filter2Size + j]);
611 
612             if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
613                 break;
614             min--;
615         }
616 
617         if (min > minFilterSize)
618             minFilterSize = min;
619     }
620 
621     if (PPC_ALTIVEC(cpu_flags)) {
622         // we can handle the special case 4, so we don't want to go the full 8
623         if (minFilterSize < 5)
624             filterAlign = 4;
625 
626         /* We really don't want to waste our time doing useless computation, so
627          * fall back on the scalar C code for very small filters.
628          * Vectorizing is worth it only if you have a decent-sized vector. */
629         if (minFilterSize < 3)
630             filterAlign = 1;
631     }
632 
633     if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
634         // special case for unscaled vertical filtering
635         if (minFilterSize == 1 && filterAlign == 2)
636             filterAlign = 1;
637     }
638 
639     av_assert0(minFilterSize > 0);
640     filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
641     av_assert0(filterSize > 0);
642     filter = av_malloc_array(dstW, filterSize * sizeof(*filter));
643     if (!filter)
644         goto nomem;
645     if (filterSize >= MAX_FILTER_SIZE * 16 /
646                       ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16)) {
647         ret = RETCODE_USE_CASCADE;
648         goto fail;
649     }
650     *outFilterSize = filterSize;
651 
652     if (flags & SWS_PRINT_INFO)
653         av_log(NULL, AV_LOG_VERBOSE,
654                "SwScaler: reducing / aligning filtersize %d -> %d\n",
655                filter2Size, filterSize);
656     /* try to reduce the filter-size (step2 reduce it) */
657     for (i = 0; i < dstW; i++) {
658         int j;
659 
660         for (j = 0; j < filterSize; j++) {
661             if (j >= filter2Size)
662                 filter[i * filterSize + j] = 0;
663             else
664                 filter[i * filterSize + j] = filter2[i * filter2Size + j];
665             if ((flags & SWS_BITEXACT) && j >= minFilterSize)
666                 filter[i * filterSize + j] = 0;
667         }
668     }
669 
670     // FIXME try to align filterPos if possible
671 
672     // fix borders
673     for (i = 0; i < dstW; i++) {
674         int j;
675         if ((*filterPos)[i] < 0) {
676             // move filter coefficients left to compensate for filterPos
677             for (j = 1; j < filterSize; j++) {
678                 int left = FFMAX(j + (*filterPos)[i], 0);
679                 filter[i * filterSize + left] += filter[i * filterSize + j];
680                 filter[i * filterSize + j]     = 0;
681             }
682             (*filterPos)[i]= 0;
683         }
684 
685         if ((*filterPos)[i] + filterSize > srcW) {
686             int shift = (*filterPos)[i] + FFMIN(filterSize - srcW, 0);
687             int64_t acc = 0;
688 
689             for (j = filterSize - 1; j >= 0; j--) {
690                 if ((*filterPos)[i] + j >= srcW) {
691                     acc += filter[i * filterSize + j];
692                     filter[i * filterSize + j] = 0;
693                 }
694             }
695             for (j = filterSize - 1; j >= 0; j--) {
696                 if (j < shift) {
697                     filter[i * filterSize + j] = 0;
698                 } else {
699                     filter[i * filterSize + j] = filter[i * filterSize + j - shift];
700                 }
701             }
702 
703             (*filterPos)[i]-= shift;
704             filter[i * filterSize + srcW - 1 - (*filterPos)[i]] += acc;
705         }
706         av_assert0((*filterPos)[i] >= 0);
707         av_assert0((*filterPos)[i] < srcW);
708         if ((*filterPos)[i] + filterSize > srcW) {
709             for (j = 0; j < filterSize; j++) {
710                 av_assert0((*filterPos)[i] + j < srcW || !filter[i * filterSize + j]);
711             }
712         }
713     }
714 
715     // Note the +1 is for the MMX scaler which reads over the end
716     /* align at 16 for AltiVec (needed by hScale_altivec_real) */
717     if (!FF_ALLOCZ_TYPED_ARRAY(*outFilter, *outFilterSize * (dstW + 3)))
718         goto nomem;
719 
720     /* normalize & store in outFilter */
721     for (i = 0; i < dstW; i++) {
722         int j;
723         int64_t error = 0;
724         int64_t sum   = 0;
725 
726         for (j = 0; j < filterSize; j++) {
727             sum += filter[i * filterSize + j];
728         }
729         sum = (sum + one / 2) / one;
730         if (!sum) {
731             av_log(NULL, AV_LOG_WARNING, "SwScaler: zero vector in scaling\n");
732             sum = 1;
733         }
734         for (j = 0; j < *outFilterSize; j++) {
735             int64_t v = filter[i * filterSize + j] + error;
736             int intV  = ROUNDED_DIV(v, sum);
737             (*outFilter)[i * (*outFilterSize) + j] = intV;
738             error                                  = v - intV * sum;
739         }
740     }
741 
742     (*filterPos)[dstW + 0] =
743     (*filterPos)[dstW + 1] =
744     (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
745                                                       * read over the end */
746     for (i = 0; i < *outFilterSize; i++) {
747         int k = (dstW - 1) * (*outFilterSize) + i;
748         (*outFilter)[k + 1 * (*outFilterSize)] =
749         (*outFilter)[k + 2 * (*outFilterSize)] =
750         (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
751     }
752 
753     ret = 0;
754     goto done;
755 nomem:
756     ret = AVERROR(ENOMEM);
757 fail:
758     if(ret < 0)
759         av_log(NULL, ret == RETCODE_USE_CASCADE ? AV_LOG_DEBUG : AV_LOG_ERROR, "sws: initFilter failed\n");
760 done:
761     av_free(filter);
762     av_free(filter2);
763     return ret;
764 }
765 
fill_rgb2yuv_table(SwsContext * c,const int table[4],int dstRange)766 static void fill_rgb2yuv_table(SwsContext *c, const int table[4], int dstRange)
767 {
768     int64_t W, V, Z, Cy, Cu, Cv;
769     int64_t vr =  table[0];
770     int64_t ub =  table[1];
771     int64_t ug = -table[2];
772     int64_t vg = -table[3];
773     int64_t ONE = 65536;
774     int64_t cy = ONE;
775     uint8_t *p = (uint8_t*)c->input_rgb2yuv_table;
776     int i;
777     static const int8_t map[] = {
778     BY_IDX, GY_IDX, -1    , BY_IDX, BY_IDX, GY_IDX, -1    , BY_IDX,
779     RY_IDX, -1    , GY_IDX, RY_IDX, RY_IDX, -1    , GY_IDX, RY_IDX,
780     RY_IDX, GY_IDX, -1    , RY_IDX, RY_IDX, GY_IDX, -1    , RY_IDX,
781     BY_IDX, -1    , GY_IDX, BY_IDX, BY_IDX, -1    , GY_IDX, BY_IDX,
782     BU_IDX, GU_IDX, -1    , BU_IDX, BU_IDX, GU_IDX, -1    , BU_IDX,
783     RU_IDX, -1    , GU_IDX, RU_IDX, RU_IDX, -1    , GU_IDX, RU_IDX,
784     RU_IDX, GU_IDX, -1    , RU_IDX, RU_IDX, GU_IDX, -1    , RU_IDX,
785     BU_IDX, -1    , GU_IDX, BU_IDX, BU_IDX, -1    , GU_IDX, BU_IDX,
786     BV_IDX, GV_IDX, -1    , BV_IDX, BV_IDX, GV_IDX, -1    , BV_IDX,
787     RV_IDX, -1    , GV_IDX, RV_IDX, RV_IDX, -1    , GV_IDX, RV_IDX,
788     RV_IDX, GV_IDX, -1    , RV_IDX, RV_IDX, GV_IDX, -1    , RV_IDX,
789     BV_IDX, -1    , GV_IDX, BV_IDX, BV_IDX, -1    , GV_IDX, BV_IDX,
790     RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX,
791     BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX, BY_IDX, RY_IDX,
792     GY_IDX, -1    , GY_IDX, -1    , GY_IDX, -1    , GY_IDX, -1    ,
793     -1    , GY_IDX, -1    , GY_IDX, -1    , GY_IDX, -1    , GY_IDX,
794     RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX,
795     BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX, BU_IDX, RU_IDX,
796     GU_IDX, -1    , GU_IDX, -1    , GU_IDX, -1    , GU_IDX, -1    ,
797     -1    , GU_IDX, -1    , GU_IDX, -1    , GU_IDX, -1    , GU_IDX,
798     RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX,
799     BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX, BV_IDX, RV_IDX,
800     GV_IDX, -1    , GV_IDX, -1    , GV_IDX, -1    , GV_IDX, -1    ,
801     -1    , GV_IDX, -1    , GV_IDX, -1    , GV_IDX, -1    , GV_IDX, //23
802     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //24
803     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //25
804     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //26
805     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //27
806     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //28
807     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //29
808     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //30
809     -1    , -1    , -1    , -1    , -1    , -1    , -1    , -1    , //31
810     BY_IDX, GY_IDX, RY_IDX, -1    , -1    , -1    , -1    , -1    , //32
811     BU_IDX, GU_IDX, RU_IDX, -1    , -1    , -1    , -1    , -1    , //33
812     BV_IDX, GV_IDX, RV_IDX, -1    , -1    , -1    , -1    , -1    , //34
813     };
814 
815     dstRange = 0; //FIXME range = 1 is handled elsewhere
816 
817     if (!dstRange) {
818         cy = cy * 255 / 219;
819     } else {
820         vr = vr * 224 / 255;
821         ub = ub * 224 / 255;
822         ug = ug * 224 / 255;
823         vg = vg * 224 / 255;
824     }
825     W = ROUNDED_DIV(ONE*ONE*ug, ub);
826     V = ROUNDED_DIV(ONE*ONE*vg, vr);
827     Z = ONE*ONE-W-V;
828 
829     Cy = ROUNDED_DIV(cy*Z, ONE);
830     Cu = ROUNDED_DIV(ub*Z, ONE);
831     Cv = ROUNDED_DIV(vr*Z, ONE);
832 
833     c->input_rgb2yuv_table[RY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V        , Cy);
834     c->input_rgb2yuv_table[GY_IDX] =  ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE  , Cy);
835     c->input_rgb2yuv_table[BY_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W        , Cy);
836 
837     c->input_rgb2yuv_table[RU_IDX] =  ROUNDED_DIV((1 << RGB2YUV_SHIFT)*V        , Cu);
838     c->input_rgb2yuv_table[GU_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE  , Cu);
839     c->input_rgb2yuv_table[BU_IDX] =  ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(Z+W)    , Cu);
840 
841     c->input_rgb2yuv_table[RV_IDX] =  ROUNDED_DIV((1 << RGB2YUV_SHIFT)*(V+Z)    , Cv);
842     c->input_rgb2yuv_table[GV_IDX] = -ROUNDED_DIV((1 << RGB2YUV_SHIFT)*ONE*ONE  , Cv);
843     c->input_rgb2yuv_table[BV_IDX] =  ROUNDED_DIV((1 << RGB2YUV_SHIFT)*W        , Cv);
844 
845     if(/*!dstRange && */!memcmp(table, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], sizeof(ff_yuv2rgb_coeffs[SWS_CS_DEFAULT]))) {
846         c->input_rgb2yuv_table[BY_IDX] =  ((int)(0.114 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
847         c->input_rgb2yuv_table[BV_IDX] = (-(int)(0.081 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
848         c->input_rgb2yuv_table[BU_IDX] =  ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
849         c->input_rgb2yuv_table[GY_IDX] =  ((int)(0.587 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
850         c->input_rgb2yuv_table[GV_IDX] = (-(int)(0.419 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
851         c->input_rgb2yuv_table[GU_IDX] = (-(int)(0.331 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
852         c->input_rgb2yuv_table[RY_IDX] =  ((int)(0.299 * 219 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
853         c->input_rgb2yuv_table[RV_IDX] =  ((int)(0.500 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
854         c->input_rgb2yuv_table[RU_IDX] = (-(int)(0.169 * 224 / 255 * (1 << RGB2YUV_SHIFT) + 0.5));
855     }
856     for(i=0; i<FF_ARRAY_ELEMS(map); i++)
857         AV_WL16(p + 16*4 + 2*i, map[i] >= 0 ? c->input_rgb2yuv_table[map[i]] : 0);
858 }
859 
fill_xyztables(struct SwsContext * c)860 static void fill_xyztables(struct SwsContext *c)
861 {
862     int i;
863     double xyzgamma = XYZ_GAMMA;
864     double rgbgamma = 1.0 / RGB_GAMMA;
865     double xyzgammainv = 1.0 / XYZ_GAMMA;
866     double rgbgammainv = RGB_GAMMA;
867     static const int16_t xyz2rgb_matrix[3][4] = {
868         {13270, -6295, -2041},
869         {-3969,  7682,   170},
870         {  228,  -835,  4329} };
871     static const int16_t rgb2xyz_matrix[3][4] = {
872         {1689, 1464,  739},
873         { 871, 2929,  296},
874         {  79,  488, 3891} };
875     static int16_t xyzgamma_tab[4096], rgbgamma_tab[4096], xyzgammainv_tab[4096], rgbgammainv_tab[4096];
876 
877     memcpy(c->xyz2rgb_matrix, xyz2rgb_matrix, sizeof(c->xyz2rgb_matrix));
878     memcpy(c->rgb2xyz_matrix, rgb2xyz_matrix, sizeof(c->rgb2xyz_matrix));
879     c->xyzgamma = xyzgamma_tab;
880     c->rgbgamma = rgbgamma_tab;
881     c->xyzgammainv = xyzgammainv_tab;
882     c->rgbgammainv = rgbgammainv_tab;
883 
884     if (rgbgamma_tab[4095])
885         return;
886 
887     /* set gamma vectors */
888     for (i = 0; i < 4096; i++) {
889         xyzgamma_tab[i] = lrint(pow(i / 4095.0, xyzgamma) * 4095.0);
890         rgbgamma_tab[i] = lrint(pow(i / 4095.0, rgbgamma) * 4095.0);
891         xyzgammainv_tab[i] = lrint(pow(i / 4095.0, xyzgammainv) * 4095.0);
892         rgbgammainv_tab[i] = lrint(pow(i / 4095.0, rgbgammainv) * 4095.0);
893     }
894 }
895 
range_override_needed(enum AVPixelFormat format)896 static int range_override_needed(enum AVPixelFormat format)
897 {
898     return !isYUV(format) && !isGray(format);
899 }
900 
sws_setColorspaceDetails(struct SwsContext * c,const int inv_table[4],int srcRange,const int table[4],int dstRange,int brightness,int contrast,int saturation)901 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
902                              int srcRange, const int table[4], int dstRange,
903                              int brightness, int contrast, int saturation)
904 {
905     const AVPixFmtDescriptor *desc_dst;
906     const AVPixFmtDescriptor *desc_src;
907     int need_reinit = 0;
908 
909     if (c->nb_slice_ctx) {
910         int parent_ret = 0;
911         for (int i = 0; i < c->nb_slice_ctx; i++) {
912             int ret = sws_setColorspaceDetails(c->slice_ctx[i], inv_table,
913                                                srcRange, table, dstRange,
914                                                brightness, contrast, saturation);
915             if (ret < 0)
916                 parent_ret = ret;
917         }
918 
919         return parent_ret;
920     }
921 
922     handle_formats(c);
923     desc_dst = av_pix_fmt_desc_get(c->dstFormat);
924     desc_src = av_pix_fmt_desc_get(c->srcFormat);
925 
926     if(range_override_needed(c->dstFormat))
927         dstRange = 0;
928     if(range_override_needed(c->srcFormat))
929         srcRange = 0;
930 
931     if (c->srcRange != srcRange ||
932         c->dstRange != dstRange ||
933         c->brightness != brightness ||
934         c->contrast   != contrast ||
935         c->saturation != saturation ||
936         memcmp(c->srcColorspaceTable, inv_table, sizeof(int) * 4) ||
937         memcmp(c->dstColorspaceTable,     table, sizeof(int) * 4)
938     )
939         need_reinit = 1;
940 
941     memmove(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
942     memmove(c->dstColorspaceTable, table, sizeof(int) * 4);
943 
944 
945 
946     c->brightness = brightness;
947     c->contrast   = contrast;
948     c->saturation = saturation;
949     c->srcRange   = srcRange;
950     c->dstRange   = dstRange;
951 
952     //The srcBpc check is possibly wrong but we seem to lack a definitive reference to test this
953     //and what we have in ticket 2939 looks better with this check
954     if (need_reinit && (c->srcBpc == 8 || !isYUV(c->srcFormat)))
955         ff_sws_init_range_convert(c);
956 
957     c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
958     c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
959 
960     if (c->cascaded_context[c->cascaded_mainindex])
961         return sws_setColorspaceDetails(c->cascaded_context[c->cascaded_mainindex],inv_table, srcRange,table, dstRange, brightness,  contrast, saturation);
962 
963     if (!need_reinit)
964         return 0;
965 
966     if ((isYUV(c->dstFormat) || isGray(c->dstFormat)) && (isYUV(c->srcFormat) || isGray(c->srcFormat))) {
967         if (!c->cascaded_context[0] &&
968             memcmp(c->dstColorspaceTable, c->srcColorspaceTable, sizeof(int) * 4) &&
969             c->srcW && c->srcH && c->dstW && c->dstH) {
970             enum AVPixelFormat tmp_format;
971             int tmp_width, tmp_height;
972             int srcW = c->srcW;
973             int srcH = c->srcH;
974             int dstW = c->dstW;
975             int dstH = c->dstH;
976             int ret;
977             av_log(c, AV_LOG_VERBOSE, "YUV color matrix differs for YUV->YUV, using intermediate RGB to convert\n");
978 
979             if (isNBPS(c->dstFormat) || is16BPS(c->dstFormat)) {
980                 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
981                     tmp_format = AV_PIX_FMT_BGRA64;
982                 } else {
983                     tmp_format = AV_PIX_FMT_BGR48;
984                 }
985             } else {
986                 if (isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) {
987                     tmp_format = AV_PIX_FMT_BGRA;
988                 } else {
989                     tmp_format = AV_PIX_FMT_BGR24;
990                 }
991             }
992 
993             if (srcW*srcH > dstW*dstH) {
994                 tmp_width  = dstW;
995                 tmp_height = dstH;
996             } else {
997                 tmp_width  = srcW;
998                 tmp_height = srcH;
999             }
1000 
1001             ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1002                                 tmp_width, tmp_height, tmp_format, 64);
1003             if (ret < 0)
1004                 return ret;
1005 
1006             c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, c->srcFormat,
1007                                                         tmp_width, tmp_height, tmp_format,
1008                                                         c->flags, c->param);
1009             if (!c->cascaded_context[0])
1010                 return -1;
1011 
1012             c->cascaded_context[0]->alphablend = c->alphablend;
1013             ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
1014             if (ret < 0)
1015                 return ret;
1016             //we set both src and dst depending on that the RGB side will be ignored
1017             sws_setColorspaceDetails(c->cascaded_context[0], inv_table,
1018                                      srcRange, table, dstRange,
1019                                      brightness, contrast, saturation);
1020 
1021             c->cascaded_context[1] = sws_alloc_set_opts(tmp_width, tmp_height, tmp_format,
1022                                                         dstW, dstH, c->dstFormat,
1023                                                         c->flags, c->param);
1024             if (!c->cascaded_context[1])
1025                 return -1;
1026             c->cascaded_context[1]->srcRange = srcRange;
1027             c->cascaded_context[1]->dstRange = dstRange;
1028             ret = sws_init_context(c->cascaded_context[1], NULL , NULL);
1029             if (ret < 0)
1030                 return ret;
1031             sws_setColorspaceDetails(c->cascaded_context[1], inv_table,
1032                                      srcRange, table, dstRange,
1033                                      0, 1 << 16, 1 << 16);
1034             return 0;
1035         }
1036         //We do not support this combination currently, we need to cascade more contexts to compensate
1037         if (c->cascaded_context[0] && memcmp(c->dstColorspaceTable, c->srcColorspaceTable, sizeof(int) * 4))
1038             return -1; //AVERROR_PATCHWELCOME;
1039         return 0;
1040     }
1041 
1042     if (!isYUV(c->dstFormat) && !isGray(c->dstFormat)) {
1043         ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
1044                                  contrast, saturation);
1045         // FIXME factorize
1046 
1047 #if ARCH_PPC
1048         ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness,
1049                                    contrast, saturation);
1050 #endif
1051     }
1052 
1053     fill_rgb2yuv_table(c, table, dstRange);
1054 
1055     return 0;
1056 }
1057 
sws_getColorspaceDetails(struct SwsContext * c,int ** inv_table,int * srcRange,int ** table,int * dstRange,int * brightness,int * contrast,int * saturation)1058 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
1059                              int *srcRange, int **table, int *dstRange,
1060                              int *brightness, int *contrast, int *saturation)
1061 {
1062     if (!c )
1063         return -1;
1064 
1065     if (c->nb_slice_ctx) {
1066         return sws_getColorspaceDetails(c->slice_ctx[0], inv_table, srcRange,
1067                                         table, dstRange, brightness, contrast,
1068                                         saturation);
1069     }
1070 
1071     *inv_table  = c->srcColorspaceTable;
1072     *table      = c->dstColorspaceTable;
1073     *srcRange   = range_override_needed(c->srcFormat) ? 1 : c->srcRange;
1074     *dstRange   = range_override_needed(c->dstFormat) ? 1 : c->dstRange;
1075     *brightness = c->brightness;
1076     *contrast   = c->contrast;
1077     *saturation = c->saturation;
1078 
1079     return 0;
1080 }
1081 
handle_jpeg(enum AVPixelFormat * format)1082 static int handle_jpeg(enum AVPixelFormat *format)
1083 {
1084     switch (*format) {
1085     case AV_PIX_FMT_YUVJ420P:
1086         *format = AV_PIX_FMT_YUV420P;
1087         return 1;
1088     case AV_PIX_FMT_YUVJ411P:
1089         *format = AV_PIX_FMT_YUV411P;
1090         return 1;
1091     case AV_PIX_FMT_YUVJ422P:
1092         *format = AV_PIX_FMT_YUV422P;
1093         return 1;
1094     case AV_PIX_FMT_YUVJ444P:
1095         *format = AV_PIX_FMT_YUV444P;
1096         return 1;
1097     case AV_PIX_FMT_YUVJ440P:
1098         *format = AV_PIX_FMT_YUV440P;
1099         return 1;
1100     case AV_PIX_FMT_GRAY8:
1101     case AV_PIX_FMT_YA8:
1102     case AV_PIX_FMT_GRAY9LE:
1103     case AV_PIX_FMT_GRAY9BE:
1104     case AV_PIX_FMT_GRAY10LE:
1105     case AV_PIX_FMT_GRAY10BE:
1106     case AV_PIX_FMT_GRAY12LE:
1107     case AV_PIX_FMT_GRAY12BE:
1108     case AV_PIX_FMT_GRAY14LE:
1109     case AV_PIX_FMT_GRAY14BE:
1110     case AV_PIX_FMT_GRAY16LE:
1111     case AV_PIX_FMT_GRAY16BE:
1112     case AV_PIX_FMT_YA16BE:
1113     case AV_PIX_FMT_YA16LE:
1114         return 1;
1115     default:
1116         return 0;
1117     }
1118 }
1119 
handle_0alpha(enum AVPixelFormat * format)1120 static int handle_0alpha(enum AVPixelFormat *format)
1121 {
1122     switch (*format) {
1123     case AV_PIX_FMT_0BGR    : *format = AV_PIX_FMT_ABGR   ; return 1;
1124     case AV_PIX_FMT_BGR0    : *format = AV_PIX_FMT_BGRA   ; return 4;
1125     case AV_PIX_FMT_0RGB    : *format = AV_PIX_FMT_ARGB   ; return 1;
1126     case AV_PIX_FMT_RGB0    : *format = AV_PIX_FMT_RGBA   ; return 4;
1127     default:                                          return 0;
1128     }
1129 }
1130 
handle_xyz(enum AVPixelFormat * format)1131 static int handle_xyz(enum AVPixelFormat *format)
1132 {
1133     switch (*format) {
1134     case AV_PIX_FMT_XYZ12BE : *format = AV_PIX_FMT_RGB48BE; return 1;
1135     case AV_PIX_FMT_XYZ12LE : *format = AV_PIX_FMT_RGB48LE; return 1;
1136     default:                                                return 0;
1137     }
1138 }
1139 
handle_formats(SwsContext * c)1140 static void handle_formats(SwsContext *c)
1141 {
1142     c->src0Alpha |= handle_0alpha(&c->srcFormat);
1143     c->dst0Alpha |= handle_0alpha(&c->dstFormat);
1144     c->srcXYZ    |= handle_xyz(&c->srcFormat);
1145     c->dstXYZ    |= handle_xyz(&c->dstFormat);
1146     if (c->srcXYZ || c->dstXYZ)
1147         fill_xyztables(c);
1148 }
1149 
sws_alloc_context(void)1150 SwsContext *sws_alloc_context(void)
1151 {
1152     SwsContext *c = av_mallocz(sizeof(SwsContext));
1153 
1154     av_assert0(offsetof(SwsContext, redDither) + DITHER32_INT == offsetof(SwsContext, dither32));
1155 
1156     if (c) {
1157         c->av_class = &ff_sws_context_class;
1158         av_opt_set_defaults(c);
1159         atomic_init(&c->stride_unaligned_warned, 0);
1160         atomic_init(&c->data_unaligned_warned,   0);
1161     }
1162 
1163     return c;
1164 }
1165 
alloc_gamma_tbl(double e)1166 static uint16_t * alloc_gamma_tbl(double e)
1167 {
1168     int i = 0;
1169     uint16_t * tbl;
1170     tbl = (uint16_t*)av_malloc(sizeof(uint16_t) * 1 << 16);
1171     if (!tbl)
1172         return NULL;
1173 
1174     for (i = 0; i < 65536; ++i) {
1175         tbl[i] = pow(i / 65535.0, e) * 65535.0;
1176     }
1177     return tbl;
1178 }
1179 
alphaless_fmt(enum AVPixelFormat fmt)1180 static enum AVPixelFormat alphaless_fmt(enum AVPixelFormat fmt)
1181 {
1182     switch(fmt) {
1183     case AV_PIX_FMT_ARGB:       return AV_PIX_FMT_RGB24;
1184     case AV_PIX_FMT_RGBA:       return AV_PIX_FMT_RGB24;
1185     case AV_PIX_FMT_ABGR:       return AV_PIX_FMT_BGR24;
1186     case AV_PIX_FMT_BGRA:       return AV_PIX_FMT_BGR24;
1187     case AV_PIX_FMT_YA8:        return AV_PIX_FMT_GRAY8;
1188 
1189     case AV_PIX_FMT_YUVA420P:   return AV_PIX_FMT_YUV420P;
1190     case AV_PIX_FMT_YUVA422P:   return AV_PIX_FMT_YUV422P;
1191     case AV_PIX_FMT_YUVA444P:           return AV_PIX_FMT_YUV444P;
1192 
1193     case AV_PIX_FMT_GBRAP:              return AV_PIX_FMT_GBRP;
1194 
1195     case AV_PIX_FMT_GBRAP10LE:          return AV_PIX_FMT_GBRP10;
1196     case AV_PIX_FMT_GBRAP10BE:          return AV_PIX_FMT_GBRP10;
1197 
1198     case AV_PIX_FMT_GBRAP12LE:          return AV_PIX_FMT_GBRP12;
1199     case AV_PIX_FMT_GBRAP12BE:          return AV_PIX_FMT_GBRP12;
1200 
1201     case AV_PIX_FMT_GBRAP16LE:          return AV_PIX_FMT_GBRP16;
1202     case AV_PIX_FMT_GBRAP16BE:          return AV_PIX_FMT_GBRP16;
1203 
1204     case AV_PIX_FMT_RGBA64LE:   return AV_PIX_FMT_RGB48;
1205     case AV_PIX_FMT_RGBA64BE:   return AV_PIX_FMT_RGB48;
1206     case AV_PIX_FMT_BGRA64LE:   return AV_PIX_FMT_BGR48;
1207     case AV_PIX_FMT_BGRA64BE:   return AV_PIX_FMT_BGR48;
1208 
1209     case AV_PIX_FMT_YA16BE:             return AV_PIX_FMT_GRAY16;
1210     case AV_PIX_FMT_YA16LE:             return AV_PIX_FMT_GRAY16;
1211 
1212     case AV_PIX_FMT_YUVA420P9BE:        return AV_PIX_FMT_YUV420P9;
1213     case AV_PIX_FMT_YUVA422P9BE:        return AV_PIX_FMT_YUV422P9;
1214     case AV_PIX_FMT_YUVA444P9BE:        return AV_PIX_FMT_YUV444P9;
1215     case AV_PIX_FMT_YUVA420P9LE:        return AV_PIX_FMT_YUV420P9;
1216     case AV_PIX_FMT_YUVA422P9LE:        return AV_PIX_FMT_YUV422P9;
1217     case AV_PIX_FMT_YUVA444P9LE:        return AV_PIX_FMT_YUV444P9;
1218     case AV_PIX_FMT_YUVA420P10BE:       return AV_PIX_FMT_YUV420P10;
1219     case AV_PIX_FMT_YUVA422P10BE:       return AV_PIX_FMT_YUV422P10;
1220     case AV_PIX_FMT_YUVA444P10BE:       return AV_PIX_FMT_YUV444P10;
1221     case AV_PIX_FMT_YUVA420P10LE:       return AV_PIX_FMT_YUV420P10;
1222     case AV_PIX_FMT_YUVA422P10LE:       return AV_PIX_FMT_YUV422P10;
1223     case AV_PIX_FMT_YUVA444P10LE:       return AV_PIX_FMT_YUV444P10;
1224     case AV_PIX_FMT_YUVA420P16BE:       return AV_PIX_FMT_YUV420P16;
1225     case AV_PIX_FMT_YUVA422P16BE:       return AV_PIX_FMT_YUV422P16;
1226     case AV_PIX_FMT_YUVA444P16BE:       return AV_PIX_FMT_YUV444P16;
1227     case AV_PIX_FMT_YUVA420P16LE:       return AV_PIX_FMT_YUV420P16;
1228     case AV_PIX_FMT_YUVA422P16LE:       return AV_PIX_FMT_YUV422P16;
1229     case AV_PIX_FMT_YUVA444P16LE:       return AV_PIX_FMT_YUV444P16;
1230 
1231 //     case AV_PIX_FMT_AYUV64LE:
1232 //     case AV_PIX_FMT_AYUV64BE:
1233 //     case AV_PIX_FMT_PAL8:
1234     default: return AV_PIX_FMT_NONE;
1235     }
1236 }
1237 
context_init_threaded(SwsContext * c,SwsFilter * src_filter,SwsFilter * dst_filter)1238 static int context_init_threaded(SwsContext *c,
1239                                  SwsFilter *src_filter, SwsFilter *dst_filter)
1240 {
1241     int ret;
1242 
1243     ret = avpriv_slicethread_create(&c->slicethread, (void*)c,
1244                                     ff_sws_slice_worker, NULL, c->nb_threads);
1245     if (ret == AVERROR(ENOSYS)) {
1246         c->nb_threads = 1;
1247         return 0;
1248     } else if (ret < 0)
1249         return ret;
1250 
1251     c->nb_threads = ret;
1252 
1253     c->slice_ctx = av_calloc(c->nb_threads, sizeof(*c->slice_ctx));
1254     c->slice_err = av_calloc(c->nb_threads, sizeof(*c->slice_err));
1255     if (!c->slice_ctx || !c->slice_err)
1256         return AVERROR(ENOMEM);
1257 
1258     for (int i = 0; i < c->nb_threads; i++) {
1259         c->slice_ctx[i] = sws_alloc_context();
1260         if (!c->slice_ctx[i])
1261             return AVERROR(ENOMEM);
1262 
1263         c->slice_ctx[i]->parent = c;
1264 
1265         ret = av_opt_copy((void*)c->slice_ctx[i], (void*)c);
1266         if (ret < 0)
1267             return ret;
1268 
1269         c->slice_ctx[i]->nb_threads = 1;
1270 
1271         ret = sws_init_context(c->slice_ctx[i], src_filter, dst_filter);
1272         if (ret < 0)
1273             return ret;
1274 
1275         c->nb_slice_ctx++;
1276 
1277         if (c->slice_ctx[i]->dither == SWS_DITHER_ED) {
1278             av_log(c, AV_LOG_VERBOSE,
1279                    "Error-diffusion dither is in use, scaling will be single-threaded.");
1280             break;
1281         }
1282     }
1283 
1284     c->frame_src = av_frame_alloc();
1285     c->frame_dst = av_frame_alloc();
1286     if (!c->frame_src || !c->frame_dst)
1287         return AVERROR(ENOMEM);
1288 
1289     return 0;
1290 }
1291 
sws_init_context(SwsContext * c,SwsFilter * srcFilter,SwsFilter * dstFilter)1292 av_cold int sws_init_context(SwsContext *c, SwsFilter *srcFilter,
1293                              SwsFilter *dstFilter)
1294 {
1295     int i;
1296     int usesVFilter, usesHFilter;
1297     int unscaled;
1298     SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
1299     int srcW              = c->srcW;
1300     int srcH              = c->srcH;
1301     int dstW              = c->dstW;
1302     int dstH              = c->dstH;
1303     int dst_stride        = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
1304     int flags, cpu_flags;
1305     enum AVPixelFormat srcFormat = c->srcFormat;
1306     enum AVPixelFormat dstFormat = c->dstFormat;
1307     const AVPixFmtDescriptor *desc_src;
1308     const AVPixFmtDescriptor *desc_dst;
1309     int ret = 0;
1310     enum AVPixelFormat tmpFmt;
1311     static const float float_mult = 1.0f / 255.0f;
1312     static AVOnce rgb2rgb_once = AV_ONCE_INIT;
1313 
1314     if (c->nb_threads != 1) {
1315         ret = context_init_threaded(c, srcFilter, dstFilter);
1316         if (ret < 0 || c->nb_threads > 1)
1317             return ret;
1318         // threading disabled in this build, init as single-threaded
1319     }
1320 
1321     cpu_flags = av_get_cpu_flags();
1322     flags     = c->flags;
1323     emms_c();
1324     if (ff_thread_once(&rgb2rgb_once, ff_sws_rgb2rgb_init) != 0)
1325         return AVERROR_UNKNOWN;
1326 
1327     unscaled = (srcW == dstW && srcH == dstH);
1328 
1329     c->srcRange |= handle_jpeg(&c->srcFormat);
1330     c->dstRange |= handle_jpeg(&c->dstFormat);
1331 
1332     if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat)
1333         av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
1334 
1335     if (!c->contrast && !c->saturation && !c->dstFormatBpp)
1336         sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], c->srcRange,
1337                                  ff_yuv2rgb_coeffs[SWS_CS_DEFAULT],
1338                                  c->dstRange, 0, 1 << 16, 1 << 16);
1339 
1340     handle_formats(c);
1341     srcFormat = c->srcFormat;
1342     dstFormat = c->dstFormat;
1343     desc_src = av_pix_fmt_desc_get(srcFormat);
1344     desc_dst = av_pix_fmt_desc_get(dstFormat);
1345 
1346     // If the source has no alpha then disable alpha blendaway
1347     if (c->src0Alpha)
1348         c->alphablend = SWS_ALPHA_BLEND_NONE;
1349 
1350     if (!(unscaled && sws_isSupportedEndiannessConversion(srcFormat) &&
1351           av_pix_fmt_swap_endianness(srcFormat) == dstFormat)) {
1352     if (!sws_isSupportedInput(srcFormat)) {
1353         av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
1354                av_get_pix_fmt_name(srcFormat));
1355         return AVERROR(EINVAL);
1356     }
1357     if (!sws_isSupportedOutput(dstFormat)) {
1358         av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
1359                av_get_pix_fmt_name(dstFormat));
1360         return AVERROR(EINVAL);
1361     }
1362     }
1363     av_assert2(desc_src && desc_dst);
1364 
1365     i = flags & (SWS_POINT         |
1366                  SWS_AREA          |
1367                  SWS_BILINEAR      |
1368                  SWS_FAST_BILINEAR |
1369                  SWS_BICUBIC       |
1370                  SWS_X             |
1371                  SWS_GAUSS         |
1372                  SWS_LANCZOS       |
1373                  SWS_SINC          |
1374                  SWS_SPLINE        |
1375                  SWS_BICUBLIN);
1376 
1377     /* provide a default scaler if not set by caller */
1378     if (!i) {
1379         if (dstW < srcW && dstH < srcH)
1380             flags |= SWS_BICUBIC;
1381         else if (dstW > srcW && dstH > srcH)
1382             flags |= SWS_BICUBIC;
1383         else
1384             flags |= SWS_BICUBIC;
1385         c->flags = flags;
1386     } else if (i & (i - 1)) {
1387         av_log(c, AV_LOG_ERROR,
1388                "Exactly one scaler algorithm must be chosen, got %X\n", i);
1389         return AVERROR(EINVAL);
1390     }
1391     /* sanity check */
1392     if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
1393         /* FIXME check if these are enough and try to lower them after
1394          * fixing the relevant parts of the code */
1395         av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
1396                srcW, srcH, dstW, dstH);
1397         return AVERROR(EINVAL);
1398     }
1399     if (flags & SWS_FAST_BILINEAR) {
1400         if (srcW < 8 || dstW < 8) {
1401             flags ^= SWS_FAST_BILINEAR | SWS_BILINEAR;
1402             c->flags = flags;
1403         }
1404     }
1405 
1406     if (!dstFilter)
1407         dstFilter = &dummyFilter;
1408     if (!srcFilter)
1409         srcFilter = &dummyFilter;
1410 
1411     c->lumXInc      = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
1412     c->lumYInc      = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
1413     c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
1414     c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
1415     c->vRounder     = 4 * 0x0001000100010001ULL;
1416 
1417     usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
1418                   (srcFilter->chrV && srcFilter->chrV->length > 1) ||
1419                   (dstFilter->lumV && dstFilter->lumV->length > 1) ||
1420                   (dstFilter->chrV && dstFilter->chrV->length > 1);
1421     usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
1422                   (srcFilter->chrH && srcFilter->chrH->length > 1) ||
1423                   (dstFilter->lumH && dstFilter->lumH->length > 1) ||
1424                   (dstFilter->chrH && dstFilter->chrH->length > 1);
1425 
1426     av_pix_fmt_get_chroma_sub_sample(srcFormat, &c->chrSrcHSubSample, &c->chrSrcVSubSample);
1427     av_pix_fmt_get_chroma_sub_sample(dstFormat, &c->chrDstHSubSample, &c->chrDstVSubSample);
1428 
1429     c->dst_slice_align = 1 << c->chrDstVSubSample;
1430 
1431     if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
1432         if (dstW&1) {
1433             av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
1434             flags |= SWS_FULL_CHR_H_INT;
1435             c->flags = flags;
1436         }
1437 
1438         if (   c->chrSrcHSubSample == 0
1439             && c->chrSrcVSubSample == 0
1440             && c->dither != SWS_DITHER_BAYER //SWS_FULL_CHR_H_INT is currently not supported with SWS_DITHER_BAYER
1441             && !(c->flags & SWS_FAST_BILINEAR)
1442         ) {
1443             av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to input having non subsampled chroma\n");
1444             flags |= SWS_FULL_CHR_H_INT;
1445             c->flags = flags;
1446         }
1447     }
1448 
1449     if (c->dither == SWS_DITHER_AUTO) {
1450         if (flags & SWS_ERROR_DIFFUSION)
1451             c->dither = SWS_DITHER_ED;
1452     }
1453 
1454     if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1455        dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1456        dstFormat == AV_PIX_FMT_BGR8 ||
1457        dstFormat == AV_PIX_FMT_RGB8) {
1458         if (c->dither == SWS_DITHER_AUTO)
1459             c->dither = (flags & SWS_FULL_CHR_H_INT) ? SWS_DITHER_ED : SWS_DITHER_BAYER;
1460         if (!(flags & SWS_FULL_CHR_H_INT)) {
1461             if (c->dither == SWS_DITHER_ED || c->dither == SWS_DITHER_A_DITHER || c->dither == SWS_DITHER_X_DITHER || c->dither == SWS_DITHER_NONE) {
1462                 av_log(c, AV_LOG_DEBUG,
1463                     "Desired dithering only supported in full chroma interpolation for destination format '%s'\n",
1464                     av_get_pix_fmt_name(dstFormat));
1465                 flags   |= SWS_FULL_CHR_H_INT;
1466                 c->flags = flags;
1467             }
1468         }
1469         if (flags & SWS_FULL_CHR_H_INT) {
1470             if (c->dither == SWS_DITHER_BAYER) {
1471                 av_log(c, AV_LOG_DEBUG,
1472                     "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1473                     av_get_pix_fmt_name(dstFormat));
1474                 c->dither = SWS_DITHER_ED;
1475             }
1476         }
1477     }
1478     if (isPlanarRGB(dstFormat)) {
1479         if (!(flags & SWS_FULL_CHR_H_INT)) {
1480             av_log(c, AV_LOG_DEBUG,
1481                    "%s output is not supported with half chroma resolution, switching to full\n",
1482                    av_get_pix_fmt_name(dstFormat));
1483             flags   |= SWS_FULL_CHR_H_INT;
1484             c->flags = flags;
1485         }
1486     }
1487 
1488     /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1489      * chroma interpolation */
1490     if (flags & SWS_FULL_CHR_H_INT &&
1491         isAnyRGB(dstFormat)        &&
1492         !isPlanarRGB(dstFormat)    &&
1493         dstFormat != AV_PIX_FMT_RGBA64LE &&
1494         dstFormat != AV_PIX_FMT_RGBA64BE &&
1495         dstFormat != AV_PIX_FMT_BGRA64LE &&
1496         dstFormat != AV_PIX_FMT_BGRA64BE &&
1497         dstFormat != AV_PIX_FMT_RGB48LE &&
1498         dstFormat != AV_PIX_FMT_RGB48BE &&
1499         dstFormat != AV_PIX_FMT_BGR48LE &&
1500         dstFormat != AV_PIX_FMT_BGR48BE &&
1501         dstFormat != AV_PIX_FMT_RGBA  &&
1502         dstFormat != AV_PIX_FMT_ARGB  &&
1503         dstFormat != AV_PIX_FMT_BGRA  &&
1504         dstFormat != AV_PIX_FMT_ABGR  &&
1505         dstFormat != AV_PIX_FMT_RGB24 &&
1506         dstFormat != AV_PIX_FMT_BGR24 &&
1507         dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1508         dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1509         dstFormat != AV_PIX_FMT_BGR8 &&
1510         dstFormat != AV_PIX_FMT_RGB8
1511     ) {
1512         av_log(c, AV_LOG_WARNING,
1513                "full chroma interpolation for destination format '%s' not yet implemented\n",
1514                av_get_pix_fmt_name(dstFormat));
1515         flags   &= ~SWS_FULL_CHR_H_INT;
1516         c->flags = flags;
1517     }
1518     if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1519         c->chrDstHSubSample = 1;
1520 
1521     // drop some chroma lines if the user wants it
1522     c->vChrDrop          = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1523                            SWS_SRC_V_CHR_DROP_SHIFT;
1524     c->chrSrcVSubSample += c->vChrDrop;
1525 
1526     /* drop every other pixel for chroma calculation unless user
1527      * wants full chroma */
1528     if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP)   &&
1529         srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1530         srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1531         srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1532         srcFormat != AV_PIX_FMT_GBRP9BE   && srcFormat != AV_PIX_FMT_GBRP9LE  &&
1533         srcFormat != AV_PIX_FMT_GBRP10BE  && srcFormat != AV_PIX_FMT_GBRP10LE &&
1534         srcFormat != AV_PIX_FMT_GBRAP10BE && srcFormat != AV_PIX_FMT_GBRAP10LE &&
1535         srcFormat != AV_PIX_FMT_GBRP12BE  && srcFormat != AV_PIX_FMT_GBRP12LE &&
1536         srcFormat != AV_PIX_FMT_GBRAP12BE && srcFormat != AV_PIX_FMT_GBRAP12LE &&
1537         srcFormat != AV_PIX_FMT_GBRP14BE  && srcFormat != AV_PIX_FMT_GBRP14LE &&
1538         srcFormat != AV_PIX_FMT_GBRP16BE  && srcFormat != AV_PIX_FMT_GBRP16LE &&
1539         srcFormat != AV_PIX_FMT_GBRAP16BE  && srcFormat != AV_PIX_FMT_GBRAP16LE &&
1540         srcFormat != AV_PIX_FMT_GBRPF32BE  && srcFormat != AV_PIX_FMT_GBRPF32LE &&
1541         srcFormat != AV_PIX_FMT_GBRAPF32BE && srcFormat != AV_PIX_FMT_GBRAPF32LE &&
1542         ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1543          (flags & SWS_FAST_BILINEAR)))
1544         c->chrSrcHSubSample = 1;
1545 
1546     // Note the AV_CEIL_RSHIFT is so that we always round toward +inf.
1547     c->chrSrcW = AV_CEIL_RSHIFT(srcW, c->chrSrcHSubSample);
1548     c->chrSrcH = AV_CEIL_RSHIFT(srcH, c->chrSrcVSubSample);
1549     c->chrDstW = AV_CEIL_RSHIFT(dstW, c->chrDstHSubSample);
1550     c->chrDstH = AV_CEIL_RSHIFT(dstH, c->chrDstVSubSample);
1551 
1552     if (!FF_ALLOCZ_TYPED_ARRAY(c->formatConvBuffer, FFALIGN(srcW * 2 + 78, 16) * 2))
1553         goto nomem;
1554 
1555     c->frame_src = av_frame_alloc();
1556     c->frame_dst = av_frame_alloc();
1557     if (!c->frame_src || !c->frame_dst)
1558         goto nomem;
1559 
1560     c->srcBpc = desc_src->comp[0].depth;
1561     if (c->srcBpc < 8)
1562         c->srcBpc = 8;
1563     c->dstBpc = desc_dst->comp[0].depth;
1564     if (c->dstBpc < 8)
1565         c->dstBpc = 8;
1566     if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1567         c->srcBpc = 16;
1568     if (c->dstBpc == 16)
1569         dst_stride <<= 1;
1570 
1571     if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1572         c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1573                              c->chrDstW >= c->chrSrcW &&
1574                              (srcW & 15) == 0;
1575         if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1576 
1577             && (flags & SWS_FAST_BILINEAR)) {
1578             if (flags & SWS_PRINT_INFO)
1579                 av_log(c, AV_LOG_INFO,
1580                        "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1581         }
1582         if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1583             c->canMMXEXTBeUsed = 0;
1584     } else
1585         c->canMMXEXTBeUsed = 0;
1586 
1587     c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1588     c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1589 
1590     /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1591      * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1592      * correct scaling.
1593      * n-2 is the last chrominance sample available.
1594      * This is not perfect, but no one should notice the difference, the more
1595      * correct variant would be like the vertical one, but that would require
1596      * some special code for the first and last pixel */
1597     if (flags & SWS_FAST_BILINEAR) {
1598         if (c->canMMXEXTBeUsed) {
1599             c->lumXInc += 20;
1600             c->chrXInc += 20;
1601         }
1602         // we don't use the x86 asm scaler if MMX is available
1603         else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1604             c->lumXInc = ((int64_t)(srcW       - 2) << 16) / (dstW       - 2) - 20;
1605             c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1606         }
1607     }
1608 
1609     // hardcoded for now
1610     c->gamma_value = 2.2;
1611     tmpFmt = AV_PIX_FMT_RGBA64LE;
1612 
1613 
1614     if (!unscaled && c->gamma_flag && (srcFormat != tmpFmt || dstFormat != tmpFmt)) {
1615         SwsContext *c2;
1616         c->cascaded_context[0] = NULL;
1617 
1618         ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1619                             srcW, srcH, tmpFmt, 64);
1620         if (ret < 0)
1621             return ret;
1622 
1623         c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1624                                                 srcW, srcH, tmpFmt,
1625                                                 flags, NULL, NULL, c->param);
1626         if (!c->cascaded_context[0]) {
1627             return AVERROR(ENOMEM);
1628         }
1629 
1630         c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFmt,
1631                                                 dstW, dstH, tmpFmt,
1632                                                 flags, srcFilter, dstFilter, c->param);
1633 
1634         if (!c->cascaded_context[1])
1635             return AVERROR(ENOMEM);
1636 
1637         c2 = c->cascaded_context[1];
1638         c2->is_internal_gamma = 1;
1639         c2->gamma     = alloc_gamma_tbl(    c->gamma_value);
1640         c2->inv_gamma = alloc_gamma_tbl(1.f/c->gamma_value);
1641         if (!c2->gamma || !c2->inv_gamma)
1642             return AVERROR(ENOMEM);
1643 
1644         // is_internal_flag is set after creating the context
1645         // to properly create the gamma convert FilterDescriptor
1646         // we have to re-initialize it
1647         ff_free_filters(c2);
1648         if ((ret = ff_init_filters(c2)) < 0) {
1649             sws_freeContext(c2);
1650             c->cascaded_context[1] = NULL;
1651             return ret;
1652         }
1653 
1654         c->cascaded_context[2] = NULL;
1655         if (dstFormat != tmpFmt) {
1656             ret = av_image_alloc(c->cascaded1_tmp, c->cascaded1_tmpStride,
1657                                 dstW, dstH, tmpFmt, 64);
1658             if (ret < 0)
1659                 return ret;
1660 
1661             c->cascaded_context[2] = sws_getContext(dstW, dstH, tmpFmt,
1662                                                 dstW, dstH, dstFormat,
1663                                                 flags, NULL, NULL, c->param);
1664             if (!c->cascaded_context[2])
1665                 return AVERROR(ENOMEM);
1666         }
1667         return 0;
1668     }
1669 
1670     if (isBayer(srcFormat)) {
1671         if (!unscaled ||
1672             (dstFormat != AV_PIX_FMT_RGB24 && dstFormat != AV_PIX_FMT_YUV420P &&
1673              dstFormat != AV_PIX_FMT_RGB48)) {
1674             enum AVPixelFormat tmpFormat = isBayer16BPS(srcFormat) ? AV_PIX_FMT_RGB48 : AV_PIX_FMT_RGB24;
1675 
1676             ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1677                                 srcW, srcH, tmpFormat, 64);
1678             if (ret < 0)
1679                 return ret;
1680 
1681             c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
1682                                                     srcW, srcH, tmpFormat,
1683                                                     flags, srcFilter, NULL, c->param);
1684             if (!c->cascaded_context[0])
1685                 return AVERROR(ENOMEM);
1686 
1687             c->cascaded_context[1] = sws_getContext(srcW, srcH, tmpFormat,
1688                                                     dstW, dstH, dstFormat,
1689                                                     flags, NULL, dstFilter, c->param);
1690             if (!c->cascaded_context[1])
1691                 return AVERROR(ENOMEM);
1692             return 0;
1693         }
1694     }
1695 
1696     if (unscaled && c->srcBpc == 8 && dstFormat == AV_PIX_FMT_GRAYF32){
1697         for (i = 0; i < 256; ++i){
1698             c->uint2float_lut[i] = (float)i * float_mult;
1699         }
1700     }
1701 
1702     // float will be converted to uint16_t
1703     if ((srcFormat == AV_PIX_FMT_GRAYF32BE || srcFormat == AV_PIX_FMT_GRAYF32LE) &&
1704         (!unscaled || unscaled && dstFormat != srcFormat && (srcFormat != AV_PIX_FMT_GRAYF32 ||
1705         dstFormat != AV_PIX_FMT_GRAY8))){
1706         c->srcBpc = 16;
1707     }
1708 
1709     if (CONFIG_SWSCALE_ALPHA && isALPHA(srcFormat) && !isALPHA(dstFormat)) {
1710         enum AVPixelFormat tmpFormat = alphaless_fmt(srcFormat);
1711 
1712         if (tmpFormat != AV_PIX_FMT_NONE && c->alphablend != SWS_ALPHA_BLEND_NONE) {
1713             if (!unscaled ||
1714                 dstFormat != tmpFormat ||
1715                 usesHFilter || usesVFilter ||
1716                 c->srcRange != c->dstRange
1717             ) {
1718                 c->cascaded_mainindex = 1;
1719                 ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
1720                                      srcW, srcH, tmpFormat, 64);
1721                 if (ret < 0)
1722                     return ret;
1723 
1724                 c->cascaded_context[0] = sws_alloc_set_opts(srcW, srcH, srcFormat,
1725                                                             srcW, srcH, tmpFormat,
1726                                                             flags, c->param);
1727                 if (!c->cascaded_context[0])
1728                     return AVERROR(EINVAL);
1729                 c->cascaded_context[0]->alphablend = c->alphablend;
1730                 ret = sws_init_context(c->cascaded_context[0], NULL , NULL);
1731                 if (ret < 0)
1732                     return ret;
1733 
1734                 c->cascaded_context[1] = sws_alloc_set_opts(srcW, srcH, tmpFormat,
1735                                                             dstW, dstH, dstFormat,
1736                                                             flags, c->param);
1737                 if (!c->cascaded_context[1])
1738                     return AVERROR(EINVAL);
1739 
1740                 c->cascaded_context[1]->srcRange = c->srcRange;
1741                 c->cascaded_context[1]->dstRange = c->dstRange;
1742                 ret = sws_init_context(c->cascaded_context[1], srcFilter , dstFilter);
1743                 if (ret < 0)
1744                     return ret;
1745 
1746                 return 0;
1747             }
1748         }
1749     }
1750 
1751 #if HAVE_MMAP && HAVE_MPROTECT && defined(MAP_ANONYMOUS)
1752 #define USE_MMAP 1
1753 #else
1754 #define USE_MMAP 0
1755 #endif
1756 
1757     /* precalculate horizontal scaler filter coefficients */
1758     {
1759 #if HAVE_MMXEXT_INLINE
1760 // can't downscale !!!
1761         if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1762             c->lumMmxextFilterCodeSize = ff_init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1763                                                              NULL, NULL, 8);
1764             c->chrMmxextFilterCodeSize = ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1765                                                              NULL, NULL, NULL, 4);
1766 
1767 #if USE_MMAP
1768             c->lumMmxextFilterCode = mmap(NULL, c->lumMmxextFilterCodeSize,
1769                                           PROT_READ | PROT_WRITE,
1770                                           MAP_PRIVATE | MAP_ANONYMOUS,
1771                                           -1, 0);
1772             c->chrMmxextFilterCode = mmap(NULL, c->chrMmxextFilterCodeSize,
1773                                           PROT_READ | PROT_WRITE,
1774                                           MAP_PRIVATE | MAP_ANONYMOUS,
1775                                           -1, 0);
1776 #elif HAVE_VIRTUALALLOC
1777             c->lumMmxextFilterCode = VirtualAlloc(NULL,
1778                                                   c->lumMmxextFilterCodeSize,
1779                                                   MEM_COMMIT,
1780                                                   PAGE_EXECUTE_READWRITE);
1781             c->chrMmxextFilterCode = VirtualAlloc(NULL,
1782                                                   c->chrMmxextFilterCodeSize,
1783                                                   MEM_COMMIT,
1784                                                   PAGE_EXECUTE_READWRITE);
1785 #else
1786             c->lumMmxextFilterCode = av_malloc(c->lumMmxextFilterCodeSize);
1787             c->chrMmxextFilterCode = av_malloc(c->chrMmxextFilterCodeSize);
1788 #endif
1789 
1790 #ifdef MAP_ANONYMOUS
1791             if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1792 #else
1793             if (!c->lumMmxextFilterCode || !c->chrMmxextFilterCode)
1794 #endif
1795             {
1796                 av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1797                 return AVERROR(ENOMEM);
1798             }
1799 
1800             if (!FF_ALLOCZ_TYPED_ARRAY(c->hLumFilter,    dstW           / 8 + 8) ||
1801                 !FF_ALLOCZ_TYPED_ARRAY(c->hChrFilter,    c->chrDstW     / 4 + 8) ||
1802                 !FF_ALLOCZ_TYPED_ARRAY(c->hLumFilterPos, dstW       / 2 / 8 + 8) ||
1803                 !FF_ALLOCZ_TYPED_ARRAY(c->hChrFilterPos, c->chrDstW / 2 / 4 + 8))
1804                 goto nomem;
1805 
1806             ff_init_hscaler_mmxext(      dstW, c->lumXInc, c->lumMmxextFilterCode,
1807                                 c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1808             ff_init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1809                                 c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1810 
1811 #if USE_MMAP
1812             if (   mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1
1813                 || mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ) == -1) {
1814                 av_log(c, AV_LOG_ERROR, "mprotect failed, cannot use fast bilinear scaler\n");
1815                 ret = AVERROR(EINVAL);
1816                 goto fail;
1817             }
1818 #endif
1819         } else
1820 #endif /* HAVE_MMXEXT_INLINE */
1821         {
1822             const int filterAlign = X86_MMX(cpu_flags)     ? 4 :
1823                                     PPC_ALTIVEC(cpu_flags) ? 8 :
1824                                     have_neon(cpu_flags)   ? 4 : 1;
1825 
1826             if ((ret = initFilter(&c->hLumFilter, &c->hLumFilterPos,
1827                            &c->hLumFilterSize, c->lumXInc,
1828                            srcW, dstW, filterAlign, 1 << 14,
1829                            (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1830                            cpu_flags, srcFilter->lumH, dstFilter->lumH,
1831                            c->param,
1832                            get_local_pos(c, 0, 0, 0),
1833                            get_local_pos(c, 0, 0, 0))) < 0)
1834                 goto fail;
1835             if (ff_shuffle_filter_coefficients(c, c->hLumFilterPos, c->hLumFilterSize, c->hLumFilter, dstW) < 0)
1836                 goto nomem;
1837             if ((ret = initFilter(&c->hChrFilter, &c->hChrFilterPos,
1838                            &c->hChrFilterSize, c->chrXInc,
1839                            c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1840                            (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1841                            cpu_flags, srcFilter->chrH, dstFilter->chrH,
1842                            c->param,
1843                            get_local_pos(c, c->chrSrcHSubSample, c->src_h_chr_pos, 0),
1844                            get_local_pos(c, c->chrDstHSubSample, c->dst_h_chr_pos, 0))) < 0)
1845                 goto fail;
1846             if (ff_shuffle_filter_coefficients(c, c->hChrFilterPos, c->hChrFilterSize, c->hChrFilter, c->chrDstW) < 0)
1847                 goto nomem;
1848         }
1849     } // initialize horizontal stuff
1850 
1851     /* precalculate vertical scaler filter coefficients */
1852     {
1853         const int filterAlign = X86_MMX(cpu_flags)     ? 2 :
1854                                 PPC_ALTIVEC(cpu_flags) ? 8 :
1855                                 have_neon(cpu_flags)   ? 2 : 1;
1856 
1857         if ((ret = initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize,
1858                        c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1859                        (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1860                        cpu_flags, srcFilter->lumV, dstFilter->lumV,
1861                        c->param,
1862                        get_local_pos(c, 0, 0, 1),
1863                        get_local_pos(c, 0, 0, 1))) < 0)
1864             goto fail;
1865         if ((ret = initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize,
1866                        c->chrYInc, c->chrSrcH, c->chrDstH,
1867                        filterAlign, (1 << 12),
1868                        (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1869                        cpu_flags, srcFilter->chrV, dstFilter->chrV,
1870                        c->param,
1871                        get_local_pos(c, c->chrSrcVSubSample, c->src_v_chr_pos, 1),
1872                        get_local_pos(c, c->chrDstVSubSample, c->dst_v_chr_pos, 1))) < 0)
1873 
1874             goto fail;
1875 
1876 #if HAVE_ALTIVEC
1877         if (!FF_ALLOC_TYPED_ARRAY(c->vYCoeffsBank, c->vLumFilterSize * c->dstH) ||
1878             !FF_ALLOC_TYPED_ARRAY(c->vCCoeffsBank, c->vChrFilterSize * c->chrDstH))
1879             goto nomem;
1880 
1881         for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1882             int j;
1883             short *p = (short *)&c->vYCoeffsBank[i];
1884             for (j = 0; j < 8; j++)
1885                 p[j] = c->vLumFilter[i];
1886         }
1887 
1888         for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1889             int j;
1890             short *p = (short *)&c->vCCoeffsBank[i];
1891             for (j = 0; j < 8; j++)
1892                 p[j] = c->vChrFilter[i];
1893         }
1894 #endif
1895     }
1896 
1897     for (i = 0; i < 4; i++)
1898         if (!FF_ALLOCZ_TYPED_ARRAY(c->dither_error[i], c->dstW + 2))
1899             goto nomem;
1900 
1901     c->needAlpha = (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat)) ? 1 : 0;
1902 
1903     // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1904     c->uv_off   = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1905     c->uv_offx2 = dst_stride + 16;
1906 
1907     av_assert0(c->chrDstH <= dstH);
1908 
1909     if (flags & SWS_PRINT_INFO) {
1910         const char *scaler = NULL, *cpucaps;
1911 
1912         for (i = 0; i < FF_ARRAY_ELEMS(scale_algorithms); i++) {
1913             if (flags & scale_algorithms[i].flag) {
1914                 scaler = scale_algorithms[i].description;
1915                 break;
1916             }
1917         }
1918         if (!scaler)
1919             scaler =  "ehh flags invalid?!";
1920         av_log(c, AV_LOG_INFO, "%s scaler, from %s to %s%s ",
1921                scaler,
1922                av_get_pix_fmt_name(srcFormat),
1923 #ifdef DITHER1XBPP
1924                dstFormat == AV_PIX_FMT_BGR555   || dstFormat == AV_PIX_FMT_BGR565   ||
1925                dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1926                dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1927                                                              "dithered " : "",
1928 #else
1929                "",
1930 #endif
1931                av_get_pix_fmt_name(dstFormat));
1932 
1933         if (INLINE_MMXEXT(cpu_flags))
1934             cpucaps = "MMXEXT";
1935         else if (INLINE_AMD3DNOW(cpu_flags))
1936             cpucaps = "3DNOW";
1937         else if (INLINE_MMX(cpu_flags))
1938             cpucaps = "MMX";
1939         else if (PPC_ALTIVEC(cpu_flags))
1940             cpucaps = "AltiVec";
1941         else
1942             cpucaps = "C";
1943 
1944         av_log(c, AV_LOG_INFO, "using %s\n", cpucaps);
1945 
1946         av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1947         av_log(c, AV_LOG_DEBUG,
1948                "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1949                c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1950         av_log(c, AV_LOG_DEBUG,
1951                "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1952                c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1953                c->chrXInc, c->chrYInc);
1954     }
1955 
1956     /* alpha blend special case, note this has been split via cascaded contexts if its scaled */
1957     if (unscaled && !usesHFilter && !usesVFilter &&
1958         c->alphablend != SWS_ALPHA_BLEND_NONE &&
1959         isALPHA(srcFormat) &&
1960         (c->srcRange == c->dstRange || isAnyRGB(dstFormat)) &&
1961         alphaless_fmt(srcFormat) == dstFormat
1962     ) {
1963         c->convert_unscaled = ff_sws_alphablendaway;
1964 
1965         if (flags & SWS_PRINT_INFO)
1966             av_log(c, AV_LOG_INFO,
1967                     "using alpha blendaway %s -> %s special converter\n",
1968                     av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1969         return 0;
1970     }
1971 
1972     /* unscaled special cases */
1973     if (unscaled && !usesHFilter && !usesVFilter &&
1974         (c->srcRange == c->dstRange || isAnyRGB(dstFormat) ||
1975          isFloat(srcFormat) || isFloat(dstFormat))){
1976         ff_get_unscaled_swscale(c);
1977 
1978         if (c->convert_unscaled) {
1979             if (flags & SWS_PRINT_INFO)
1980                 av_log(c, AV_LOG_INFO,
1981                        "using unscaled %s -> %s special converter\n",
1982                        av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1983             return 0;
1984         }
1985     }
1986 
1987     ff_sws_init_scale(c);
1988 
1989     return ff_init_filters(c);
1990 nomem:
1991     ret = AVERROR(ENOMEM);
1992 fail: // FIXME replace things by appropriate error codes
1993     if (ret == RETCODE_USE_CASCADE)  {
1994         int tmpW = sqrt(srcW * (int64_t)dstW);
1995         int tmpH = sqrt(srcH * (int64_t)dstH);
1996         enum AVPixelFormat tmpFormat = AV_PIX_FMT_YUV420P;
1997 
1998         if (isALPHA(srcFormat))
1999             tmpFormat = AV_PIX_FMT_YUVA420P;
2000 
2001         if (srcW*(int64_t)srcH <= 4LL*dstW*dstH)
2002             return AVERROR(EINVAL);
2003 
2004         ret = av_image_alloc(c->cascaded_tmp, c->cascaded_tmpStride,
2005                              tmpW, tmpH, tmpFormat, 64);
2006         if (ret < 0)
2007             return ret;
2008 
2009         c->cascaded_context[0] = sws_getContext(srcW, srcH, srcFormat,
2010                                                 tmpW, tmpH, tmpFormat,
2011                                                 flags, srcFilter, NULL, c->param);
2012         if (!c->cascaded_context[0])
2013             return AVERROR(ENOMEM);
2014 
2015         c->cascaded_context[1] = sws_getContext(tmpW, tmpH, tmpFormat,
2016                                                 dstW, dstH, dstFormat,
2017                                                 flags, NULL, dstFilter, c->param);
2018         if (!c->cascaded_context[1])
2019             return AVERROR(ENOMEM);
2020         return 0;
2021     }
2022     return ret;
2023 }
2024 
sws_alloc_set_opts(int srcW,int srcH,enum AVPixelFormat srcFormat,int dstW,int dstH,enum AVPixelFormat dstFormat,int flags,const double * param)2025 SwsContext *sws_alloc_set_opts(int srcW, int srcH, enum AVPixelFormat srcFormat,
2026                                int dstW, int dstH, enum AVPixelFormat dstFormat,
2027                                int flags, const double *param)
2028 {
2029     SwsContext *c;
2030 
2031     if (!(c = sws_alloc_context()))
2032         return NULL;
2033 
2034     c->flags     = flags;
2035     c->srcW      = srcW;
2036     c->srcH      = srcH;
2037     c->dstW      = dstW;
2038     c->dstH      = dstH;
2039     c->srcFormat = srcFormat;
2040     c->dstFormat = dstFormat;
2041 
2042     if (param) {
2043         c->param[0] = param[0];
2044         c->param[1] = param[1];
2045     }
2046 
2047     return c;
2048 }
2049 
sws_getContext(int srcW,int srcH,enum AVPixelFormat srcFormat,int dstW,int dstH,enum AVPixelFormat dstFormat,int flags,SwsFilter * srcFilter,SwsFilter * dstFilter,const double * param)2050 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
2051                            int dstW, int dstH, enum AVPixelFormat dstFormat,
2052                            int flags, SwsFilter *srcFilter,
2053                            SwsFilter *dstFilter, const double *param)
2054 {
2055     SwsContext *c;
2056 
2057     c = sws_alloc_set_opts(srcW, srcH, srcFormat,
2058                            dstW, dstH, dstFormat,
2059                            flags, param);
2060     if (!c)
2061         return NULL;
2062 
2063     if (sws_init_context(c, srcFilter, dstFilter) < 0) {
2064         sws_freeContext(c);
2065         return NULL;
2066     }
2067 
2068     return c;
2069 }
2070 
isnan_vec(SwsVector * a)2071 static int isnan_vec(SwsVector *a)
2072 {
2073     int i;
2074     for (i=0; i<a->length; i++)
2075         if (isnan(a->coeff[i]))
2076             return 1;
2077     return 0;
2078 }
2079 
makenan_vec(SwsVector * a)2080 static void makenan_vec(SwsVector *a)
2081 {
2082     int i;
2083     for (i=0; i<a->length; i++)
2084         a->coeff[i] = NAN;
2085 }
2086 
sws_getDefaultFilter(float lumaGBlur,float chromaGBlur,float lumaSharpen,float chromaSharpen,float chromaHShift,float chromaVShift,int verbose)2087 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
2088                                 float lumaSharpen, float chromaSharpen,
2089                                 float chromaHShift, float chromaVShift,
2090                                 int verbose)
2091 {
2092     SwsFilter *filter = av_malloc(sizeof(SwsFilter));
2093     if (!filter)
2094         return NULL;
2095 
2096     if (lumaGBlur != 0.0) {
2097         filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
2098         filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
2099     } else {
2100         filter->lumH = sws_getIdentityVec();
2101         filter->lumV = sws_getIdentityVec();
2102     }
2103 
2104     if (chromaGBlur != 0.0) {
2105         filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
2106         filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
2107     } else {
2108         filter->chrH = sws_getIdentityVec();
2109         filter->chrV = sws_getIdentityVec();
2110     }
2111 
2112     if (!filter->lumH || !filter->lumV || !filter->chrH || !filter->chrV)
2113         goto fail;
2114 
2115     if (chromaSharpen != 0.0) {
2116         SwsVector *id = sws_getIdentityVec();
2117         if (!id)
2118             goto fail;
2119         sws_scaleVec(filter->chrH, -chromaSharpen);
2120         sws_scaleVec(filter->chrV, -chromaSharpen);
2121         sws_addVec(filter->chrH, id);
2122         sws_addVec(filter->chrV, id);
2123         sws_freeVec(id);
2124     }
2125 
2126     if (lumaSharpen != 0.0) {
2127         SwsVector *id = sws_getIdentityVec();
2128         if (!id)
2129             goto fail;
2130         sws_scaleVec(filter->lumH, -lumaSharpen);
2131         sws_scaleVec(filter->lumV, -lumaSharpen);
2132         sws_addVec(filter->lumH, id);
2133         sws_addVec(filter->lumV, id);
2134         sws_freeVec(id);
2135     }
2136 
2137     if (chromaHShift != 0.0)
2138         sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
2139 
2140     if (chromaVShift != 0.0)
2141         sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
2142 
2143     sws_normalizeVec(filter->chrH, 1.0);
2144     sws_normalizeVec(filter->chrV, 1.0);
2145     sws_normalizeVec(filter->lumH, 1.0);
2146     sws_normalizeVec(filter->lumV, 1.0);
2147 
2148     if (isnan_vec(filter->chrH) ||
2149         isnan_vec(filter->chrV) ||
2150         isnan_vec(filter->lumH) ||
2151         isnan_vec(filter->lumV))
2152         goto fail;
2153 
2154     if (verbose)
2155         sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
2156     if (verbose)
2157         sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
2158 
2159     return filter;
2160 
2161 fail:
2162     sws_freeVec(filter->lumH);
2163     sws_freeVec(filter->lumV);
2164     sws_freeVec(filter->chrH);
2165     sws_freeVec(filter->chrV);
2166     av_freep(&filter);
2167     return NULL;
2168 }
2169 
sws_allocVec(int length)2170 SwsVector *sws_allocVec(int length)
2171 {
2172     SwsVector *vec;
2173 
2174     if(length <= 0 || length > INT_MAX/ sizeof(double))
2175         return NULL;
2176 
2177     vec = av_malloc(sizeof(SwsVector));
2178     if (!vec)
2179         return NULL;
2180     vec->length = length;
2181     vec->coeff  = av_malloc(sizeof(double) * length);
2182     if (!vec->coeff)
2183         av_freep(&vec);
2184     return vec;
2185 }
2186 
sws_getGaussianVec(double variance,double quality)2187 SwsVector *sws_getGaussianVec(double variance, double quality)
2188 {
2189     const int length = (int)(variance * quality + 0.5) | 1;
2190     int i;
2191     double middle  = (length - 1) * 0.5;
2192     SwsVector *vec;
2193 
2194     if(variance < 0 || quality < 0)
2195         return NULL;
2196 
2197     vec = sws_allocVec(length);
2198 
2199     if (!vec)
2200         return NULL;
2201 
2202     for (i = 0; i < length; i++) {
2203         double dist = i - middle;
2204         vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
2205                         sqrt(2 * variance * M_PI);
2206     }
2207 
2208     sws_normalizeVec(vec, 1.0);
2209 
2210     return vec;
2211 }
2212 
2213 /**
2214  * Allocate and return a vector with length coefficients, all
2215  * with the same value c.
2216  */
2217 static
sws_getConstVec(double c,int length)2218 SwsVector *sws_getConstVec(double c, int length)
2219 {
2220     int i;
2221     SwsVector *vec = sws_allocVec(length);
2222 
2223     if (!vec)
2224         return NULL;
2225 
2226     for (i = 0; i < length; i++)
2227         vec->coeff[i] = c;
2228 
2229     return vec;
2230 }
2231 
2232 /**
2233  * Allocate and return a vector with just one coefficient, with
2234  * value 1.0.
2235  */
2236 static
sws_getIdentityVec(void)2237 SwsVector *sws_getIdentityVec(void)
2238 {
2239     return sws_getConstVec(1.0, 1);
2240 }
2241 
sws_dcVec(SwsVector * a)2242 static double sws_dcVec(SwsVector *a)
2243 {
2244     int i;
2245     double sum = 0;
2246 
2247     for (i = 0; i < a->length; i++)
2248         sum += a->coeff[i];
2249 
2250     return sum;
2251 }
2252 
sws_scaleVec(SwsVector * a,double scalar)2253 void sws_scaleVec(SwsVector *a, double scalar)
2254 {
2255     int i;
2256 
2257     for (i = 0; i < a->length; i++)
2258         a->coeff[i] *= scalar;
2259 }
2260 
sws_normalizeVec(SwsVector * a,double height)2261 void sws_normalizeVec(SwsVector *a, double height)
2262 {
2263     sws_scaleVec(a, height / sws_dcVec(a));
2264 }
2265 
sws_sumVec(SwsVector * a,SwsVector * b)2266 static SwsVector *sws_sumVec(SwsVector *a, SwsVector *b)
2267 {
2268     int length = FFMAX(a->length, b->length);
2269     int i;
2270     SwsVector *vec = sws_getConstVec(0.0, length);
2271 
2272     if (!vec)
2273         return NULL;
2274 
2275     for (i = 0; i < a->length; i++)
2276         vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
2277     for (i = 0; i < b->length; i++)
2278         vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
2279 
2280     return vec;
2281 }
2282 
2283 /* shift left / or right if "shift" is negative */
sws_getShiftedVec(SwsVector * a,int shift)2284 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
2285 {
2286     int length = a->length + FFABS(shift) * 2;
2287     int i;
2288     SwsVector *vec = sws_getConstVec(0.0, length);
2289 
2290     if (!vec)
2291         return NULL;
2292 
2293     for (i = 0; i < a->length; i++) {
2294         vec->coeff[i + (length    - 1) / 2 -
2295                        (a->length - 1) / 2 - shift] = a->coeff[i];
2296     }
2297 
2298     return vec;
2299 }
2300 
2301 static
sws_shiftVec(SwsVector * a,int shift)2302 void sws_shiftVec(SwsVector *a, int shift)
2303 {
2304     SwsVector *shifted = sws_getShiftedVec(a, shift);
2305     if (!shifted) {
2306         makenan_vec(a);
2307         return;
2308     }
2309     av_free(a->coeff);
2310     a->coeff  = shifted->coeff;
2311     a->length = shifted->length;
2312     av_free(shifted);
2313 }
2314 
2315 static
sws_addVec(SwsVector * a,SwsVector * b)2316 void sws_addVec(SwsVector *a, SwsVector *b)
2317 {
2318     SwsVector *sum = sws_sumVec(a, b);
2319     if (!sum) {
2320         makenan_vec(a);
2321         return;
2322     }
2323     av_free(a->coeff);
2324     a->coeff  = sum->coeff;
2325     a->length = sum->length;
2326     av_free(sum);
2327 }
2328 
2329 /**
2330  * Print with av_log() a textual representation of the vector a
2331  * if log_level <= av_log_level.
2332  */
2333 static
sws_printVec2(SwsVector * a,AVClass * log_ctx,int log_level)2334 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
2335 {
2336     int i;
2337     double max = 0;
2338     double min = 0;
2339     double range;
2340 
2341     for (i = 0; i < a->length; i++)
2342         if (a->coeff[i] > max)
2343             max = a->coeff[i];
2344 
2345     for (i = 0; i < a->length; i++)
2346         if (a->coeff[i] < min)
2347             min = a->coeff[i];
2348 
2349     range = max - min;
2350 
2351     for (i = 0; i < a->length; i++) {
2352         int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
2353         av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
2354         for (; x > 0; x--)
2355             av_log(log_ctx, log_level, " ");
2356         av_log(log_ctx, log_level, "|\n");
2357     }
2358 }
2359 
sws_freeVec(SwsVector * a)2360 void sws_freeVec(SwsVector *a)
2361 {
2362     if (!a)
2363         return;
2364     av_freep(&a->coeff);
2365     a->length = 0;
2366     av_free(a);
2367 }
2368 
sws_freeFilter(SwsFilter * filter)2369 void sws_freeFilter(SwsFilter *filter)
2370 {
2371     if (!filter)
2372         return;
2373 
2374     sws_freeVec(filter->lumH);
2375     sws_freeVec(filter->lumV);
2376     sws_freeVec(filter->chrH);
2377     sws_freeVec(filter->chrV);
2378     av_free(filter);
2379 }
2380 
sws_freeContext(SwsContext * c)2381 void sws_freeContext(SwsContext *c)
2382 {
2383     int i;
2384     if (!c)
2385         return;
2386 
2387     for (i = 0; i < c->nb_slice_ctx; i++)
2388         sws_freeContext(c->slice_ctx[i]);
2389     av_freep(&c->slice_ctx);
2390     av_freep(&c->slice_err);
2391 
2392     avpriv_slicethread_free(&c->slicethread);
2393 
2394     for (i = 0; i < 4; i++)
2395         av_freep(&c->dither_error[i]);
2396 
2397     av_frame_free(&c->frame_src);
2398     av_frame_free(&c->frame_dst);
2399 
2400     av_freep(&c->src_ranges.ranges);
2401 
2402     av_freep(&c->vLumFilter);
2403     av_freep(&c->vChrFilter);
2404     av_freep(&c->hLumFilter);
2405     av_freep(&c->hChrFilter);
2406 #if HAVE_ALTIVEC
2407     av_freep(&c->vYCoeffsBank);
2408     av_freep(&c->vCCoeffsBank);
2409 #endif
2410 
2411     av_freep(&c->vLumFilterPos);
2412     av_freep(&c->vChrFilterPos);
2413     av_freep(&c->hLumFilterPos);
2414     av_freep(&c->hChrFilterPos);
2415 
2416 #if HAVE_MMX_INLINE
2417 #if USE_MMAP
2418     if (c->lumMmxextFilterCode)
2419         munmap(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize);
2420     if (c->chrMmxextFilterCode)
2421         munmap(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize);
2422 #elif HAVE_VIRTUALALLOC
2423     if (c->lumMmxextFilterCode)
2424         VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
2425     if (c->chrMmxextFilterCode)
2426         VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
2427 #else
2428     av_free(c->lumMmxextFilterCode);
2429     av_free(c->chrMmxextFilterCode);
2430 #endif
2431     c->lumMmxextFilterCode = NULL;
2432     c->chrMmxextFilterCode = NULL;
2433 #endif /* HAVE_MMX_INLINE */
2434 
2435     av_freep(&c->yuvTable);
2436     av_freep(&c->formatConvBuffer);
2437 
2438     sws_freeContext(c->cascaded_context[0]);
2439     sws_freeContext(c->cascaded_context[1]);
2440     sws_freeContext(c->cascaded_context[2]);
2441     memset(c->cascaded_context, 0, sizeof(c->cascaded_context));
2442     av_freep(&c->cascaded_tmp[0]);
2443     av_freep(&c->cascaded1_tmp[0]);
2444 
2445     av_freep(&c->gamma);
2446     av_freep(&c->inv_gamma);
2447 
2448     av_freep(&c->rgb0_scratch);
2449     av_freep(&c->xyz_scratch);
2450 
2451     ff_free_filters(c);
2452 
2453     av_free(c);
2454 }
2455 
sws_getCachedContext(struct SwsContext * context,int srcW,int srcH,enum AVPixelFormat srcFormat,int dstW,int dstH,enum AVPixelFormat dstFormat,int flags,SwsFilter * srcFilter,SwsFilter * dstFilter,const double * param)2456 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
2457                                         int srcH, enum AVPixelFormat srcFormat,
2458                                         int dstW, int dstH,
2459                                         enum AVPixelFormat dstFormat, int flags,
2460                                         SwsFilter *srcFilter,
2461                                         SwsFilter *dstFilter,
2462                                         const double *param)
2463 {
2464     static const double default_param[2] = { SWS_PARAM_DEFAULT,
2465                                              SWS_PARAM_DEFAULT };
2466     int64_t src_h_chr_pos = -513, dst_h_chr_pos = -513,
2467             src_v_chr_pos = -513, dst_v_chr_pos = -513;
2468 
2469     if (!param)
2470         param = default_param;
2471 
2472     if (context &&
2473         (context->srcW      != srcW      ||
2474          context->srcH      != srcH      ||
2475          context->srcFormat != srcFormat ||
2476          context->dstW      != dstW      ||
2477          context->dstH      != dstH      ||
2478          context->dstFormat != dstFormat ||
2479          context->flags     != flags     ||
2480          context->param[0]  != param[0]  ||
2481          context->param[1]  != param[1])) {
2482 
2483         av_opt_get_int(context, "src_h_chr_pos", 0, &src_h_chr_pos);
2484         av_opt_get_int(context, "src_v_chr_pos", 0, &src_v_chr_pos);
2485         av_opt_get_int(context, "dst_h_chr_pos", 0, &dst_h_chr_pos);
2486         av_opt_get_int(context, "dst_v_chr_pos", 0, &dst_v_chr_pos);
2487         sws_freeContext(context);
2488         context = NULL;
2489     }
2490 
2491     if (!context) {
2492         if (!(context = sws_alloc_context()))
2493             return NULL;
2494         context->srcW      = srcW;
2495         context->srcH      = srcH;
2496         context->srcFormat = srcFormat;
2497         context->dstW      = dstW;
2498         context->dstH      = dstH;
2499         context->dstFormat = dstFormat;
2500         context->flags     = flags;
2501         context->param[0]  = param[0];
2502         context->param[1]  = param[1];
2503 
2504         av_opt_set_int(context, "src_h_chr_pos", src_h_chr_pos, 0);
2505         av_opt_set_int(context, "src_v_chr_pos", src_v_chr_pos, 0);
2506         av_opt_set_int(context, "dst_h_chr_pos", dst_h_chr_pos, 0);
2507         av_opt_set_int(context, "dst_v_chr_pos", dst_v_chr_pos, 0);
2508 
2509         if (sws_init_context(context, srcFilter, dstFilter) < 0) {
2510             sws_freeContext(context);
2511             return NULL;
2512         }
2513     }
2514     return context;
2515 }
2516 
ff_range_add(RangeList * rl,unsigned int start,unsigned int len)2517 int ff_range_add(RangeList *rl, unsigned int start, unsigned int len)
2518 {
2519     Range *tmp;
2520     unsigned int idx;
2521 
2522     /* find the first existing range after the new one */
2523     for (idx = 0; idx < rl->nb_ranges; idx++)
2524         if (rl->ranges[idx].start > start)
2525             break;
2526 
2527     /* check for overlap */
2528     if (idx > 0) {
2529         Range *prev = &rl->ranges[idx - 1];
2530         if (prev->start + prev->len > start)
2531             return AVERROR(EINVAL);
2532     }
2533     if (idx < rl->nb_ranges) {
2534         Range *next = &rl->ranges[idx];
2535         if (start + len > next->start)
2536             return AVERROR(EINVAL);
2537     }
2538 
2539     tmp = av_fast_realloc(rl->ranges, &rl->ranges_allocated,
2540                           (rl->nb_ranges + 1) * sizeof(*rl->ranges));
2541     if (!tmp)
2542         return AVERROR(ENOMEM);
2543     rl->ranges = tmp;
2544 
2545     memmove(rl->ranges + idx + 1, rl->ranges + idx,
2546             sizeof(*rl->ranges) * (rl->nb_ranges - idx));
2547     rl->ranges[idx].start = start;
2548     rl->ranges[idx].len   = len;
2549     rl->nb_ranges++;
2550 
2551     /* merge ranges */
2552     if (idx > 0) {
2553         Range *prev = &rl->ranges[idx - 1];
2554         Range *cur  = &rl->ranges[idx];
2555         if (prev->start + prev->len == cur->start) {
2556             prev->len += cur->len;
2557             memmove(rl->ranges + idx - 1, rl->ranges + idx,
2558                     sizeof(*rl->ranges) * (rl->nb_ranges - idx));
2559             rl->nb_ranges--;
2560             idx--;
2561         }
2562     }
2563     if (idx < rl->nb_ranges - 1) {
2564         Range *cur  = &rl->ranges[idx];
2565         Range *next = &rl->ranges[idx + 1];
2566         if (cur->start + cur->len == next->start) {
2567             cur->len += next->len;
2568             memmove(rl->ranges + idx, rl->ranges + idx + 1,
2569                     sizeof(*rl->ranges) * (rl->nb_ranges - idx - 1));
2570             rl->nb_ranges--;
2571         }
2572     }
2573 
2574     return 0;
2575 }
2576