FFmpeg  1.2.12
utils.c
Go to the documentation of this file.
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 _SVID_SOURCE // needed for MAP_ANONYMOUS
24 #define _DARWIN_C_SOURCE // needed for MAP_ANON
25 #include <inttypes.h>
26 #include <math.h>
27 #include <stdio.h>
28 #include <string.h>
29 #if HAVE_SYS_MMAN_H
30 #include <sys/mman.h>
31 #if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
32 #define MAP_ANONYMOUS MAP_ANON
33 #endif
34 #endif
35 #if HAVE_VIRTUALALLOC
36 #define WIN32_LEAN_AND_MEAN
37 #include <windows.h>
38 #endif
39 
40 #include "libavutil/attributes.h"
41 #include "libavutil/avassert.h"
42 #include "libavutil/avutil.h"
43 #include "libavutil/bswap.h"
44 #include "libavutil/cpu.h"
45 #include "libavutil/intreadwrite.h"
46 #include "libavutil/mathematics.h"
47 #include "libavutil/opt.h"
48 #include "libavutil/pixdesc.h"
49 #include "libavutil/x86/asm.h"
50 #include "libavutil/x86/cpu.h"
51 #include "rgb2rgb.h"
52 #include "swscale.h"
53 #include "swscale_internal.h"
54 
55 unsigned swscale_version(void)
56 {
59 }
60 
61 const char *swscale_configuration(void)
62 {
63  return FFMPEG_CONFIGURATION;
64 }
65 
66 const char *swscale_license(void)
67 {
68 #define LICENSE_PREFIX "libswscale license: "
69  return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
70 }
71 
72 #define RET 0xC3 // near return opcode for x86
73 
74 typedef struct FormatEntry {
76 } FormatEntry;
77 
79  [AV_PIX_FMT_YUV420P] = { 1, 1 },
80  [AV_PIX_FMT_YUYV422] = { 1, 1 },
81  [AV_PIX_FMT_RGB24] = { 1, 1 },
82  [AV_PIX_FMT_BGR24] = { 1, 1 },
83  [AV_PIX_FMT_YUV422P] = { 1, 1 },
84  [AV_PIX_FMT_YUV444P] = { 1, 1 },
85  [AV_PIX_FMT_YUV410P] = { 1, 1 },
86  [AV_PIX_FMT_YUV411P] = { 1, 1 },
87  [AV_PIX_FMT_GRAY8] = { 1, 1 },
88  [AV_PIX_FMT_MONOWHITE] = { 1, 1 },
89  [AV_PIX_FMT_MONOBLACK] = { 1, 1 },
90  [AV_PIX_FMT_PAL8] = { 1, 0 },
91  [AV_PIX_FMT_YUVJ420P] = { 1, 1 },
92  [AV_PIX_FMT_YUVJ422P] = { 1, 1 },
93  [AV_PIX_FMT_YUVJ444P] = { 1, 1 },
94  [AV_PIX_FMT_UYVY422] = { 1, 1 },
95  [AV_PIX_FMT_UYYVYY411] = { 0, 0 },
96  [AV_PIX_FMT_BGR8] = { 1, 1 },
97  [AV_PIX_FMT_BGR4] = { 0, 1 },
98  [AV_PIX_FMT_BGR4_BYTE] = { 1, 1 },
99  [AV_PIX_FMT_RGB8] = { 1, 1 },
100  [AV_PIX_FMT_RGB4] = { 0, 1 },
101  [AV_PIX_FMT_RGB4_BYTE] = { 1, 1 },
102  [AV_PIX_FMT_NV12] = { 1, 1 },
103  [AV_PIX_FMT_NV21] = { 1, 1 },
104  [AV_PIX_FMT_ARGB] = { 1, 1 },
105  [AV_PIX_FMT_RGBA] = { 1, 1 },
106  [AV_PIX_FMT_ABGR] = { 1, 1 },
107  [AV_PIX_FMT_BGRA] = { 1, 1 },
108  [AV_PIX_FMT_0RGB] = { 1, 1 },
109  [AV_PIX_FMT_RGB0] = { 1, 1 },
110  [AV_PIX_FMT_0BGR] = { 1, 1 },
111  [AV_PIX_FMT_BGR0] = { 1, 1 },
112  [AV_PIX_FMT_GRAY16BE] = { 1, 1 },
113  [AV_PIX_FMT_GRAY16LE] = { 1, 1 },
114  [AV_PIX_FMT_YUV440P] = { 1, 1 },
115  [AV_PIX_FMT_YUVJ440P] = { 1, 1 },
116  [AV_PIX_FMT_YUVA420P] = { 1, 1 },
117  [AV_PIX_FMT_YUVA422P] = { 1, 1 },
118  [AV_PIX_FMT_YUVA444P] = { 1, 1 },
119  [AV_PIX_FMT_YUVA420P9BE] = { 1, 1 },
120  [AV_PIX_FMT_YUVA420P9LE] = { 1, 1 },
121  [AV_PIX_FMT_YUVA422P9BE] = { 1, 1 },
122  [AV_PIX_FMT_YUVA422P9LE] = { 1, 1 },
123  [AV_PIX_FMT_YUVA444P9BE] = { 1, 1 },
124  [AV_PIX_FMT_YUVA444P9LE] = { 1, 1 },
125  [AV_PIX_FMT_YUVA420P10BE]= { 1, 1 },
126  [AV_PIX_FMT_YUVA420P10LE]= { 1, 1 },
127  [AV_PIX_FMT_YUVA422P10BE]= { 1, 1 },
128  [AV_PIX_FMT_YUVA422P10LE]= { 1, 1 },
129  [AV_PIX_FMT_YUVA444P10BE]= { 1, 1 },
130  [AV_PIX_FMT_YUVA444P10LE]= { 1, 1 },
131  [AV_PIX_FMT_YUVA420P16BE]= { 1, 1 },
132  [AV_PIX_FMT_YUVA420P16LE]= { 1, 1 },
133  [AV_PIX_FMT_YUVA422P16BE]= { 1, 1 },
134  [AV_PIX_FMT_YUVA422P16LE]= { 1, 1 },
135  [AV_PIX_FMT_YUVA444P16BE]= { 1, 1 },
136  [AV_PIX_FMT_YUVA444P16LE]= { 1, 1 },
137  [AV_PIX_FMT_RGB48BE] = { 1, 1 },
138  [AV_PIX_FMT_RGB48LE] = { 1, 1 },
139  [AV_PIX_FMT_RGBA64BE] = { 1, 0 },
140  [AV_PIX_FMT_RGBA64LE] = { 1, 0 },
141  [AV_PIX_FMT_RGB565BE] = { 1, 1 },
142  [AV_PIX_FMT_RGB565LE] = { 1, 1 },
143  [AV_PIX_FMT_RGB555BE] = { 1, 1 },
144  [AV_PIX_FMT_RGB555LE] = { 1, 1 },
145  [AV_PIX_FMT_BGR565BE] = { 1, 1 },
146  [AV_PIX_FMT_BGR565LE] = { 1, 1 },
147  [AV_PIX_FMT_BGR555BE] = { 1, 1 },
148  [AV_PIX_FMT_BGR555LE] = { 1, 1 },
149  [AV_PIX_FMT_YUV420P16LE] = { 1, 1 },
150  [AV_PIX_FMT_YUV420P16BE] = { 1, 1 },
151  [AV_PIX_FMT_YUV422P16LE] = { 1, 1 },
152  [AV_PIX_FMT_YUV422P16BE] = { 1, 1 },
153  [AV_PIX_FMT_YUV444P16LE] = { 1, 1 },
154  [AV_PIX_FMT_YUV444P16BE] = { 1, 1 },
155  [AV_PIX_FMT_RGB444LE] = { 1, 1 },
156  [AV_PIX_FMT_RGB444BE] = { 1, 1 },
157  [AV_PIX_FMT_BGR444LE] = { 1, 1 },
158  [AV_PIX_FMT_BGR444BE] = { 1, 1 },
159  [AV_PIX_FMT_Y400A] = { 1, 0 },
160  [AV_PIX_FMT_BGR48BE] = { 1, 1 },
161  [AV_PIX_FMT_BGR48LE] = { 1, 1 },
162  [AV_PIX_FMT_BGRA64BE] = { 0, 0 },
163  [AV_PIX_FMT_BGRA64LE] = { 0, 0 },
164  [AV_PIX_FMT_YUV420P9BE] = { 1, 1 },
165  [AV_PIX_FMT_YUV420P9LE] = { 1, 1 },
166  [AV_PIX_FMT_YUV420P10BE] = { 1, 1 },
167  [AV_PIX_FMT_YUV420P10LE] = { 1, 1 },
168  [AV_PIX_FMT_YUV420P12BE] = { 1, 1 },
169  [AV_PIX_FMT_YUV420P12LE] = { 1, 1 },
170  [AV_PIX_FMT_YUV420P14BE] = { 1, 1 },
171  [AV_PIX_FMT_YUV420P14LE] = { 1, 1 },
172  [AV_PIX_FMT_YUV422P9BE] = { 1, 1 },
173  [AV_PIX_FMT_YUV422P9LE] = { 1, 1 },
174  [AV_PIX_FMT_YUV422P10BE] = { 1, 1 },
175  [AV_PIX_FMT_YUV422P10LE] = { 1, 1 },
176  [AV_PIX_FMT_YUV422P12BE] = { 1, 1 },
177  [AV_PIX_FMT_YUV422P12LE] = { 1, 1 },
178  [AV_PIX_FMT_YUV422P14BE] = { 1, 1 },
179  [AV_PIX_FMT_YUV422P14LE] = { 1, 1 },
180  [AV_PIX_FMT_YUV444P9BE] = { 1, 1 },
181  [AV_PIX_FMT_YUV444P9LE] = { 1, 1 },
182  [AV_PIX_FMT_YUV444P10BE] = { 1, 1 },
183  [AV_PIX_FMT_YUV444P10LE] = { 1, 1 },
184  [AV_PIX_FMT_YUV444P12BE] = { 1, 1 },
185  [AV_PIX_FMT_YUV444P12LE] = { 1, 1 },
186  [AV_PIX_FMT_YUV444P14BE] = { 1, 1 },
187  [AV_PIX_FMT_YUV444P14LE] = { 1, 1 },
188  [AV_PIX_FMT_GBRP] = { 1, 1 },
189  [AV_PIX_FMT_GBRP9LE] = { 1, 1 },
190  [AV_PIX_FMT_GBRP9BE] = { 1, 1 },
191  [AV_PIX_FMT_GBRP10LE] = { 1, 1 },
192  [AV_PIX_FMT_GBRP10BE] = { 1, 1 },
193  [AV_PIX_FMT_GBRP12LE] = { 1, 1 },
194  [AV_PIX_FMT_GBRP12BE] = { 1, 1 },
195  [AV_PIX_FMT_GBRP14LE] = { 1, 1 },
196  [AV_PIX_FMT_GBRP14BE] = { 1, 1 },
197  [AV_PIX_FMT_GBRP16LE] = { 1, 0 },
198  [AV_PIX_FMT_GBRP16BE] = { 1, 0 },
199 };
200 
202 {
203  return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
204  format_entries[pix_fmt].is_supported_in : 0;
205 }
206 
208 {
209  return (unsigned)pix_fmt < AV_PIX_FMT_NB ?
210  format_entries[pix_fmt].is_supported_out : 0;
211 }
212 
213 extern const int32_t ff_yuv2rgb_coeffs[8][4];
214 
215 #if FF_API_SWS_FORMAT_NAME
216 const char *sws_format_name(enum AVPixelFormat format)
217 {
218  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);
219  if (desc)
220  return desc->name;
221  else
222  return "Unknown format";
223 }
224 #endif
225 
226 static double getSplineCoeff(double a, double b, double c, double d,
227  double dist)
228 {
229  if (dist <= 1.0)
230  return ((d * dist + c) * dist + b) * dist + a;
231  else
232  return getSplineCoeff(0.0,
233  b + 2.0 * c + 3.0 * d,
234  c + 3.0 * d,
235  -b - 3.0 * c - 6.0 * d,
236  dist - 1.0);
237 }
238 
239 static int initFilter(int16_t **outFilter, int32_t **filterPos,
240  int *outFilterSize, int xInc, int srcW, int dstW,
241  int filterAlign, int one, int flags, int cpu_flags,
242  SwsVector *srcFilter, SwsVector *dstFilter,
243  double param[2])
244 {
245  int i;
246  int filterSize;
247  int filter2Size;
248  int minFilterSize;
249  int64_t *filter = NULL;
250  int64_t *filter2 = NULL;
251  const int64_t fone = 1LL << (54 - FFMIN(av_log2(srcW/dstW), 8));
252  int ret = -1;
253 
254  emms_c(); // FIXME should not be required but IS (even for non-MMX versions)
255 
256  // NOTE: the +3 is for the MMX(+1) / SSE(+3) scaler which reads over the end
257  FF_ALLOC_OR_GOTO(NULL, *filterPos, (dstW + 3) * sizeof(**filterPos), fail);
258 
259  if (FFABS(xInc - 0x10000) < 10) { // unscaled
260  int i;
261  filterSize = 1;
262  FF_ALLOCZ_OR_GOTO(NULL, filter,
263  dstW * sizeof(*filter) * filterSize, fail);
264 
265  for (i = 0; i < dstW; i++) {
266  filter[i * filterSize] = fone;
267  (*filterPos)[i] = i;
268  }
269  } else if (flags & SWS_POINT) { // lame looking point sampling mode
270  int i;
271  int64_t xDstInSrc;
272  filterSize = 1;
273  FF_ALLOC_OR_GOTO(NULL, filter,
274  dstW * sizeof(*filter) * filterSize, fail);
275 
276  xDstInSrc = xInc / 2 - 0x8000;
277  for (i = 0; i < dstW; i++) {
278  int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
279 
280  (*filterPos)[i] = xx;
281  filter[i] = fone;
282  xDstInSrc += xInc;
283  }
284  } else if ((xInc <= (1 << 16) && (flags & SWS_AREA)) ||
285  (flags & SWS_FAST_BILINEAR)) { // bilinear upscale
286  int i;
287  int64_t xDstInSrc;
288  filterSize = 2;
289  FF_ALLOC_OR_GOTO(NULL, filter,
290  dstW * sizeof(*filter) * filterSize, fail);
291 
292  xDstInSrc = xInc / 2 - 0x8000;
293  for (i = 0; i < dstW; i++) {
294  int xx = (xDstInSrc - ((filterSize - 1) << 15) + (1 << 15)) >> 16;
295  int j;
296 
297  (*filterPos)[i] = xx;
298  // bilinear upscale / linear interpolate / area averaging
299  for (j = 0; j < filterSize; j++) {
300  int64_t coeff= fone - FFABS(((int64_t)xx<<16) - xDstInSrc)*(fone>>16);
301  if (coeff < 0)
302  coeff = 0;
303  filter[i * filterSize + j] = coeff;
304  xx++;
305  }
306  xDstInSrc += xInc;
307  }
308  } else {
309  int64_t xDstInSrc;
310  int sizeFactor;
311 
312  if (flags & SWS_BICUBIC)
313  sizeFactor = 4;
314  else if (flags & SWS_X)
315  sizeFactor = 8;
316  else if (flags & SWS_AREA)
317  sizeFactor = 1; // downscale only, for upscale it is bilinear
318  else if (flags & SWS_GAUSS)
319  sizeFactor = 8; // infinite ;)
320  else if (flags & SWS_LANCZOS)
321  sizeFactor = param[0] != SWS_PARAM_DEFAULT ? ceil(2 * param[0]) : 6;
322  else if (flags & SWS_SINC)
323  sizeFactor = 20; // infinite ;)
324  else if (flags & SWS_SPLINE)
325  sizeFactor = 20; // infinite ;)
326  else if (flags & SWS_BILINEAR)
327  sizeFactor = 2;
328  else {
329  av_assert0(0);
330  }
331 
332  if (xInc <= 1 << 16)
333  filterSize = 1 + sizeFactor; // upscale
334  else
335  filterSize = 1 + (sizeFactor * srcW + dstW - 1) / dstW;
336 
337  filterSize = FFMIN(filterSize, srcW - 2);
338  filterSize = FFMAX(filterSize, 1);
339 
340  FF_ALLOC_OR_GOTO(NULL, filter,
341  dstW * sizeof(*filter) * filterSize, fail);
342 
343  xDstInSrc = xInc - 0x10000;
344  for (i = 0; i < dstW; i++) {
345  int xx = (xDstInSrc - ((filterSize - 2) << 16)) / (1 << 17);
346  int j;
347  (*filterPos)[i] = xx;
348  for (j = 0; j < filterSize; j++) {
349  int64_t d = (FFABS(((int64_t)xx << 17) - xDstInSrc)) << 13;
350  double floatd;
351  int64_t coeff;
352 
353  if (xInc > 1 << 16)
354  d = d * dstW / srcW;
355  floatd = d * (1.0 / (1 << 30));
356 
357  if (flags & SWS_BICUBIC) {
358  int64_t B = (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1 << 24);
359  int64_t C = (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1 << 24);
360 
361  if (d >= 1LL << 31) {
362  coeff = 0.0;
363  } else {
364  int64_t dd = (d * d) >> 30;
365  int64_t ddd = (dd * d) >> 30;
366 
367  if (d < 1LL << 30)
368  coeff = (12 * (1 << 24) - 9 * B - 6 * C) * ddd +
369  (-18 * (1 << 24) + 12 * B + 6 * C) * dd +
370  (6 * (1 << 24) - 2 * B) * (1 << 30);
371  else
372  coeff = (-B - 6 * C) * ddd +
373  (6 * B + 30 * C) * dd +
374  (-12 * B - 48 * C) * d +
375  (8 * B + 24 * C) * (1 << 30);
376  }
377  coeff /= (1LL<<54)/fone;
378  }
379 #if 0
380  else if (flags & SWS_X) {
381  double p = param ? param * 0.01 : 0.3;
382  coeff = d ? sin(d * M_PI) / (d * M_PI) : 1.0;
383  coeff *= pow(2.0, -p * d * d);
384  }
385 #endif
386  else if (flags & SWS_X) {
387  double A = param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
388  double c;
389 
390  if (floatd < 1.0)
391  c = cos(floatd * M_PI);
392  else
393  c = -1.0;
394  if (c < 0.0)
395  c = -pow(-c, A);
396  else
397  c = pow(c, A);
398  coeff = (c * 0.5 + 0.5) * fone;
399  } else if (flags & SWS_AREA) {
400  int64_t d2 = d - (1 << 29);
401  if (d2 * xInc < -(1LL << (29 + 16)))
402  coeff = 1.0 * (1LL << (30 + 16));
403  else if (d2 * xInc < (1LL << (29 + 16)))
404  coeff = -d2 * xInc + (1LL << (29 + 16));
405  else
406  coeff = 0.0;
407  coeff *= fone >> (30 + 16);
408  } else if (flags & SWS_GAUSS) {
409  double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
410  coeff = (pow(2.0, -p * floatd * floatd)) * fone;
411  } else if (flags & SWS_SINC) {
412  coeff = (d ? sin(floatd * M_PI) / (floatd * M_PI) : 1.0) * fone;
413  } else if (flags & SWS_LANCZOS) {
414  double p = param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
415  coeff = (d ? sin(floatd * M_PI) * sin(floatd * M_PI / p) /
416  (floatd * floatd * M_PI * M_PI / p) : 1.0) * fone;
417  if (floatd > p)
418  coeff = 0;
419  } else if (flags & SWS_BILINEAR) {
420  coeff = (1 << 30) - d;
421  if (coeff < 0)
422  coeff = 0;
423  coeff *= fone >> 30;
424  } else if (flags & SWS_SPLINE) {
425  double p = -2.196152422706632;
426  coeff = getSplineCoeff(1.0, 0.0, p, -p - 1.0, floatd) * fone;
427  } else {
428  av_assert0(0);
429  }
430 
431  filter[i * filterSize + j] = coeff;
432  xx++;
433  }
434  xDstInSrc += 2 * xInc;
435  }
436  }
437 
438  /* apply src & dst Filter to filter -> filter2
439  * av_free(filter);
440  */
441  av_assert0(filterSize > 0);
442  filter2Size = filterSize;
443  if (srcFilter)
444  filter2Size += srcFilter->length - 1;
445  if (dstFilter)
446  filter2Size += dstFilter->length - 1;
447  av_assert0(filter2Size > 0);
448  FF_ALLOCZ_OR_GOTO(NULL, filter2, filter2Size * dstW * sizeof(*filter2), fail);
449 
450  for (i = 0; i < dstW; i++) {
451  int j, k;
452 
453  if (srcFilter) {
454  for (k = 0; k < srcFilter->length; k++) {
455  for (j = 0; j < filterSize; j++)
456  filter2[i * filter2Size + k + j] +=
457  srcFilter->coeff[k] * filter[i * filterSize + j];
458  }
459  } else {
460  for (j = 0; j < filterSize; j++)
461  filter2[i * filter2Size + j] = filter[i * filterSize + j];
462  }
463  // FIXME dstFilter
464 
465  (*filterPos)[i] += (filterSize - 1) / 2 - (filter2Size - 1) / 2;
466  }
467  av_freep(&filter);
468 
469  /* try to reduce the filter-size (step1 find size and shift left) */
470  // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
471  minFilterSize = 0;
472  for (i = dstW - 1; i >= 0; i--) {
473  int min = filter2Size;
474  int j;
475  int64_t cutOff = 0.0;
476 
477  /* get rid of near zero elements on the left by shifting left */
478  for (j = 0; j < filter2Size; j++) {
479  int k;
480  cutOff += FFABS(filter2[i * filter2Size]);
481 
482  if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
483  break;
484 
485  /* preserve monotonicity because the core can't handle the
486  * filter otherwise */
487  if (i < dstW - 1 && (*filterPos)[i] >= (*filterPos)[i + 1])
488  break;
489 
490  // move filter coefficients left
491  for (k = 1; k < filter2Size; k++)
492  filter2[i * filter2Size + k - 1] = filter2[i * filter2Size + k];
493  filter2[i * filter2Size + k - 1] = 0;
494  (*filterPos)[i]++;
495  }
496 
497  cutOff = 0;
498  /* count near zeros on the right */
499  for (j = filter2Size - 1; j > 0; j--) {
500  cutOff += FFABS(filter2[i * filter2Size + j]);
501 
502  if (cutOff > SWS_MAX_REDUCE_CUTOFF * fone)
503  break;
504  min--;
505  }
506 
507  if (min > minFilterSize)
508  minFilterSize = min;
509  }
510 
511  if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) {
512  // we can handle the special case 4, so we don't want to go the full 8
513  if (minFilterSize < 5)
514  filterAlign = 4;
515 
516  /* We really don't want to waste our time doing useless computation, so
517  * fall back on the scalar C code for very small filters.
518  * Vectorizing is worth it only if you have a decent-sized vector. */
519  if (minFilterSize < 3)
520  filterAlign = 1;
521  }
522 
523  if (INLINE_MMX(cpu_flags)) {
524  // special case for unscaled vertical filtering
525  if (minFilterSize == 1 && filterAlign == 2)
526  filterAlign = 1;
527  }
528 
529  av_assert0(minFilterSize > 0);
530  filterSize = (minFilterSize + (filterAlign - 1)) & (~(filterAlign - 1));
531  av_assert0(filterSize > 0);
532  filter = av_malloc(filterSize * dstW * sizeof(*filter));
533  if (filterSize >= MAX_FILTER_SIZE * 16 /
534  ((flags & SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter) {
535  av_log(NULL, AV_LOG_ERROR, "sws: filterSize %d is too large, try less extreem scaling or increase MAX_FILTER_SIZE and recompile\n", filterSize);
536  goto fail;
537  }
538  *outFilterSize = filterSize;
539 
540  if (flags & SWS_PRINT_INFO)
542  "SwScaler: reducing / aligning filtersize %d -> %d\n",
543  filter2Size, filterSize);
544  /* try to reduce the filter-size (step2 reduce it) */
545  for (i = 0; i < dstW; i++) {
546  int j;
547 
548  for (j = 0; j < filterSize; j++) {
549  if (j >= filter2Size)
550  filter[i * filterSize + j] = 0;
551  else
552  filter[i * filterSize + j] = filter2[i * filter2Size + j];
553  if ((flags & SWS_BITEXACT) && j >= minFilterSize)
554  filter[i * filterSize + j] = 0;
555  }
556  }
557 
558  // FIXME try to align filterPos if possible
559 
560  // fix borders
561  for (i = 0; i < dstW; i++) {
562  int j;
563  if ((*filterPos)[i] < 0) {
564  // move filter coefficients left to compensate for filterPos
565  for (j = 1; j < filterSize; j++) {
566  int left = FFMAX(j + (*filterPos)[i], 0);
567  filter[i * filterSize + left] += filter[i * filterSize + j];
568  filter[i * filterSize + j] = 0;
569  }
570  (*filterPos)[i]= 0;
571  }
572 
573  if ((*filterPos)[i] + filterSize > srcW) {
574  int shift = (*filterPos)[i] + FFMIN(filterSize - srcW, 0);
575 
576  // move filter coefficients right to compensate for filterPos
577  for (j = filterSize - 2; j >= 0; j--) {
578  int right = FFMIN(j + shift, filterSize - 1);
579  filter[i * filterSize + right] += filter[i * filterSize + j];
580  filter[i * filterSize + j] = 0;
581  }
582  (*filterPos)[i]-= shift;
583  }
584  }
585 
586  // Note the +1 is for the MMX scaler which reads over the end
587  /* align at 16 for AltiVec (needed by hScale_altivec_real) */
588  FF_ALLOCZ_OR_GOTO(NULL, *outFilter,
589  *outFilterSize * (dstW + 3) * sizeof(int16_t), fail);
590 
591  /* normalize & store in outFilter */
592  for (i = 0; i < dstW; i++) {
593  int j;
594  int64_t error = 0;
595  int64_t sum = 0;
596 
597  for (j = 0; j < filterSize; j++) {
598  sum += filter[i * filterSize + j];
599  }
600  sum = (sum + one / 2) / one;
601  for (j = 0; j < *outFilterSize; j++) {
602  int64_t v = filter[i * filterSize + j] + error;
603  int intV = ROUNDED_DIV(v, sum);
604  (*outFilter)[i * (*outFilterSize) + j] = intV;
605  error = v - intV * sum;
606  }
607  }
608 
609  (*filterPos)[dstW + 0] =
610  (*filterPos)[dstW + 1] =
611  (*filterPos)[dstW + 2] = (*filterPos)[dstW - 1]; /* the MMX/SSE scaler will
612  * read over the end */
613  for (i = 0; i < *outFilterSize; i++) {
614  int k = (dstW - 1) * (*outFilterSize) + i;
615  (*outFilter)[k + 1 * (*outFilterSize)] =
616  (*outFilter)[k + 2 * (*outFilterSize)] =
617  (*outFilter)[k + 3 * (*outFilterSize)] = (*outFilter)[k];
618  }
619 
620  ret = 0;
621 
622 fail:
623  if(ret < 0)
624  av_log(NULL, AV_LOG_ERROR, "sws: initFilter failed\n");
625  av_free(filter);
626  av_free(filter2);
627  return ret;
628 }
629 
630 #if HAVE_MMXEXT_INLINE
631 static int init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
632  int16_t *filter, int32_t *filterPos,
633  int numSplits)
634 {
635  uint8_t *fragmentA;
636  x86_reg imm8OfPShufW1A;
637  x86_reg imm8OfPShufW2A;
638  x86_reg fragmentLengthA;
639  uint8_t *fragmentB;
640  x86_reg imm8OfPShufW1B;
641  x86_reg imm8OfPShufW2B;
642  x86_reg fragmentLengthB;
643  int fragmentPos;
644 
645  int xpos, i;
646 
647  // create an optimized horizontal scaling routine
648  /* This scaler is made of runtime-generated MMXEXT code using specially tuned
649  * pshufw instructions. For every four output pixels, if four input pixels
650  * are enough for the fast bilinear scaling, then a chunk of fragmentB is
651  * used. If five input pixels are needed, then a chunk of fragmentA is used.
652  */
653 
654  // code fragment
655 
656  __asm__ volatile (
657  "jmp 9f \n\t"
658  // Begin
659  "0: \n\t"
660  "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
661  "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
662  "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
663  "punpcklbw %%mm7, %%mm1 \n\t"
664  "punpcklbw %%mm7, %%mm0 \n\t"
665  "pshufw $0xFF, %%mm1, %%mm1 \n\t"
666  "1: \n\t"
667  "pshufw $0xFF, %%mm0, %%mm0 \n\t"
668  "2: \n\t"
669  "psubw %%mm1, %%mm0 \n\t"
670  "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
671  "pmullw %%mm3, %%mm0 \n\t"
672  "psllw $7, %%mm1 \n\t"
673  "paddw %%mm1, %%mm0 \n\t"
674 
675  "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
676 
677  "add $8, %%"REG_a" \n\t"
678  // End
679  "9: \n\t"
680  // "int $3 \n\t"
681  "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
682  "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
683  "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
684  "dec %1 \n\t"
685  "dec %2 \n\t"
686  "sub %0, %1 \n\t"
687  "sub %0, %2 \n\t"
688  "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
689  "sub %0, %3 \n\t"
690 
691 
692  : "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
693  "=r" (fragmentLengthA)
694  );
695 
696  __asm__ volatile (
697  "jmp 9f \n\t"
698  // Begin
699  "0: \n\t"
700  "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
701  "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
702  "punpcklbw %%mm7, %%mm0 \n\t"
703  "pshufw $0xFF, %%mm0, %%mm1 \n\t"
704  "1: \n\t"
705  "pshufw $0xFF, %%mm0, %%mm0 \n\t"
706  "2: \n\t"
707  "psubw %%mm1, %%mm0 \n\t"
708  "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
709  "pmullw %%mm3, %%mm0 \n\t"
710  "psllw $7, %%mm1 \n\t"
711  "paddw %%mm1, %%mm0 \n\t"
712 
713  "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
714 
715  "add $8, %%"REG_a" \n\t"
716  // End
717  "9: \n\t"
718  // "int $3 \n\t"
719  "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
720  "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
721  "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
722  "dec %1 \n\t"
723  "dec %2 \n\t"
724  "sub %0, %1 \n\t"
725  "sub %0, %2 \n\t"
726  "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
727  "sub %0, %3 \n\t"
728 
729 
730  : "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
731  "=r" (fragmentLengthB)
732  );
733 
734  xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers
735  fragmentPos = 0;
736 
737  for (i = 0; i < dstW / numSplits; i++) {
738  int xx = xpos >> 16;
739 
740  if ((i & 3) == 0) {
741  int a = 0;
742  int b = ((xpos + xInc) >> 16) - xx;
743  int c = ((xpos + xInc * 2) >> 16) - xx;
744  int d = ((xpos + xInc * 3) >> 16) - xx;
745  int inc = (d + 1 < 4);
746  uint8_t *fragment = (d + 1 < 4) ? fragmentB : fragmentA;
747  x86_reg imm8OfPShufW1 = (d + 1 < 4) ? imm8OfPShufW1B : imm8OfPShufW1A;
748  x86_reg imm8OfPShufW2 = (d + 1 < 4) ? imm8OfPShufW2B : imm8OfPShufW2A;
749  x86_reg fragmentLength = (d + 1 < 4) ? fragmentLengthB : fragmentLengthA;
750  int maxShift = 3 - (d + inc);
751  int shift = 0;
752 
753  if (filterCode) {
754  filter[i] = ((xpos & 0xFFFF) ^ 0xFFFF) >> 9;
755  filter[i + 1] = (((xpos + xInc) & 0xFFFF) ^ 0xFFFF) >> 9;
756  filter[i + 2] = (((xpos + xInc * 2) & 0xFFFF) ^ 0xFFFF) >> 9;
757  filter[i + 3] = (((xpos + xInc * 3) & 0xFFFF) ^ 0xFFFF) >> 9;
758  filterPos[i / 2] = xx;
759 
760  memcpy(filterCode + fragmentPos, fragment, fragmentLength);
761 
762  filterCode[fragmentPos + imm8OfPShufW1] = (a + inc) |
763  ((b + inc) << 2) |
764  ((c + inc) << 4) |
765  ((d + inc) << 6);
766  filterCode[fragmentPos + imm8OfPShufW2] = a | (b << 2) |
767  (c << 4) |
768  (d << 6);
769 
770  if (i + 4 - inc >= dstW)
771  shift = maxShift; // avoid overread
772  else if ((filterPos[i / 2] & 3) <= maxShift)
773  shift = filterPos[i / 2] & 3; // align
774 
775  if (shift && i >= shift) {
776  filterCode[fragmentPos + imm8OfPShufW1] += 0x55 * shift;
777  filterCode[fragmentPos + imm8OfPShufW2] += 0x55 * shift;
778  filterPos[i / 2] -= shift;
779  }
780  }
781 
782  fragmentPos += fragmentLength;
783 
784  if (filterCode)
785  filterCode[fragmentPos] = RET;
786  }
787  xpos += xInc;
788  }
789  if (filterCode)
790  filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part
791 
792  return fragmentPos + 1;
793 }
794 #endif /* HAVE_MMXEXT_INLINE */
795 
796 static void getSubSampleFactors(int *h, int *v, enum AVPixelFormat format)
797 {
798  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(format);
799  *h = desc->log2_chroma_w;
800  *v = desc->log2_chroma_h;
801 }
802 
803 int sws_setColorspaceDetails(struct SwsContext *c, const int inv_table[4],
804  int srcRange, const int table[4], int dstRange,
805  int brightness, int contrast, int saturation)
806 {
807  const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(c->dstFormat);
808  const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(c->srcFormat);
809  memcpy(c->srcColorspaceTable, inv_table, sizeof(int) * 4);
810  memcpy(c->dstColorspaceTable, table, sizeof(int) * 4);
811 
812  if(!isYUV(c->dstFormat) && !isGray(c->dstFormat))
813  dstRange = 0;
814  if(!isYUV(c->srcFormat) && !isGray(c->srcFormat))
815  srcRange = 0;
816 
817  c->brightness = brightness;
818  c->contrast = contrast;
819  c->saturation = saturation;
820  c->srcRange = srcRange;
821  c->dstRange = dstRange;
822 
823  if (isYUV(c->dstFormat) || isGray(c->dstFormat))
824  return -1;
825 
826  c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
827  c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
828 
829  ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness,
830  contrast, saturation);
831  // FIXME factorize
832 
834  ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness,
835  contrast, saturation);
836  return 0;
837 }
838 
839 int sws_getColorspaceDetails(struct SwsContext *c, int **inv_table,
840  int *srcRange, int **table, int *dstRange,
841  int *brightness, int *contrast, int *saturation)
842 {
843  if (!c || isYUV(c->dstFormat) || isGray(c->dstFormat))
844  return -1;
845 
846  *inv_table = c->srcColorspaceTable;
847  *table = c->dstColorspaceTable;
848  *srcRange = c->srcRange;
849  *dstRange = c->dstRange;
850  *brightness = c->brightness;
851  *contrast = c->contrast;
852  *saturation = c->saturation;
853 
854  return 0;
855 }
856 
857 static int handle_jpeg(enum AVPixelFormat *format)
858 {
859  switch (*format) {
860  case AV_PIX_FMT_YUVJ420P:
861  *format = AV_PIX_FMT_YUV420P;
862  return 1;
863  case AV_PIX_FMT_YUVJ422P:
864  *format = AV_PIX_FMT_YUV422P;
865  return 1;
866  case AV_PIX_FMT_YUVJ444P:
867  *format = AV_PIX_FMT_YUV444P;
868  return 1;
869  case AV_PIX_FMT_YUVJ440P:
870  *format = AV_PIX_FMT_YUV440P;
871  return 1;
872  default:
873  return 0;
874  }
875 }
876 
877 static int handle_0alpha(enum AVPixelFormat *format)
878 {
879  switch (*format) {
880  case AV_PIX_FMT_0BGR : *format = AV_PIX_FMT_ABGR ; return 1;
881  case AV_PIX_FMT_BGR0 : *format = AV_PIX_FMT_BGRA ; return 4;
882  case AV_PIX_FMT_0RGB : *format = AV_PIX_FMT_ARGB ; return 1;
883  case AV_PIX_FMT_RGB0 : *format = AV_PIX_FMT_RGBA ; return 4;
884  default: return 0;
885  }
886 }
887 
889 {
890  SwsContext *c = av_mallocz(sizeof(SwsContext));
891 
892  if (c) {
895  }
896 
897  return c;
898 }
899 
901  SwsFilter *dstFilter)
902 {
903  int i, j;
904  int usesVFilter, usesHFilter;
905  int unscaled;
906  SwsFilter dummyFilter = { NULL, NULL, NULL, NULL };
907  int srcW = c->srcW;
908  int srcH = c->srcH;
909  int dstW = c->dstW;
910  int dstH = c->dstH;
911  int dst_stride = FFALIGN(dstW * sizeof(int16_t) + 66, 16);
912  int flags, cpu_flags;
913  enum AVPixelFormat srcFormat = c->srcFormat;
914  enum AVPixelFormat dstFormat = c->dstFormat;
915  const AVPixFmtDescriptor *desc_src = av_pix_fmt_desc_get(srcFormat);
916  const AVPixFmtDescriptor *desc_dst = av_pix_fmt_desc_get(dstFormat);
917 
918  cpu_flags = av_get_cpu_flags();
919  flags = c->flags;
920  emms_c();
921  if (!rgb15to16)
923 
924  unscaled = (srcW == dstW && srcH == dstH);
925 
926  handle_jpeg(&srcFormat);
927  handle_jpeg(&dstFormat);
928  handle_0alpha(&srcFormat);
929  handle_0alpha(&dstFormat);
930 
931  if(srcFormat!=c->srcFormat || dstFormat!=c->dstFormat){
932  av_log(c, AV_LOG_WARNING, "deprecated pixel format used, make sure you did set range correctly\n");
933  c->srcFormat= srcFormat;
934  c->dstFormat= dstFormat;
935  }
936 
937  if (!sws_isSupportedInput(srcFormat)) {
938  av_log(c, AV_LOG_ERROR, "%s is not supported as input pixel format\n",
939  av_get_pix_fmt_name(srcFormat));
940  return AVERROR(EINVAL);
941  }
942  if (!sws_isSupportedOutput(dstFormat)) {
943  av_log(c, AV_LOG_ERROR, "%s is not supported as output pixel format\n",
944  av_get_pix_fmt_name(dstFormat));
945  return AVERROR(EINVAL);
946  }
947 
948  i = flags & (SWS_POINT |
949  SWS_AREA |
950  SWS_BILINEAR |
952  SWS_BICUBIC |
953  SWS_X |
954  SWS_GAUSS |
955  SWS_LANCZOS |
956  SWS_SINC |
957  SWS_SPLINE |
958  SWS_BICUBLIN);
959  if (!i || (i & (i - 1))) {
960  av_log(c, AV_LOG_ERROR, "Exactly one scaler algorithm must be chosen, got %X\n", i);
961  return AVERROR(EINVAL);
962  }
963  /* sanity check */
964  if (srcW < 1 || srcH < 1 || dstW < 1 || dstH < 1) {
965  /* FIXME check if these are enough and try to lower them after
966  * fixing the relevant parts of the code */
967  av_log(c, AV_LOG_ERROR, "%dx%d -> %dx%d is invalid scaling dimension\n",
968  srcW, srcH, dstW, dstH);
969  return AVERROR(EINVAL);
970  }
971 
972  if (!dstFilter)
973  dstFilter = &dummyFilter;
974  if (!srcFilter)
975  srcFilter = &dummyFilter;
976 
977  c->lumXInc = (((int64_t)srcW << 16) + (dstW >> 1)) / dstW;
978  c->lumYInc = (((int64_t)srcH << 16) + (dstH >> 1)) / dstH;
979  c->dstFormatBpp = av_get_bits_per_pixel(desc_dst);
980  c->srcFormatBpp = av_get_bits_per_pixel(desc_src);
981  c->vRounder = 4 * 0x0001000100010001ULL;
982 
983  usesVFilter = (srcFilter->lumV && srcFilter->lumV->length > 1) ||
984  (srcFilter->chrV && srcFilter->chrV->length > 1) ||
985  (dstFilter->lumV && dstFilter->lumV->length > 1) ||
986  (dstFilter->chrV && dstFilter->chrV->length > 1);
987  usesHFilter = (srcFilter->lumH && srcFilter->lumH->length > 1) ||
988  (srcFilter->chrH && srcFilter->chrH->length > 1) ||
989  (dstFilter->lumH && dstFilter->lumH->length > 1) ||
990  (dstFilter->chrH && dstFilter->chrH->length > 1);
991 
994 
995  if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) {
996  if (dstW&1) {
997  av_log(c, AV_LOG_DEBUG, "Forcing full internal H chroma due to odd output size\n");
998  flags |= SWS_FULL_CHR_H_INT;
999  c->flags = flags;
1000  }
1001  }
1002 
1003  if(dstFormat == AV_PIX_FMT_BGR4_BYTE ||
1004  dstFormat == AV_PIX_FMT_RGB4_BYTE ||
1005  dstFormat == AV_PIX_FMT_BGR8 ||
1006  dstFormat == AV_PIX_FMT_RGB8) {
1007  if (flags & SWS_ERROR_DIFFUSION && !(flags & SWS_FULL_CHR_H_INT)) {
1008  av_log(c, AV_LOG_DEBUG,
1009  "Error diffusion dither is only supported in full chroma interpolation for destination format '%s'\n",
1010  av_get_pix_fmt_name(dstFormat));
1011  flags |= SWS_FULL_CHR_H_INT;
1012  c->flags = flags;
1013  }
1014  if (!(flags & SWS_ERROR_DIFFUSION) && (flags & SWS_FULL_CHR_H_INT)) {
1015  av_log(c, AV_LOG_DEBUG,
1016  "Ordered dither is not supported in full chroma interpolation for destination format '%s'\n",
1017  av_get_pix_fmt_name(dstFormat));
1018  flags |= SWS_ERROR_DIFFUSION;
1019  c->flags = flags;
1020  }
1021  }
1022  if (isPlanarRGB(dstFormat)) {
1023  if (!(flags & SWS_FULL_CHR_H_INT)) {
1024  av_log(c, AV_LOG_DEBUG,
1025  "%s output is not supported with half chroma resolution, switching to full\n",
1026  av_get_pix_fmt_name(dstFormat));
1027  flags |= SWS_FULL_CHR_H_INT;
1028  c->flags = flags;
1029  }
1030  }
1031 
1032  /* reuse chroma for 2 pixels RGB/BGR unless user wants full
1033  * chroma interpolation */
1034  if (flags & SWS_FULL_CHR_H_INT &&
1035  isAnyRGB(dstFormat) &&
1036  !isPlanarRGB(dstFormat) &&
1037  dstFormat != AV_PIX_FMT_RGBA &&
1038  dstFormat != AV_PIX_FMT_ARGB &&
1039  dstFormat != AV_PIX_FMT_BGRA &&
1040  dstFormat != AV_PIX_FMT_ABGR &&
1041  dstFormat != AV_PIX_FMT_RGB24 &&
1042  dstFormat != AV_PIX_FMT_BGR24 &&
1043  dstFormat != AV_PIX_FMT_BGR4_BYTE &&
1044  dstFormat != AV_PIX_FMT_RGB4_BYTE &&
1045  dstFormat != AV_PIX_FMT_BGR8 &&
1046  dstFormat != AV_PIX_FMT_RGB8
1047  ) {
1049  "full chroma interpolation for destination format '%s' not yet implemented\n",
1050  av_get_pix_fmt_name(dstFormat));
1051  flags &= ~SWS_FULL_CHR_H_INT;
1052  c->flags = flags;
1053  }
1054  if (isAnyRGB(dstFormat) && !(flags & SWS_FULL_CHR_H_INT))
1055  c->chrDstHSubSample = 1;
1056 
1057  // drop some chroma lines if the user wants it
1058  c->vChrDrop = (flags & SWS_SRC_V_CHR_DROP_MASK) >>
1060  c->chrSrcVSubSample += c->vChrDrop;
1061 
1062  /* drop every other pixel for chroma calculation unless user
1063  * wants full chroma */
1064  if (isAnyRGB(srcFormat) && !(flags & SWS_FULL_CHR_H_INP) &&
1065  srcFormat != AV_PIX_FMT_RGB8 && srcFormat != AV_PIX_FMT_BGR8 &&
1066  srcFormat != AV_PIX_FMT_RGB4 && srcFormat != AV_PIX_FMT_BGR4 &&
1067  srcFormat != AV_PIX_FMT_RGB4_BYTE && srcFormat != AV_PIX_FMT_BGR4_BYTE &&
1068  srcFormat != AV_PIX_FMT_GBRP9BE && srcFormat != AV_PIX_FMT_GBRP9LE &&
1069  srcFormat != AV_PIX_FMT_GBRP10BE && srcFormat != AV_PIX_FMT_GBRP10LE &&
1070  srcFormat != AV_PIX_FMT_GBRP12BE && srcFormat != AV_PIX_FMT_GBRP12LE &&
1071  srcFormat != AV_PIX_FMT_GBRP14BE && srcFormat != AV_PIX_FMT_GBRP14LE &&
1072  srcFormat != AV_PIX_FMT_GBRP16BE && srcFormat != AV_PIX_FMT_GBRP16LE &&
1073  ((dstW >> c->chrDstHSubSample) <= (srcW >> 1) ||
1074  (flags & SWS_FAST_BILINEAR)))
1075  c->chrSrcHSubSample = 1;
1076 
1077  // Note the -((-x)>>y) is so that we always round toward +inf.
1078  c->chrSrcW = -((-srcW) >> c->chrSrcHSubSample);
1079  c->chrSrcH = -((-srcH) >> c->chrSrcVSubSample);
1080  c->chrDstW = -((-dstW) >> c->chrDstHSubSample);
1081  c->chrDstH = -((-dstH) >> c->chrDstVSubSample);
1082 
1083  FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW*2+78, 16) * 2, fail);
1084 
1085  /* unscaled special cases */
1086  if (unscaled && !usesHFilter && !usesVFilter &&
1087  (c->srcRange == c->dstRange || isAnyRGB(dstFormat))) {
1089 
1090  if (c->swScale) {
1091  if (flags & SWS_PRINT_INFO)
1092  av_log(c, AV_LOG_INFO,
1093  "using unscaled %s -> %s special converter\n",
1094  av_get_pix_fmt_name(srcFormat), av_get_pix_fmt_name(dstFormat));
1095  return 0;
1096  }
1097  }
1098 
1099  c->srcBpc = 1 + desc_src->comp[0].depth_minus1;
1100  if (c->srcBpc < 8)
1101  c->srcBpc = 8;
1102  c->dstBpc = 1 + desc_dst->comp[0].depth_minus1;
1103  if (c->dstBpc < 8)
1104  c->dstBpc = 8;
1105  if (isAnyRGB(srcFormat) || srcFormat == AV_PIX_FMT_PAL8)
1106  c->srcBpc = 16;
1107  if (c->dstBpc == 16)
1108  dst_stride <<= 1;
1109 
1110  if (INLINE_MMXEXT(cpu_flags) && c->srcBpc == 8 && c->dstBpc <= 14) {
1111  c->canMMXEXTBeUsed = dstW >= srcW && (dstW & 31) == 0 &&
1112  c->chrDstW >= c->chrSrcW &&
1113  (srcW & 15) == 0;
1114  if (!c->canMMXEXTBeUsed && dstW >= srcW && c->chrDstW >= c->chrSrcW && (srcW & 15) == 0
1115 
1116  && (flags & SWS_FAST_BILINEAR)) {
1117  if (flags & SWS_PRINT_INFO)
1118  av_log(c, AV_LOG_INFO,
1119  "output width is not a multiple of 32 -> no MMXEXT scaler\n");
1120  }
1121  if (usesHFilter || isNBPS(c->srcFormat) || is16BPS(c->srcFormat) || isAnyRGB(c->srcFormat))
1122  c->canMMXEXTBeUsed = 0;
1123  } else
1124  c->canMMXEXTBeUsed = 0;
1125 
1126  c->chrXInc = (((int64_t)c->chrSrcW << 16) + (c->chrDstW >> 1)) / c->chrDstW;
1127  c->chrYInc = (((int64_t)c->chrSrcH << 16) + (c->chrDstH >> 1)) / c->chrDstH;
1128 
1129  /* Match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src
1130  * to pixel n-2 of dst, but only for the FAST_BILINEAR mode otherwise do
1131  * correct scaling.
1132  * n-2 is the last chrominance sample available.
1133  * This is not perfect, but no one should notice the difference, the more
1134  * correct variant would be like the vertical one, but that would require
1135  * some special code for the first and last pixel */
1136  if (flags & SWS_FAST_BILINEAR) {
1137  if (c->canMMXEXTBeUsed) {
1138  c->lumXInc += 20;
1139  c->chrXInc += 20;
1140  }
1141  // we don't use the x86 asm scaler if MMX is available
1142  else if (INLINE_MMX(cpu_flags) && c->dstBpc <= 14) {
1143  c->lumXInc = ((int64_t)(srcW - 2) << 16) / (dstW - 2) - 20;
1144  c->chrXInc = ((int64_t)(c->chrSrcW - 2) << 16) / (c->chrDstW - 2) - 20;
1145  }
1146  }
1147 
1148 #define USE_MMAP (HAVE_MMAP && HAVE_MPROTECT && defined MAP_ANONYMOUS)
1149 
1150  /* precalculate horizontal scaler filter coefficients */
1151  {
1152 #if HAVE_MMXEXT_INLINE
1153 // can't downscale !!!
1154  if (c->canMMXEXTBeUsed && (flags & SWS_FAST_BILINEAR)) {
1155  c->lumMmxextFilterCodeSize = init_hscaler_mmxext(dstW, c->lumXInc, NULL,
1156  NULL, NULL, 8);
1157  c->chrMmxextFilterCodeSize = init_hscaler_mmxext(c->chrDstW, c->chrXInc,
1158  NULL, NULL, NULL, 4);
1159 
1160 #if USE_MMAP
1162  PROT_READ | PROT_WRITE,
1163  MAP_PRIVATE | MAP_ANONYMOUS,
1164  -1, 0);
1166  PROT_READ | PROT_WRITE,
1167  MAP_PRIVATE | MAP_ANONYMOUS,
1168  -1, 0);
1169 #elif HAVE_VIRTUALALLOC
1170  c->lumMmxextFilterCode = VirtualAlloc(NULL,
1172  MEM_COMMIT,
1173  PAGE_EXECUTE_READWRITE);
1174  c->chrMmxextFilterCode = VirtualAlloc(NULL,
1176  MEM_COMMIT,
1177  PAGE_EXECUTE_READWRITE);
1178 #else
1181 #endif
1182 
1183 #ifdef MAP_ANONYMOUS
1184  if (c->lumMmxextFilterCode == MAP_FAILED || c->chrMmxextFilterCode == MAP_FAILED)
1185 #else
1187 #endif
1188  {
1189  av_log(c, AV_LOG_ERROR, "Failed to allocate MMX2FilterCode\n");
1190  return AVERROR(ENOMEM);
1191  }
1192 
1193  FF_ALLOCZ_OR_GOTO(c, c->hLumFilter, (dstW / 8 + 8) * sizeof(int16_t), fail);
1194  FF_ALLOCZ_OR_GOTO(c, c->hChrFilter, (c->chrDstW / 4 + 8) * sizeof(int16_t), fail);
1195  FF_ALLOCZ_OR_GOTO(c, c->hLumFilterPos, (dstW / 2 / 8 + 8) * sizeof(int32_t), fail);
1196  FF_ALLOCZ_OR_GOTO(c, c->hChrFilterPos, (c->chrDstW / 2 / 4 + 8) * sizeof(int32_t), fail);
1197 
1198  init_hscaler_mmxext( dstW, c->lumXInc, c->lumMmxextFilterCode,
1199  c->hLumFilter, (uint32_t*)c->hLumFilterPos, 8);
1200  init_hscaler_mmxext(c->chrDstW, c->chrXInc, c->chrMmxextFilterCode,
1201  c->hChrFilter, (uint32_t*)c->hChrFilterPos, 4);
1202 
1203 #if USE_MMAP
1204  mprotect(c->lumMmxextFilterCode, c->lumMmxextFilterCodeSize, PROT_EXEC | PROT_READ);
1205  mprotect(c->chrMmxextFilterCode, c->chrMmxextFilterCodeSize, PROT_EXEC | PROT_READ);
1206 #endif
1207  } else
1208 #endif /* HAVE_MMXEXT_INLINE */
1209  {
1210  const int filterAlign =
1211  (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 4 :
1212  (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1213  1;
1214 
1215  if (initFilter(&c->hLumFilter, &c->hLumFilterPos,
1216  &c->hLumFilterSize, c->lumXInc,
1217  srcW, dstW, filterAlign, 1 << 14,
1218  (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1219  cpu_flags, srcFilter->lumH, dstFilter->lumH,
1220  c->param) < 0)
1221  goto fail;
1222  if (initFilter(&c->hChrFilter, &c->hChrFilterPos,
1223  &c->hChrFilterSize, c->chrXInc,
1224  c->chrSrcW, c->chrDstW, filterAlign, 1 << 14,
1225  (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1226  cpu_flags, srcFilter->chrH, dstFilter->chrH,
1227  c->param) < 0)
1228  goto fail;
1229  }
1230  } // initialize horizontal stuff
1231 
1232  /* precalculate vertical scaler filter coefficients */
1233  {
1234  const int filterAlign =
1235  (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) ? 2 :
1236  (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC) ? 8 :
1237  1;
1238 
1240  c->lumYInc, srcH, dstH, filterAlign, (1 << 12),
1241  (flags & SWS_BICUBLIN) ? (flags | SWS_BICUBIC) : flags,
1242  cpu_flags, srcFilter->lumV, dstFilter->lumV,
1243  c->param) < 0)
1244  goto fail;
1246  c->chrYInc, c->chrSrcH, c->chrDstH,
1247  filterAlign, (1 << 12),
1248  (flags & SWS_BICUBLIN) ? (flags | SWS_BILINEAR) : flags,
1249  cpu_flags, srcFilter->chrV, dstFilter->chrV,
1250  c->param) < 0)
1251  goto fail;
1252 
1253 #if HAVE_ALTIVEC
1254  FF_ALLOC_OR_GOTO(c, c->vYCoeffsBank, sizeof(vector signed short) * c->vLumFilterSize * c->dstH, fail);
1255  FF_ALLOC_OR_GOTO(c, c->vCCoeffsBank, sizeof(vector signed short) * c->vChrFilterSize * c->chrDstH, fail);
1256 
1257  for (i = 0; i < c->vLumFilterSize * c->dstH; i++) {
1258  int j;
1259  short *p = (short *)&c->vYCoeffsBank[i];
1260  for (j = 0; j < 8; j++)
1261  p[j] = c->vLumFilter[i];
1262  }
1263 
1264  for (i = 0; i < c->vChrFilterSize * c->chrDstH; i++) {
1265  int j;
1266  short *p = (short *)&c->vCCoeffsBank[i];
1267  for (j = 0; j < 8; j++)
1268  p[j] = c->vChrFilter[i];
1269  }
1270 #endif
1271  }
1272 
1273  // calculate buffer sizes so that they won't run out while handling these damn slices
1274  c->vLumBufSize = c->vLumFilterSize;
1275  c->vChrBufSize = c->vChrFilterSize;
1276  for (i = 0; i < dstH; i++) {
1277  int chrI = (int64_t)i * c->chrDstH / dstH;
1278  int nextSlice = FFMAX(c->vLumFilterPos[i] + c->vLumFilterSize - 1,
1279  ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)
1280  << c->chrSrcVSubSample));
1281 
1282  nextSlice >>= c->chrSrcVSubSample;
1283  nextSlice <<= c->chrSrcVSubSample;
1284  if (c->vLumFilterPos[i] + c->vLumBufSize < nextSlice)
1285  c->vLumBufSize = nextSlice - c->vLumFilterPos[i];
1286  if (c->vChrFilterPos[chrI] + c->vChrBufSize <
1287  (nextSlice >> c->chrSrcVSubSample))
1288  c->vChrBufSize = (nextSlice >> c->chrSrcVSubSample) -
1289  c->vChrFilterPos[chrI];
1290  }
1291 
1292  for (i = 0; i < 4; i++)
1293  FF_ALLOCZ_OR_GOTO(c, c->dither_error[i], (c->dstW+2) * sizeof(int), fail);
1294 
1295  /* Allocate pixbufs (we use dynamic allocation because otherwise we would
1296  * need to allocate several megabytes to handle all possible cases) */
1297  FF_ALLOC_OR_GOTO(c, c->lumPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1298  FF_ALLOC_OR_GOTO(c, c->chrUPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1299  FF_ALLOC_OR_GOTO(c, c->chrVPixBuf, c->vChrBufSize * 3 * sizeof(int16_t *), fail);
1301  FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf, c->vLumBufSize * 3 * sizeof(int16_t *), fail);
1302  /* Note we need at least one pixel more at the end because of the MMX code
1303  * (just in case someone wants to replace the 4000/8000). */
1304  /* align at 16 bytes for AltiVec */
1305  for (i = 0; i < c->vLumBufSize; i++) {
1306  FF_ALLOCZ_OR_GOTO(c, c->lumPixBuf[i + c->vLumBufSize],
1307  dst_stride + 16, fail);
1308  c->lumPixBuf[i] = c->lumPixBuf[i + c->vLumBufSize];
1309  }
1310  // 64 / c->scalingBpp is the same as 16 / sizeof(scaling_intermediate)
1311  c->uv_off = (dst_stride>>1) + 64 / (c->dstBpc &~ 7);
1312  c->uv_offx2 = dst_stride + 16;
1313  for (i = 0; i < c->vChrBufSize; i++) {
1314  FF_ALLOC_OR_GOTO(c, c->chrUPixBuf[i + c->vChrBufSize],
1315  dst_stride * 2 + 32, fail);
1316  c->chrUPixBuf[i] = c->chrUPixBuf[i + c->vChrBufSize];
1317  c->chrVPixBuf[i] = c->chrVPixBuf[i + c->vChrBufSize]
1318  = c->chrUPixBuf[i] + (dst_stride >> 1) + 8;
1319  }
1320  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1321  for (i = 0; i < c->vLumBufSize; i++) {
1322  FF_ALLOCZ_OR_GOTO(c, c->alpPixBuf[i + c->vLumBufSize],
1323  dst_stride + 16, fail);
1324  c->alpPixBuf[i] = c->alpPixBuf[i + c->vLumBufSize];
1325  }
1326 
1327  // try to avoid drawing green stuff between the right end and the stride end
1328  for (i = 0; i < c->vChrBufSize; i++)
1329  if(desc_dst->comp[0].depth_minus1 == 15){
1330  av_assert0(c->dstBpc > 14);
1331  for(j=0; j<dst_stride/2+1; j++)
1332  ((int32_t*)(c->chrUPixBuf[i]))[j] = 1<<18;
1333  } else
1334  for(j=0; j<dst_stride+1; j++)
1335  ((int16_t*)(c->chrUPixBuf[i]))[j] = 1<<14;
1336 
1337  av_assert0(c->chrDstH <= dstH);
1338 
1339  if (flags & SWS_PRINT_INFO) {
1340  if (flags & SWS_FAST_BILINEAR)
1341  av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
1342  else if (flags & SWS_BILINEAR)
1343  av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
1344  else if (flags & SWS_BICUBIC)
1345  av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
1346  else if (flags & SWS_X)
1347  av_log(c, AV_LOG_INFO, "Experimental scaler, ");
1348  else if (flags & SWS_POINT)
1349  av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
1350  else if (flags & SWS_AREA)
1351  av_log(c, AV_LOG_INFO, "Area Averaging scaler, ");
1352  else if (flags & SWS_BICUBLIN)
1353  av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
1354  else if (flags & SWS_GAUSS)
1355  av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
1356  else if (flags & SWS_SINC)
1357  av_log(c, AV_LOG_INFO, "Sinc scaler, ");
1358  else if (flags & SWS_LANCZOS)
1359  av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
1360  else if (flags & SWS_SPLINE)
1361  av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
1362  else
1363  av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
1364 
1365  av_log(c, AV_LOG_INFO, "from %s to %s%s ",
1366  av_get_pix_fmt_name(srcFormat),
1367 #ifdef DITHER1XBPP
1368  dstFormat == AV_PIX_FMT_BGR555 || dstFormat == AV_PIX_FMT_BGR565 ||
1369  dstFormat == AV_PIX_FMT_RGB444BE || dstFormat == AV_PIX_FMT_RGB444LE ||
1370  dstFormat == AV_PIX_FMT_BGR444BE || dstFormat == AV_PIX_FMT_BGR444LE ?
1371  "dithered " : "",
1372 #else
1373  "",
1374 #endif
1375  av_get_pix_fmt_name(dstFormat));
1376 
1377  if (INLINE_MMXEXT(cpu_flags))
1378  av_log(c, AV_LOG_INFO, "using MMXEXT\n");
1379  else if (INLINE_AMD3DNOW(cpu_flags))
1380  av_log(c, AV_LOG_INFO, "using 3DNOW\n");
1381  else if (INLINE_MMX(cpu_flags))
1382  av_log(c, AV_LOG_INFO, "using MMX\n");
1383  else if (HAVE_ALTIVEC && cpu_flags & AV_CPU_FLAG_ALTIVEC)
1384  av_log(c, AV_LOG_INFO, "using AltiVec\n");
1385  else
1386  av_log(c, AV_LOG_INFO, "using C\n");
1387 
1388  av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
1389  av_log(c, AV_LOG_DEBUG,
1390  "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1391  c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
1392  av_log(c, AV_LOG_DEBUG,
1393  "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
1394  c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH,
1395  c->chrXInc, c->chrYInc);
1396  }
1397 
1398  c->swScale = ff_getSwsFunc(c);
1399  return 0;
1400 fail: // FIXME replace things by appropriate error codes
1401  return -1;
1402 }
1403 
1404 #if FF_API_SWS_GETCONTEXT
1405 SwsContext *sws_getContext(int srcW, int srcH, enum AVPixelFormat srcFormat,
1406  int dstW, int dstH, enum AVPixelFormat dstFormat,
1407  int flags, SwsFilter *srcFilter,
1408  SwsFilter *dstFilter, const double *param)
1409 {
1410  SwsContext *c;
1411 
1412  if (!(c = sws_alloc_context()))
1413  return NULL;
1414 
1415  c->flags = flags;
1416  c->srcW = srcW;
1417  c->srcH = srcH;
1418  c->dstW = dstW;
1419  c->dstH = dstH;
1420  c->srcRange = handle_jpeg(&srcFormat);
1421  c->dstRange = handle_jpeg(&dstFormat);
1422  c->src0Alpha = handle_0alpha(&srcFormat);
1423  c->dst0Alpha = handle_0alpha(&dstFormat);
1424  c->srcFormat = srcFormat;
1425  c->dstFormat = dstFormat;
1426 
1427  if (param) {
1428  c->param[0] = param[0];
1429  c->param[1] = param[1];
1430  }
1432  ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/,
1433  c->dstRange, 0, 1 << 16, 1 << 16);
1434 
1435  if (sws_init_context(c, srcFilter, dstFilter) < 0) {
1436  sws_freeContext(c);
1437  return NULL;
1438  }
1439 
1440  return c;
1441 }
1442 #endif
1443 
1444 SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
1445  float lumaSharpen, float chromaSharpen,
1446  float chromaHShift, float chromaVShift,
1447  int verbose)
1448 {
1449  SwsFilter *filter = av_malloc(sizeof(SwsFilter));
1450  if (!filter)
1451  return NULL;
1452 
1453  if (lumaGBlur != 0.0) {
1454  filter->lumH = sws_getGaussianVec(lumaGBlur, 3.0);
1455  filter->lumV = sws_getGaussianVec(lumaGBlur, 3.0);
1456  } else {
1457  filter->lumH = sws_getIdentityVec();
1458  filter->lumV = sws_getIdentityVec();
1459  }
1460 
1461  if (chromaGBlur != 0.0) {
1462  filter->chrH = sws_getGaussianVec(chromaGBlur, 3.0);
1463  filter->chrV = sws_getGaussianVec(chromaGBlur, 3.0);
1464  } else {
1465  filter->chrH = sws_getIdentityVec();
1466  filter->chrV = sws_getIdentityVec();
1467  }
1468 
1469  if (chromaSharpen != 0.0) {
1470  SwsVector *id = sws_getIdentityVec();
1471  sws_scaleVec(filter->chrH, -chromaSharpen);
1472  sws_scaleVec(filter->chrV, -chromaSharpen);
1473  sws_addVec(filter->chrH, id);
1474  sws_addVec(filter->chrV, id);
1475  sws_freeVec(id);
1476  }
1477 
1478  if (lumaSharpen != 0.0) {
1479  SwsVector *id = sws_getIdentityVec();
1480  sws_scaleVec(filter->lumH, -lumaSharpen);
1481  sws_scaleVec(filter->lumV, -lumaSharpen);
1482  sws_addVec(filter->lumH, id);
1483  sws_addVec(filter->lumV, id);
1484  sws_freeVec(id);
1485  }
1486 
1487  if (chromaHShift != 0.0)
1488  sws_shiftVec(filter->chrH, (int)(chromaHShift + 0.5));
1489 
1490  if (chromaVShift != 0.0)
1491  sws_shiftVec(filter->chrV, (int)(chromaVShift + 0.5));
1492 
1493  sws_normalizeVec(filter->chrH, 1.0);
1494  sws_normalizeVec(filter->chrV, 1.0);
1495  sws_normalizeVec(filter->lumH, 1.0);
1496  sws_normalizeVec(filter->lumV, 1.0);
1497 
1498  if (verbose)
1499  sws_printVec2(filter->chrH, NULL, AV_LOG_DEBUG);
1500  if (verbose)
1501  sws_printVec2(filter->lumH, NULL, AV_LOG_DEBUG);
1502 
1503  return filter;
1504 }
1505 
1507 {
1508  SwsVector *vec;
1509 
1510  if(length <= 0 || length > INT_MAX/ sizeof(double))
1511  return NULL;
1512 
1513  vec = av_malloc(sizeof(SwsVector));
1514  if (!vec)
1515  return NULL;
1516  vec->length = length;
1517  vec->coeff = av_malloc(sizeof(double) * length);
1518  if (!vec->coeff)
1519  av_freep(&vec);
1520  return vec;
1521 }
1522 
1523 SwsVector *sws_getGaussianVec(double variance, double quality)
1524 {
1525  const int length = (int)(variance * quality + 0.5) | 1;
1526  int i;
1527  double middle = (length - 1) * 0.5;
1528  SwsVector *vec;
1529 
1530  if(variance < 0 || quality < 0)
1531  return NULL;
1532 
1533  vec = sws_allocVec(length);
1534 
1535  if (!vec)
1536  return NULL;
1537 
1538  for (i = 0; i < length; i++) {
1539  double dist = i - middle;
1540  vec->coeff[i] = exp(-dist * dist / (2 * variance * variance)) /
1541  sqrt(2 * variance * M_PI);
1542  }
1543 
1544  sws_normalizeVec(vec, 1.0);
1545 
1546  return vec;
1547 }
1548 
1549 SwsVector *sws_getConstVec(double c, int length)
1550 {
1551  int i;
1552  SwsVector *vec = sws_allocVec(length);
1553 
1554  if (!vec)
1555  return NULL;
1556 
1557  for (i = 0; i < length; i++)
1558  vec->coeff[i] = c;
1559 
1560  return vec;
1561 }
1562 
1564 {
1565  return sws_getConstVec(1.0, 1);
1566 }
1567 
1568 static double sws_dcVec(SwsVector *a)
1569 {
1570  int i;
1571  double sum = 0;
1572 
1573  for (i = 0; i < a->length; i++)
1574  sum += a->coeff[i];
1575 
1576  return sum;
1577 }
1578 
1579 void sws_scaleVec(SwsVector *a, double scalar)
1580 {
1581  int i;
1582 
1583  for (i = 0; i < a->length; i++)
1584  a->coeff[i] *= scalar;
1585 }
1586 
1588 {
1589  sws_scaleVec(a, height / sws_dcVec(a));
1590 }
1591 
1593 {
1594  int length = a->length + b->length - 1;
1595  int i, j;
1596  SwsVector *vec = sws_getConstVec(0.0, length);
1597 
1598  if (!vec)
1599  return NULL;
1600 
1601  for (i = 0; i < a->length; i++) {
1602  for (j = 0; j < b->length; j++) {
1603  vec->coeff[i + j] += a->coeff[i] * b->coeff[j];
1604  }
1605  }
1606 
1607  return vec;
1608 }
1609 
1611 {
1612  int length = FFMAX(a->length, b->length);
1613  int i;
1614  SwsVector *vec = sws_getConstVec(0.0, length);
1615 
1616  if (!vec)
1617  return NULL;
1618 
1619  for (i = 0; i < a->length; i++)
1620  vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1621  for (i = 0; i < b->length; i++)
1622  vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] += b->coeff[i];
1623 
1624  return vec;
1625 }
1626 
1628 {
1629  int length = FFMAX(a->length, b->length);
1630  int i;
1631  SwsVector *vec = sws_getConstVec(0.0, length);
1632 
1633  if (!vec)
1634  return NULL;
1635 
1636  for (i = 0; i < a->length; i++)
1637  vec->coeff[i + (length - 1) / 2 - (a->length - 1) / 2] += a->coeff[i];
1638  for (i = 0; i < b->length; i++)
1639  vec->coeff[i + (length - 1) / 2 - (b->length - 1) / 2] -= b->coeff[i];
1640 
1641  return vec;
1642 }
1643 
1644 /* shift left / or right if "shift" is negative */
1645 static SwsVector *sws_getShiftedVec(SwsVector *a, int shift)
1646 {
1647  int length = a->length + FFABS(shift) * 2;
1648  int i;
1649  SwsVector *vec = sws_getConstVec(0.0, length);
1650 
1651  if (!vec)
1652  return NULL;
1653 
1654  for (i = 0; i < a->length; i++) {
1655  vec->coeff[i + (length - 1) / 2 -
1656  (a->length - 1) / 2 - shift] = a->coeff[i];
1657  }
1658 
1659  return vec;
1660 }
1661 
1662 void sws_shiftVec(SwsVector *a, int shift)
1663 {
1664  SwsVector *shifted = sws_getShiftedVec(a, shift);
1665  av_free(a->coeff);
1666  a->coeff = shifted->coeff;
1667  a->length = shifted->length;
1668  av_free(shifted);
1669 }
1670 
1672 {
1673  SwsVector *sum = sws_sumVec(a, b);
1674  av_free(a->coeff);
1675  a->coeff = sum->coeff;
1676  a->length = sum->length;
1677  av_free(sum);
1678 }
1679 
1681 {
1682  SwsVector *diff = sws_diffVec(a, b);
1683  av_free(a->coeff);
1684  a->coeff = diff->coeff;
1685  a->length = diff->length;
1686  av_free(diff);
1687 }
1688 
1690 {
1691  SwsVector *conv = sws_getConvVec(a, b);
1692  av_free(a->coeff);
1693  a->coeff = conv->coeff;
1694  a->length = conv->length;
1695  av_free(conv);
1696 }
1697 
1699 {
1700  int i;
1701  SwsVector *vec = sws_allocVec(a->length);
1702 
1703  if (!vec)
1704  return NULL;
1705 
1706  for (i = 0; i < a->length; i++)
1707  vec->coeff[i] = a->coeff[i];
1708 
1709  return vec;
1710 }
1711 
1712 void sws_printVec2(SwsVector *a, AVClass *log_ctx, int log_level)
1713 {
1714  int i;
1715  double max = 0;
1716  double min = 0;
1717  double range;
1718 
1719  for (i = 0; i < a->length; i++)
1720  if (a->coeff[i] > max)
1721  max = a->coeff[i];
1722 
1723  for (i = 0; i < a->length; i++)
1724  if (a->coeff[i] < min)
1725  min = a->coeff[i];
1726 
1727  range = max - min;
1728 
1729  for (i = 0; i < a->length; i++) {
1730  int x = (int)((a->coeff[i] - min) * 60.0 / range + 0.5);
1731  av_log(log_ctx, log_level, "%1.3f ", a->coeff[i]);
1732  for (; x > 0; x--)
1733  av_log(log_ctx, log_level, " ");
1734  av_log(log_ctx, log_level, "|\n");
1735  }
1736 }
1737 
1739 {
1740  if (!a)
1741  return;
1742  av_freep(&a->coeff);
1743  a->length = 0;
1744  av_free(a);
1745 }
1746 
1748 {
1749  if (!filter)
1750  return;
1751 
1752  if (filter->lumH)
1753  sws_freeVec(filter->lumH);
1754  if (filter->lumV)
1755  sws_freeVec(filter->lumV);
1756  if (filter->chrH)
1757  sws_freeVec(filter->chrH);
1758  if (filter->chrV)
1759  sws_freeVec(filter->chrV);
1760  av_free(filter);
1761 }
1762 
1764 {
1765  int i;
1766  if (!c)
1767  return;
1768 
1769  if (c->lumPixBuf) {
1770  for (i = 0; i < c->vLumBufSize; i++)
1771  av_freep(&c->lumPixBuf[i]);
1772  av_freep(&c->lumPixBuf);
1773  }
1774 
1775  if (c->chrUPixBuf) {
1776  for (i = 0; i < c->vChrBufSize; i++)
1777  av_freep(&c->chrUPixBuf[i]);
1778  av_freep(&c->chrUPixBuf);
1779  av_freep(&c->chrVPixBuf);
1780  }
1781 
1782  if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
1783  for (i = 0; i < c->vLumBufSize; i++)
1784  av_freep(&c->alpPixBuf[i]);
1785  av_freep(&c->alpPixBuf);
1786  }
1787 
1788  for (i = 0; i < 4; i++)
1789  av_freep(&c->dither_error[i]);
1790 
1791  av_freep(&c->vLumFilter);
1792  av_freep(&c->vChrFilter);
1793  av_freep(&c->hLumFilter);
1794  av_freep(&c->hChrFilter);
1795 #if HAVE_ALTIVEC
1796  av_freep(&c->vYCoeffsBank);
1797  av_freep(&c->vCCoeffsBank);
1798 #endif
1799 
1800  av_freep(&c->vLumFilterPos);
1801  av_freep(&c->vChrFilterPos);
1802  av_freep(&c->hLumFilterPos);
1803  av_freep(&c->hChrFilterPos);
1804 
1805 #if HAVE_MMX_INLINE
1806 #if USE_MMAP
1807  if (c->lumMmxextFilterCode)
1809  if (c->chrMmxextFilterCode)
1811 #elif HAVE_VIRTUALALLOC
1812  if (c->lumMmxextFilterCode)
1813  VirtualFree(c->lumMmxextFilterCode, 0, MEM_RELEASE);
1814  if (c->chrMmxextFilterCode)
1815  VirtualFree(c->chrMmxextFilterCode, 0, MEM_RELEASE);
1816 #else
1819 #endif
1822 #endif /* HAVE_MMX_INLINE */
1823 
1824  av_freep(&c->yuvTable);
1826 
1827  av_free(c);
1828 }
1829 
1830 struct SwsContext *sws_getCachedContext(struct SwsContext *context, int srcW,
1831  int srcH, enum AVPixelFormat srcFormat,
1832  int dstW, int dstH,
1833  enum AVPixelFormat dstFormat, int flags,
1834  SwsFilter *srcFilter,
1835  SwsFilter *dstFilter,
1836  const double *param)
1837 {
1838  static const double default_param[2] = { SWS_PARAM_DEFAULT,
1840 
1841  if (!param)
1842  param = default_param;
1843 
1844  if (context &&
1845  (context->srcW != srcW ||
1846  context->srcH != srcH ||
1847  context->srcFormat != srcFormat ||
1848  context->dstW != dstW ||
1849  context->dstH != dstH ||
1850  context->dstFormat != dstFormat ||
1851  context->flags != flags ||
1852  context->param[0] != param[0] ||
1853  context->param[1] != param[1])) {
1854  sws_freeContext(context);
1855  context = NULL;
1856  }
1857 
1858  if (!context) {
1859  if (!(context = sws_alloc_context()))
1860  return NULL;
1861  context->srcW = srcW;
1862  context->srcH = srcH;
1863  context->srcRange = handle_jpeg(&srcFormat);
1864  context->src0Alpha = handle_0alpha(&srcFormat);
1865  context->srcFormat = srcFormat;
1866  context->dstW = dstW;
1867  context->dstH = dstH;
1868  context->dstRange = handle_jpeg(&dstFormat);
1869  context->dst0Alpha = handle_0alpha(&dstFormat);
1870  context->dstFormat = dstFormat;
1871  context->flags = flags;
1872  context->param[0] = param[0];
1873  context->param[1] = param[1];
1875  context->srcRange,
1876  ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/,
1877  context->dstRange, 0, 1 << 16, 1 << 16);
1878  if (sws_init_context(context, srcFilter, dstFilter) < 0) {
1879  sws_freeContext(context);
1880  return NULL;
1881  }
1882  }
1883  return context;
1884 }