FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
output.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2001-2012 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 <assert.h>
22 #include <math.h>
23 #include <stdint.h>
24 #include <stdio.h>
25 #include <string.h>
26 
27 #include "libavutil/attributes.h"
28 #include "libavutil/avutil.h"
29 #include "libavutil/avassert.h"
30 #include "libavutil/bswap.h"
31 #include "libavutil/cpu.h"
32 #include "libavutil/intreadwrite.h"
33 #include "libavutil/mathematics.h"
34 #include "libavutil/pixdesc.h"
35 #include "config.h"
36 #include "rgb2rgb.h"
37 #include "swscale.h"
38 #include "swscale_internal.h"
39 
40 DECLARE_ALIGNED(8, const uint8_t, dither_2x2_4)[][8]={
41 { 1, 3, 1, 3, 1, 3, 1, 3, },
42 { 2, 0, 2, 0, 2, 0, 2, 0, },
43 { 1, 3, 1, 3, 1, 3, 1, 3, },
44 };
45 
46 DECLARE_ALIGNED(8, const uint8_t, dither_2x2_8)[][8]={
47 { 6, 2, 6, 2, 6, 2, 6, 2, },
48 { 0, 4, 0, 4, 0, 4, 0, 4, },
49 { 6, 2, 6, 2, 6, 2, 6, 2, },
50 };
51 
52 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[][8]={
53 { 8, 4, 11, 7, 8, 4, 11, 7, },
54 { 2, 14, 1, 13, 2, 14, 1, 13, },
55 { 10, 6, 9, 5, 10, 6, 9, 5, },
56 { 0, 12, 3, 15, 0, 12, 3, 15, },
57 { 8, 4, 11, 7, 8, 4, 11, 7, },
58 };
59 
60 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[][8]={
61 { 17, 9, 23, 15, 16, 8, 22, 14, },
62 { 5, 29, 3, 27, 4, 28, 2, 26, },
63 { 21, 13, 19, 11, 20, 12, 18, 10, },
64 { 0, 24, 6, 30, 1, 25, 7, 31, },
65 { 16, 8, 22, 14, 17, 9, 23, 15, },
66 { 4, 28, 2, 26, 5, 29, 3, 27, },
67 { 20, 12, 18, 10, 21, 13, 19, 11, },
68 { 1, 25, 7, 31, 0, 24, 6, 30, },
69 { 17, 9, 23, 15, 16, 8, 22, 14, },
70 };
71 
72 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[][8]={
73 { 0, 55, 14, 68, 3, 58, 17, 72, },
74 { 37, 18, 50, 32, 40, 22, 54, 35, },
75 { 9, 64, 5, 59, 13, 67, 8, 63, },
76 { 46, 27, 41, 23, 49, 31, 44, 26, },
77 { 2, 57, 16, 71, 1, 56, 15, 70, },
78 { 39, 21, 52, 34, 38, 19, 51, 33, },
79 { 11, 66, 7, 62, 10, 65, 6, 60, },
80 { 48, 30, 43, 25, 47, 29, 42, 24, },
81 { 0, 55, 14, 68, 3, 58, 17, 72, },
82 };
83 
84 #if 1
85 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[][8]={
86 {117, 62, 158, 103, 113, 58, 155, 100, },
87 { 34, 199, 21, 186, 31, 196, 17, 182, },
88 {144, 89, 131, 76, 141, 86, 127, 72, },
89 { 0, 165, 41, 206, 10, 175, 52, 217, },
90 {110, 55, 151, 96, 120, 65, 162, 107, },
91 { 28, 193, 14, 179, 38, 203, 24, 189, },
92 {138, 83, 124, 69, 148, 93, 134, 79, },
93 { 7, 172, 48, 213, 3, 168, 45, 210, },
94 {117, 62, 158, 103, 113, 58, 155, 100, },
95 };
96 #elif 1
97 // tries to correct a gamma of 1.5
98 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[][8]={
99 { 0, 143, 18, 200, 2, 156, 25, 215, },
100 { 78, 28, 125, 64, 89, 36, 138, 74, },
101 { 10, 180, 3, 161, 16, 195, 8, 175, },
102 {109, 51, 93, 38, 121, 60, 105, 47, },
103 { 1, 152, 23, 210, 0, 147, 20, 205, },
104 { 85, 33, 134, 71, 81, 30, 130, 67, },
105 { 14, 190, 6, 171, 12, 185, 5, 166, },
106 {117, 57, 101, 44, 113, 54, 97, 41, },
107 { 0, 143, 18, 200, 2, 156, 25, 215, },
108 };
109 #elif 1
110 // tries to correct a gamma of 2.0
111 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[][8]={
112 { 0, 124, 8, 193, 0, 140, 12, 213, },
113 { 55, 14, 104, 42, 66, 19, 119, 52, },
114 { 3, 168, 1, 145, 6, 187, 3, 162, },
115 { 86, 31, 70, 21, 99, 39, 82, 28, },
116 { 0, 134, 11, 206, 0, 129, 9, 200, },
117 { 62, 17, 114, 48, 58, 16, 109, 45, },
118 { 5, 181, 2, 157, 4, 175, 1, 151, },
119 { 95, 36, 78, 26, 90, 34, 74, 24, },
120 { 0, 124, 8, 193, 0, 140, 12, 213, },
121 };
122 #else
123 // tries to correct a gamma of 2.5
124 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[][8]={
125 { 0, 107, 3, 187, 0, 125, 6, 212, },
126 { 39, 7, 86, 28, 49, 11, 102, 36, },
127 { 1, 158, 0, 131, 3, 180, 1, 151, },
128 { 68, 19, 52, 12, 81, 25, 64, 17, },
129 { 0, 119, 5, 203, 0, 113, 4, 195, },
130 { 45, 9, 96, 33, 42, 8, 91, 30, },
131 { 2, 172, 1, 144, 2, 165, 0, 137, },
132 { 77, 23, 60, 15, 72, 21, 56, 14, },
133 { 0, 107, 3, 187, 0, 125, 6, 212, },
134 };
135 #endif
136 
137 #define output_pixel(pos, val, bias, signedness) \
138  if (big_endian) { \
139  AV_WB16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
140  } else { \
141  AV_WL16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
142  }
143 
144 static av_always_inline void
145 yuv2plane1_16_c_template(const int32_t *src, uint16_t *dest, int dstW,
146  int big_endian, int output_bits)
147 {
148  int i;
149  int shift = 3;
150  av_assert0(output_bits == 16);
151 
152  for (i = 0; i < dstW; i++) {
153  int val = src[i] + (1 << (shift - 1));
154  output_pixel(&dest[i], val, 0, uint);
155  }
156 }
157 
158 static av_always_inline void
159 yuv2planeX_16_c_template(const int16_t *filter, int filterSize,
160  const int32_t **src, uint16_t *dest, int dstW,
161  int big_endian, int output_bits)
162 {
163  int i;
164  int shift = 15;
165  av_assert0(output_bits == 16);
166 
167  for (i = 0; i < dstW; i++) {
168  int val = 1 << (shift - 1);
169  int j;
170 
171  /* range of val is [0,0x7FFFFFFF], so 31 bits, but with lanczos/spline
172  * filters (or anything with negative coeffs, the range can be slightly
173  * wider in both directions. To account for this overflow, we subtract
174  * a constant so it always fits in the signed range (assuming a
175  * reasonable filterSize), and re-add that at the end. */
176  val -= 0x40000000;
177  for (j = 0; j < filterSize; j++)
178  val += src[j][i] * (unsigned)filter[j];
179 
180  output_pixel(&dest[i], val, 0x8000, int);
181  }
182 }
183 
184 #undef output_pixel
185 
186 #define output_pixel(pos, val) \
187  if (big_endian) { \
188  AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
189  } else { \
190  AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
191  }
192 
193 static av_always_inline void
194 yuv2plane1_10_c_template(const int16_t *src, uint16_t *dest, int dstW,
195  int big_endian, int output_bits)
196 {
197  int i;
198  int shift = 15 - output_bits;
199 
200  for (i = 0; i < dstW; i++) {
201  int val = src[i] + (1 << (shift - 1));
202  output_pixel(&dest[i], val);
203  }
204 }
205 
206 static av_always_inline void
207 yuv2planeX_10_c_template(const int16_t *filter, int filterSize,
208  const int16_t **src, uint16_t *dest, int dstW,
209  int big_endian, int output_bits)
210 {
211  int i;
212  int shift = 11 + 16 - output_bits;
213 
214  for (i = 0; i < dstW; i++) {
215  int val = 1 << (shift - 1);
216  int j;
217 
218  for (j = 0; j < filterSize; j++)
219  val += src[j][i] * filter[j];
220 
221  output_pixel(&dest[i], val);
222  }
223 }
224 
225 #undef output_pixel
226 
227 #define yuv2NBPS(bits, BE_LE, is_be, template_size, typeX_t) \
228 static void yuv2plane1_ ## bits ## BE_LE ## _c(const int16_t *src, \
229  uint8_t *dest, int dstW, \
230  const uint8_t *dither, int offset)\
231 { \
232  yuv2plane1_ ## template_size ## _c_template((const typeX_t *) src, \
233  (uint16_t *) dest, dstW, is_be, bits); \
234 }\
235 static void yuv2planeX_ ## bits ## BE_LE ## _c(const int16_t *filter, int filterSize, \
236  const int16_t **src, uint8_t *dest, int dstW, \
237  const uint8_t *dither, int offset)\
238 { \
239  yuv2planeX_## template_size ## _c_template(filter, \
240  filterSize, (const typeX_t **) src, \
241  (uint16_t *) dest, dstW, is_be, bits); \
242 }
243 yuv2NBPS( 9, BE, 1, 10, int16_t)
244 yuv2NBPS( 9, LE, 0, 10, int16_t)
245 yuv2NBPS(10, BE, 1, 10, int16_t)
246 yuv2NBPS(10, LE, 0, 10, int16_t)
247 yuv2NBPS(12, BE, 1, 10, int16_t)
248 yuv2NBPS(12, LE, 0, 10, int16_t)
249 yuv2NBPS(14, BE, 1, 10, int16_t)
250 yuv2NBPS(14, LE, 0, 10, int16_t)
251 yuv2NBPS(16, BE, 1, 16, int32_t)
252 yuv2NBPS(16, LE, 0, 16, int32_t)
253 
254 static void yuv2planeX_8_c(const int16_t *filter, int filterSize,
255  const int16_t **src, uint8_t *dest, int dstW,
256  const uint8_t *dither, int offset)
257 {
258  int i;
259  for (i=0; i<dstW; i++) {
260  int val = dither[(i + offset) & 7] << 12;
261  int j;
262  for (j=0; j<filterSize; j++)
263  val += src[j][i] * filter[j];
264 
265  dest[i]= av_clip_uint8(val>>19);
266  }
267 }
268 
269 static void yuv2plane1_8_c(const int16_t *src, uint8_t *dest, int dstW,
270  const uint8_t *dither, int offset)
271 {
272  int i;
273  for (i=0; i<dstW; i++) {
274  int val = (src[i] + dither[(i + offset) & 7]) >> 7;
275  dest[i]= av_clip_uint8(val);
276  }
277 }
278 
279 static void yuv2nv12cX_c(SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
280  const int16_t **chrUSrc, const int16_t **chrVSrc,
281  uint8_t *dest, int chrDstW)
282 {
283  enum AVPixelFormat dstFormat = c->dstFormat;
284  const uint8_t *chrDither = c->chrDither8;
285  int i;
286 
287  if (dstFormat == AV_PIX_FMT_NV12)
288  for (i=0; i<chrDstW; i++) {
289  int u = chrDither[i & 7] << 12;
290  int v = chrDither[(i + 3) & 7] << 12;
291  int j;
292  for (j=0; j<chrFilterSize; j++) {
293  u += chrUSrc[j][i] * chrFilter[j];
294  v += chrVSrc[j][i] * chrFilter[j];
295  }
296 
297  dest[2*i]= av_clip_uint8(u>>19);
298  dest[2*i+1]= av_clip_uint8(v>>19);
299  }
300  else
301  for (i=0; i<chrDstW; i++) {
302  int u = chrDither[i & 7] << 12;
303  int v = chrDither[(i + 3) & 7] << 12;
304  int j;
305  for (j=0; j<chrFilterSize; j++) {
306  u += chrUSrc[j][i] * chrFilter[j];
307  v += chrVSrc[j][i] * chrFilter[j];
308  }
309 
310  dest[2*i]= av_clip_uint8(v>>19);
311  dest[2*i+1]= av_clip_uint8(u>>19);
312  }
313 }
314 
315 #define accumulate_bit(acc, val) \
316  acc <<= 1; \
317  acc |= (val) >= (128 + 110)
318 #define output_pixel(pos, acc) \
319  if (target == AV_PIX_FMT_MONOBLACK) { \
320  pos = acc; \
321  } else { \
322  pos = ~acc; \
323  }
324 
325 static av_always_inline void
326 yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
327  const int16_t **lumSrc, int lumFilterSize,
328  const int16_t *chrFilter, const int16_t **chrUSrc,
329  const int16_t **chrVSrc, int chrFilterSize,
330  const int16_t **alpSrc, uint8_t *dest, int dstW,
331  int y, enum AVPixelFormat target)
332 {
333  const uint8_t * const d128=dither_8x8_220[y&7];
334  int i;
335  unsigned acc = 0;
336  int err = 0;
337 
338  for (i = 0; i < dstW; i += 2) {
339  int j;
340  int Y1 = 1 << 18;
341  int Y2 = 1 << 18;
342 
343  for (j = 0; j < lumFilterSize; j++) {
344  Y1 += lumSrc[j][i] * lumFilter[j];
345  Y2 += lumSrc[j][i+1] * lumFilter[j];
346  }
347  Y1 >>= 19;
348  Y2 >>= 19;
349  if ((Y1 | Y2) & 0x100) {
350  Y1 = av_clip_uint8(Y1);
351  Y2 = av_clip_uint8(Y2);
352  }
353  if (c->flags & SWS_ERROR_DIFFUSION) {
354  Y1 += (7*err + 1*c->dither_error[0][i] + 5*c->dither_error[0][i+1] + 3*c->dither_error[0][i+2] + 8 - 256)>>4;
355  c->dither_error[0][i] = err;
356  acc = 2*acc + (Y1 >= 128);
357  Y1 -= 220*(acc&1);
358 
359  err = Y2 + ((7*Y1 + 1*c->dither_error[0][i+1] + 5*c->dither_error[0][i+2] + 3*c->dither_error[0][i+3] + 8 - 256)>>4);
360  c->dither_error[0][i+1] = Y1;
361  acc = 2*acc + (err >= 128);
362  err -= 220*(acc&1);
363  } else {
364  accumulate_bit(acc, Y1 + d128[(i + 0) & 7]);
365  accumulate_bit(acc, Y2 + d128[(i + 1) & 7]);
366  }
367  if ((i & 7) == 6) {
368  output_pixel(*dest++, acc);
369  }
370  }
371  c->dither_error[0][i] = err;
372 
373  if (i & 6) {
374  output_pixel(*dest, acc);
375  }
376 }
377 
378 static av_always_inline void
379 yuv2mono_2_c_template(SwsContext *c, const int16_t *buf[2],
380  const int16_t *ubuf[2], const int16_t *vbuf[2],
381  const int16_t *abuf[2], uint8_t *dest, int dstW,
382  int yalpha, int uvalpha, int y,
383  enum AVPixelFormat target)
384 {
385  const int16_t *buf0 = buf[0], *buf1 = buf[1];
386  const uint8_t * const d128 = dither_8x8_220[y & 7];
387  int yalpha1 = 4096 - yalpha;
388  int i;
389 
390  if (c->flags & SWS_ERROR_DIFFUSION) {
391  int err = 0;
392  int acc = 0;
393  for (i = 0; i < dstW; i +=2) {
394  int Y;
395 
396  Y = (buf0[i + 0] * yalpha1 + buf1[i + 0] * yalpha) >> 19;
397  Y += (7*err + 1*c->dither_error[0][i] + 5*c->dither_error[0][i+1] + 3*c->dither_error[0][i+2] + 8 - 256)>>4;
398  c->dither_error[0][i] = err;
399  acc = 2*acc + (Y >= 128);
400  Y -= 220*(acc&1);
401 
402  err = (buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19;
403  err += (7*Y + 1*c->dither_error[0][i+1] + 5*c->dither_error[0][i+2] + 3*c->dither_error[0][i+3] + 8 - 256)>>4;
404  c->dither_error[0][i+1] = Y;
405  acc = 2*acc + (err >= 128);
406  err -= 220*(acc&1);
407 
408  if ((i & 7) == 6)
409  output_pixel(*dest++, acc);
410  }
411  c->dither_error[0][i] = err;
412  } else {
413  for (i = 0; i < dstW; i += 8) {
414  int Y, acc = 0;
415 
416  Y = (buf0[i + 0] * yalpha1 + buf1[i + 0] * yalpha) >> 19;
417  accumulate_bit(acc, Y + d128[0]);
418  Y = (buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19;
419  accumulate_bit(acc, Y + d128[1]);
420  Y = (buf0[i + 2] * yalpha1 + buf1[i + 2] * yalpha) >> 19;
421  accumulate_bit(acc, Y + d128[2]);
422  Y = (buf0[i + 3] * yalpha1 + buf1[i + 3] * yalpha) >> 19;
423  accumulate_bit(acc, Y + d128[3]);
424  Y = (buf0[i + 4] * yalpha1 + buf1[i + 4] * yalpha) >> 19;
425  accumulate_bit(acc, Y + d128[4]);
426  Y = (buf0[i + 5] * yalpha1 + buf1[i + 5] * yalpha) >> 19;
427  accumulate_bit(acc, Y + d128[5]);
428  Y = (buf0[i + 6] * yalpha1 + buf1[i + 6] * yalpha) >> 19;
429  accumulate_bit(acc, Y + d128[6]);
430  Y = (buf0[i + 7] * yalpha1 + buf1[i + 7] * yalpha) >> 19;
431  accumulate_bit(acc, Y + d128[7]);
432 
433  output_pixel(*dest++, acc);
434  }
435  }
436 }
437 
438 static av_always_inline void
439 yuv2mono_1_c_template(SwsContext *c, const int16_t *buf0,
440  const int16_t *ubuf[2], const int16_t *vbuf[2],
441  const int16_t *abuf0, uint8_t *dest, int dstW,
442  int uvalpha, int y, enum AVPixelFormat target)
443 {
444  const uint8_t * const d128 = dither_8x8_220[y & 7];
445  int i;
446 
447  if (c->flags & SWS_ERROR_DIFFUSION) {
448  int err = 0;
449  int acc = 0;
450  for (i = 0; i < dstW; i +=2) {
451  int Y;
452 
453  Y = ((buf0[i + 0] + 64) >> 7);
454  Y += (7*err + 1*c->dither_error[0][i] + 5*c->dither_error[0][i+1] + 3*c->dither_error[0][i+2] + 8 - 256)>>4;
455  c->dither_error[0][i] = err;
456  acc = 2*acc + (Y >= 128);
457  Y -= 220*(acc&1);
458 
459  err = ((buf0[i + 1] + 64) >> 7);
460  err += (7*Y + 1*c->dither_error[0][i+1] + 5*c->dither_error[0][i+2] + 3*c->dither_error[0][i+3] + 8 - 256)>>4;
461  c->dither_error[0][i+1] = Y;
462  acc = 2*acc + (err >= 128);
463  err -= 220*(acc&1);
464 
465  if ((i & 7) == 6)
466  output_pixel(*dest++, acc);
467  }
468  c->dither_error[0][i] = err;
469  } else {
470  for (i = 0; i < dstW; i += 8) {
471  int acc = 0;
472  accumulate_bit(acc, ((buf0[i + 0] + 64) >> 7) + d128[0]);
473  accumulate_bit(acc, ((buf0[i + 1] + 64) >> 7) + d128[1]);
474  accumulate_bit(acc, ((buf0[i + 2] + 64) >> 7) + d128[2]);
475  accumulate_bit(acc, ((buf0[i + 3] + 64) >> 7) + d128[3]);
476  accumulate_bit(acc, ((buf0[i + 4] + 64) >> 7) + d128[4]);
477  accumulate_bit(acc, ((buf0[i + 5] + 64) >> 7) + d128[5]);
478  accumulate_bit(acc, ((buf0[i + 6] + 64) >> 7) + d128[6]);
479  accumulate_bit(acc, ((buf0[i + 7] + 64) >> 7) + d128[7]);
480 
481  output_pixel(*dest++, acc);
482  }
483  }
484 }
485 
486 #undef output_pixel
487 #undef accumulate_bit
488 
489 #define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
490 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
491  const int16_t **lumSrc, int lumFilterSize, \
492  const int16_t *chrFilter, const int16_t **chrUSrc, \
493  const int16_t **chrVSrc, int chrFilterSize, \
494  const int16_t **alpSrc, uint8_t *dest, int dstW, \
495  int y) \
496 { \
497  name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
498  chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
499  alpSrc, dest, dstW, y, fmt); \
500 } \
501  \
502 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
503  const int16_t *ubuf[2], const int16_t *vbuf[2], \
504  const int16_t *abuf[2], uint8_t *dest, int dstW, \
505  int yalpha, int uvalpha, int y) \
506 { \
507  name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
508  dest, dstW, yalpha, uvalpha, y, fmt); \
509 } \
510  \
511 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
512  const int16_t *ubuf[2], const int16_t *vbuf[2], \
513  const int16_t *abuf0, uint8_t *dest, int dstW, \
514  int uvalpha, int y) \
515 { \
516  name ## base ## _1_c_template(c, buf0, ubuf, vbuf, \
517  abuf0, dest, dstW, uvalpha, \
518  y, fmt); \
519 }
520 
521 YUV2PACKEDWRAPPER(yuv2mono,, white, AV_PIX_FMT_MONOWHITE)
522 YUV2PACKEDWRAPPER(yuv2mono,, black, AV_PIX_FMT_MONOBLACK)
523 
524 #define output_pixels(pos, Y1, U, Y2, V) \
525  if (target == AV_PIX_FMT_YUYV422) { \
526  dest[pos + 0] = Y1; \
527  dest[pos + 1] = U; \
528  dest[pos + 2] = Y2; \
529  dest[pos + 3] = V; \
530  } else { \
531  dest[pos + 0] = U; \
532  dest[pos + 1] = Y1; \
533  dest[pos + 2] = V; \
534  dest[pos + 3] = Y2; \
535  }
536 
537 static av_always_inline void
538 yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
539  const int16_t **lumSrc, int lumFilterSize,
540  const int16_t *chrFilter, const int16_t **chrUSrc,
541  const int16_t **chrVSrc, int chrFilterSize,
542  const int16_t **alpSrc, uint8_t *dest, int dstW,
543  int y, enum AVPixelFormat target)
544 {
545  int i;
546 
547  for (i = 0; i < ((dstW + 1) >> 1); i++) {
548  int j;
549  int Y1 = 1 << 18;
550  int Y2 = 1 << 18;
551  int U = 1 << 18;
552  int V = 1 << 18;
553 
554  for (j = 0; j < lumFilterSize; j++) {
555  Y1 += lumSrc[j][i * 2] * lumFilter[j];
556  Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
557  }
558  for (j = 0; j < chrFilterSize; j++) {
559  U += chrUSrc[j][i] * chrFilter[j];
560  V += chrVSrc[j][i] * chrFilter[j];
561  }
562  Y1 >>= 19;
563  Y2 >>= 19;
564  U >>= 19;
565  V >>= 19;
566  if ((Y1 | Y2 | U | V) & 0x100) {
567  Y1 = av_clip_uint8(Y1);
568  Y2 = av_clip_uint8(Y2);
569  U = av_clip_uint8(U);
570  V = av_clip_uint8(V);
571  }
572  output_pixels(4*i, Y1, U, Y2, V);
573  }
574 }
575 
576 static av_always_inline void
577 yuv2422_2_c_template(SwsContext *c, const int16_t *buf[2],
578  const int16_t *ubuf[2], const int16_t *vbuf[2],
579  const int16_t *abuf[2], uint8_t *dest, int dstW,
580  int yalpha, int uvalpha, int y,
581  enum AVPixelFormat target)
582 {
583  const int16_t *buf0 = buf[0], *buf1 = buf[1],
584  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
585  *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
586  int yalpha1 = 4096 - yalpha;
587  int uvalpha1 = 4096 - uvalpha;
588  int i;
589 
590  for (i = 0; i < ((dstW + 1) >> 1); i++) {
591  int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
592  int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
593  int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
594  int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
595 
596  if ((Y1 | Y2 | U | V) & 0x100) {
597  Y1 = av_clip_uint8(Y1);
598  Y2 = av_clip_uint8(Y2);
599  U = av_clip_uint8(U);
600  V = av_clip_uint8(V);
601  }
602 
603  output_pixels(i * 4, Y1, U, Y2, V);
604  }
605 }
606 
607 static av_always_inline void
608 yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
609  const int16_t *ubuf[2], const int16_t *vbuf[2],
610  const int16_t *abuf0, uint8_t *dest, int dstW,
611  int uvalpha, int y, enum AVPixelFormat target)
612 {
613  const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
614  int i;
615 
616  if (uvalpha < 2048) {
617  for (i = 0; i < ((dstW + 1) >> 1); i++) {
618  int Y1 = (buf0[i * 2 ]+64) >> 7;
619  int Y2 = (buf0[i * 2 + 1]+64) >> 7;
620  int U = (ubuf0[i] +64) >> 7;
621  int V = (vbuf0[i] +64) >> 7;
622 
623  if ((Y1 | Y2 | U | V) & 0x100) {
624  Y1 = av_clip_uint8(Y1);
625  Y2 = av_clip_uint8(Y2);
626  U = av_clip_uint8(U);
627  V = av_clip_uint8(V);
628  }
629 
630  Y1 = av_clip_uint8(Y1);
631  Y2 = av_clip_uint8(Y2);
632  U = av_clip_uint8(U);
633  V = av_clip_uint8(V);
634 
635  output_pixels(i * 4, Y1, U, Y2, V);
636  }
637  } else {
638  const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
639  for (i = 0; i < ((dstW + 1) >> 1); i++) {
640  int Y1 = (buf0[i * 2 ] + 64) >> 7;
641  int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
642  int U = (ubuf0[i] + ubuf1[i]+128) >> 8;
643  int V = (vbuf0[i] + vbuf1[i]+128) >> 8;
644 
645  if ((Y1 | Y2 | U | V) & 0x100) {
646  Y1 = av_clip_uint8(Y1);
647  Y2 = av_clip_uint8(Y2);
648  U = av_clip_uint8(U);
649  V = av_clip_uint8(V);
650  }
651 
652  Y1 = av_clip_uint8(Y1);
653  Y2 = av_clip_uint8(Y2);
654  U = av_clip_uint8(U);
655  V = av_clip_uint8(V);
656 
657  output_pixels(i * 4, Y1, U, Y2, V);
658  }
659  }
660 }
661 
662 #undef output_pixels
663 
664 YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, AV_PIX_FMT_YUYV422)
665 YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, AV_PIX_FMT_UYVY422)
666 
667 #define R_B ((target == AV_PIX_FMT_RGB48LE || target == AV_PIX_FMT_RGB48BE) ? R : B)
668 #define B_R ((target == AV_PIX_FMT_RGB48LE || target == AV_PIX_FMT_RGB48BE) ? B : R)
669 #define output_pixel(pos, val) \
670  if (isBE(target)) { \
671  AV_WB16(pos, val); \
672  } else { \
673  AV_WL16(pos, val); \
674  }
675 
676 static av_always_inline void
677 yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
678  const int32_t **lumSrc, int lumFilterSize,
679  const int16_t *chrFilter, const int32_t **chrUSrc,
680  const int32_t **chrVSrc, int chrFilterSize,
681  const int32_t **alpSrc, uint16_t *dest, int dstW,
682  int y, enum AVPixelFormat target)
683 {
684  int i;
685 
686  for (i = 0; i < ((dstW + 1) >> 1); i++) {
687  int j;
688  int Y1 = -0x40000000;
689  int Y2 = -0x40000000;
690  int U = -128 << 23; // 19
691  int V = -128 << 23;
692  int R, G, B;
693 
694  for (j = 0; j < lumFilterSize; j++) {
695  Y1 += lumSrc[j][i * 2] * (unsigned)lumFilter[j];
696  Y2 += lumSrc[j][i * 2 + 1] * (unsigned)lumFilter[j];
697  }
698  for (j = 0; j < chrFilterSize; j++) {;
699  U += chrUSrc[j][i] * (unsigned)chrFilter[j];
700  V += chrVSrc[j][i] * (unsigned)chrFilter[j];
701  }
702 
703  // 8bit: 12+15=27; 16-bit: 12+19=31
704  Y1 >>= 14; // 10
705  Y1 += 0x10000;
706  Y2 >>= 14;
707  Y2 += 0x10000;
708  U >>= 14;
709  V >>= 14;
710 
711  // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
712  Y1 -= c->yuv2rgb_y_offset;
713  Y2 -= c->yuv2rgb_y_offset;
714  Y1 *= c->yuv2rgb_y_coeff;
715  Y2 *= c->yuv2rgb_y_coeff;
716  Y1 += 1 << 13; // 21
717  Y2 += 1 << 13;
718  // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
719 
720  R = V * c->yuv2rgb_v2r_coeff;
721  G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
722  B = U * c->yuv2rgb_u2b_coeff;
723 
724  // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
725  output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
726  output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
727  output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
728  output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
729  output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
730  output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
731  dest += 6;
732  }
733 }
734 
735 static av_always_inline void
736 yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
737  const int32_t *ubuf[2], const int32_t *vbuf[2],
738  const int32_t *abuf[2], uint16_t *dest, int dstW,
739  int yalpha, int uvalpha, int y,
740  enum AVPixelFormat target)
741 {
742  const int32_t *buf0 = buf[0], *buf1 = buf[1],
743  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
744  *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
745  int yalpha1 = 4096 - yalpha;
746  int uvalpha1 = 4096 - uvalpha;
747  int i;
748 
749  for (i = 0; i < ((dstW + 1) >> 1); i++) {
750  int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 14;
751  int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 14;
752  int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha + (-128 << 23)) >> 14;
753  int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha + (-128 << 23)) >> 14;
754  int R, G, B;
755 
756  Y1 -= c->yuv2rgb_y_offset;
757  Y2 -= c->yuv2rgb_y_offset;
758  Y1 *= c->yuv2rgb_y_coeff;
759  Y2 *= c->yuv2rgb_y_coeff;
760  Y1 += 1 << 13;
761  Y2 += 1 << 13;
762 
763  R = V * c->yuv2rgb_v2r_coeff;
764  G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
765  B = U * c->yuv2rgb_u2b_coeff;
766 
767  output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
768  output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
769  output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
770  output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
771  output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
772  output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
773  dest += 6;
774  }
775 }
776 
777 static av_always_inline void
778 yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
779  const int32_t *ubuf[2], const int32_t *vbuf[2],
780  const int32_t *abuf0, uint16_t *dest, int dstW,
781  int uvalpha, int y, enum AVPixelFormat target)
782 {
783  const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
784  int i;
785 
786  if (uvalpha < 2048) {
787  for (i = 0; i < ((dstW + 1) >> 1); i++) {
788  int Y1 = (buf0[i * 2] ) >> 2;
789  int Y2 = (buf0[i * 2 + 1]) >> 2;
790  int U = (ubuf0[i] + (-128 << 11)) >> 2;
791  int V = (vbuf0[i] + (-128 << 11)) >> 2;
792  int R, G, B;
793 
794  Y1 -= c->yuv2rgb_y_offset;
795  Y2 -= c->yuv2rgb_y_offset;
796  Y1 *= c->yuv2rgb_y_coeff;
797  Y2 *= c->yuv2rgb_y_coeff;
798  Y1 += 1 << 13;
799  Y2 += 1 << 13;
800 
801  R = V * c->yuv2rgb_v2r_coeff;
802  G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
803  B = U * c->yuv2rgb_u2b_coeff;
804 
805  output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
806  output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
807  output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
808  output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
809  output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
810  output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
811  dest += 6;
812  }
813  } else {
814  const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
815  for (i = 0; i < ((dstW + 1) >> 1); i++) {
816  int Y1 = (buf0[i * 2] ) >> 2;
817  int Y2 = (buf0[i * 2 + 1]) >> 2;
818  int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
819  int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
820  int R, G, B;
821 
822  Y1 -= c->yuv2rgb_y_offset;
823  Y2 -= c->yuv2rgb_y_offset;
824  Y1 *= c->yuv2rgb_y_coeff;
825  Y2 *= c->yuv2rgb_y_coeff;
826  Y1 += 1 << 13;
827  Y2 += 1 << 13;
828 
829  R = V * c->yuv2rgb_v2r_coeff;
830  G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
831  B = U * c->yuv2rgb_u2b_coeff;
832 
833  output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
834  output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
835  output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
836  output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
837  output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
838  output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
839  dest += 6;
840  }
841  }
842 }
843 
844 #undef output_pixel
845 #undef r_b
846 #undef b_r
847 
848 #define YUV2PACKED16WRAPPER(name, base, ext, fmt) \
849 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
850  const int16_t **_lumSrc, int lumFilterSize, \
851  const int16_t *chrFilter, const int16_t **_chrUSrc, \
852  const int16_t **_chrVSrc, int chrFilterSize, \
853  const int16_t **_alpSrc, uint8_t *_dest, int dstW, \
854  int y) \
855 { \
856  const int32_t **lumSrc = (const int32_t **) _lumSrc, \
857  **chrUSrc = (const int32_t **) _chrUSrc, \
858  **chrVSrc = (const int32_t **) _chrVSrc, \
859  **alpSrc = (const int32_t **) _alpSrc; \
860  uint16_t *dest = (uint16_t *) _dest; \
861  name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
862  chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
863  alpSrc, dest, dstW, y, fmt); \
864 } \
865  \
866 static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
867  const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
868  const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
869  int yalpha, int uvalpha, int y) \
870 { \
871  const int32_t **buf = (const int32_t **) _buf, \
872  **ubuf = (const int32_t **) _ubuf, \
873  **vbuf = (const int32_t **) _vbuf, \
874  **abuf = (const int32_t **) _abuf; \
875  uint16_t *dest = (uint16_t *) _dest; \
876  name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
877  dest, dstW, yalpha, uvalpha, y, fmt); \
878 } \
879  \
880 static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
881  const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
882  const int16_t *_abuf0, uint8_t *_dest, int dstW, \
883  int uvalpha, int y) \
884 { \
885  const int32_t *buf0 = (const int32_t *) _buf0, \
886  **ubuf = (const int32_t **) _ubuf, \
887  **vbuf = (const int32_t **) _vbuf, \
888  *abuf0 = (const int32_t *) _abuf0; \
889  uint16_t *dest = (uint16_t *) _dest; \
890  name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
891  dstW, uvalpha, y, fmt); \
892 }
893 
894 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, AV_PIX_FMT_RGB48BE)
895 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, AV_PIX_FMT_RGB48LE)
896 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, AV_PIX_FMT_BGR48BE)
897 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, AV_PIX_FMT_BGR48LE)
898 
899 /*
900  * Write out 2 RGB pixels in the target pixel format. This function takes a
901  * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
902  * things like endianness conversion and shifting. The caller takes care of
903  * setting the correct offset in these tables from the chroma (U/V) values.
904  * This function then uses the luminance (Y1/Y2) values to write out the
905  * correct RGB values into the destination buffer.
906  */
907 static av_always_inline void
908 yuv2rgb_write(uint8_t *_dest, int i, int Y1, int Y2,
909  unsigned A1, unsigned A2,
910  const void *_r, const void *_g, const void *_b, int y,
911  enum AVPixelFormat target, int hasAlpha)
912 {
913  if (target == AV_PIX_FMT_ARGB || target == AV_PIX_FMT_RGBA ||
914  target == AV_PIX_FMT_ABGR || target == AV_PIX_FMT_BGRA) {
915  uint32_t *dest = (uint32_t *) _dest;
916  const uint32_t *r = (const uint32_t *) _r;
917  const uint32_t *g = (const uint32_t *) _g;
918  const uint32_t *b = (const uint32_t *) _b;
919 
920 #if CONFIG_SMALL
921  int sh = hasAlpha ? ((target == AV_PIX_FMT_RGB32_1 || target == AV_PIX_FMT_BGR32_1) ? 0 : 24) : 0;
922 
923  dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
924  dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
925 #else
926  if (hasAlpha) {
927  int sh = (target == AV_PIX_FMT_RGB32_1 || target == AV_PIX_FMT_BGR32_1) ? 0 : 24;
928 
929  dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
930  dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
931  } else {
932  dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
933  dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
934  }
935 #endif
936  } else if (target == AV_PIX_FMT_RGB24 || target == AV_PIX_FMT_BGR24) {
937  uint8_t *dest = (uint8_t *) _dest;
938  const uint8_t *r = (const uint8_t *) _r;
939  const uint8_t *g = (const uint8_t *) _g;
940  const uint8_t *b = (const uint8_t *) _b;
941 
942 #define r_b ((target == AV_PIX_FMT_RGB24) ? r : b)
943 #define b_r ((target == AV_PIX_FMT_RGB24) ? b : r)
944 
945  dest[i * 6 + 0] = r_b[Y1];
946  dest[i * 6 + 1] = g[Y1];
947  dest[i * 6 + 2] = b_r[Y1];
948  dest[i * 6 + 3] = r_b[Y2];
949  dest[i * 6 + 4] = g[Y2];
950  dest[i * 6 + 5] = b_r[Y2];
951 #undef r_b
952 #undef b_r
953  } else if (target == AV_PIX_FMT_RGB565 || target == AV_PIX_FMT_BGR565 ||
954  target == AV_PIX_FMT_RGB555 || target == AV_PIX_FMT_BGR555 ||
955  target == AV_PIX_FMT_RGB444 || target == AV_PIX_FMT_BGR444) {
956  uint16_t *dest = (uint16_t *) _dest;
957  const uint16_t *r = (const uint16_t *) _r;
958  const uint16_t *g = (const uint16_t *) _g;
959  const uint16_t *b = (const uint16_t *) _b;
960  int dr1, dg1, db1, dr2, dg2, db2;
961 
962  if (target == AV_PIX_FMT_RGB565 || target == AV_PIX_FMT_BGR565) {
963  dr1 = dither_2x2_8[ y & 1 ][0];
964  dg1 = dither_2x2_4[ y & 1 ][0];
965  db1 = dither_2x2_8[(y & 1) ^ 1][0];
966  dr2 = dither_2x2_8[ y & 1 ][1];
967  dg2 = dither_2x2_4[ y & 1 ][1];
968  db2 = dither_2x2_8[(y & 1) ^ 1][1];
969  } else if (target == AV_PIX_FMT_RGB555 || target == AV_PIX_FMT_BGR555) {
970  dr1 = dither_2x2_8[ y & 1 ][0];
971  dg1 = dither_2x2_8[ y & 1 ][1];
972  db1 = dither_2x2_8[(y & 1) ^ 1][0];
973  dr2 = dither_2x2_8[ y & 1 ][1];
974  dg2 = dither_2x2_8[ y & 1 ][0];
975  db2 = dither_2x2_8[(y & 1) ^ 1][1];
976  } else {
977  dr1 = dither_4x4_16[ y & 3 ][0];
978  dg1 = dither_4x4_16[ y & 3 ][1];
979  db1 = dither_4x4_16[(y & 3) ^ 3][0];
980  dr2 = dither_4x4_16[ y & 3 ][1];
981  dg2 = dither_4x4_16[ y & 3 ][0];
982  db2 = dither_4x4_16[(y & 3) ^ 3][1];
983  }
984 
985  dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
986  dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
987  } else /* 8/4-bit */ {
988  uint8_t *dest = (uint8_t *) _dest;
989  const uint8_t *r = (const uint8_t *) _r;
990  const uint8_t *g = (const uint8_t *) _g;
991  const uint8_t *b = (const uint8_t *) _b;
992  int dr1, dg1, db1, dr2, dg2, db2;
993 
994  if (target == AV_PIX_FMT_RGB8 || target == AV_PIX_FMT_BGR8) {
995  const uint8_t * const d64 = dither_8x8_73[y & 7];
996  const uint8_t * const d32 = dither_8x8_32[y & 7];
997  dr1 = dg1 = d32[(i * 2 + 0) & 7];
998  db1 = d64[(i * 2 + 0) & 7];
999  dr2 = dg2 = d32[(i * 2 + 1) & 7];
1000  db2 = d64[(i * 2 + 1) & 7];
1001  } else {
1002  const uint8_t * const d64 = dither_8x8_73 [y & 7];
1003  const uint8_t * const d128 = dither_8x8_220[y & 7];
1004  dr1 = db1 = d128[(i * 2 + 0) & 7];
1005  dg1 = d64[(i * 2 + 0) & 7];
1006  dr2 = db2 = d128[(i * 2 + 1) & 7];
1007  dg2 = d64[(i * 2 + 1) & 7];
1008  }
1009 
1010  if (target == AV_PIX_FMT_RGB4 || target == AV_PIX_FMT_BGR4) {
1011  dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
1012  ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
1013  } else {
1014  dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
1015  dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
1016  }
1017  }
1018 }
1019 
1020 static av_always_inline void
1021 yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
1022  const int16_t **lumSrc, int lumFilterSize,
1023  const int16_t *chrFilter, const int16_t **chrUSrc,
1024  const int16_t **chrVSrc, int chrFilterSize,
1025  const int16_t **alpSrc, uint8_t *dest, int dstW,
1026  int y, enum AVPixelFormat target, int hasAlpha)
1027 {
1028  int i;
1029 
1030  for (i = 0; i < ((dstW + 1) >> 1); i++) {
1031  int j, A1, A2;
1032  int Y1 = 1 << 18;
1033  int Y2 = 1 << 18;
1034  int U = 1 << 18;
1035  int V = 1 << 18;
1036  const void *r, *g, *b;
1037 
1038  for (j = 0; j < lumFilterSize; j++) {
1039  Y1 += lumSrc[j][i * 2] * lumFilter[j];
1040  Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
1041  }
1042  for (j = 0; j < chrFilterSize; j++) {
1043  U += chrUSrc[j][i] * chrFilter[j];
1044  V += chrVSrc[j][i] * chrFilter[j];
1045  }
1046  Y1 >>= 19;
1047  Y2 >>= 19;
1048  U >>= 19;
1049  V >>= 19;
1050  if (hasAlpha) {
1051  A1 = 1 << 18;
1052  A2 = 1 << 18;
1053  for (j = 0; j < lumFilterSize; j++) {
1054  A1 += alpSrc[j][i * 2 ] * lumFilter[j];
1055  A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
1056  }
1057  A1 >>= 19;
1058  A2 >>= 19;
1059  if ((A1 | A2) & 0x100) {
1060  A1 = av_clip_uint8(A1);
1061  A2 = av_clip_uint8(A2);
1062  }
1063  }
1064 
1065  r = c->table_rV[V + YUVRGB_TABLE_HEADROOM];
1066  g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]);
1067  b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1068 
1069  yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1070  r, g, b, y, target, hasAlpha);
1071  }
1072 }
1073 
1074 static av_always_inline void
1075 yuv2rgb_2_c_template(SwsContext *c, const int16_t *buf[2],
1076  const int16_t *ubuf[2], const int16_t *vbuf[2],
1077  const int16_t *abuf[2], uint8_t *dest, int dstW,
1078  int yalpha, int uvalpha, int y,
1079  enum AVPixelFormat target, int hasAlpha)
1080 {
1081  const int16_t *buf0 = buf[0], *buf1 = buf[1],
1082  *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
1083  *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
1084  *abuf0 = hasAlpha ? abuf[0] : NULL,
1085  *abuf1 = hasAlpha ? abuf[1] : NULL;
1086  int yalpha1 = 4096 - yalpha;
1087  int uvalpha1 = 4096 - uvalpha;
1088  int i;
1089 
1090  for (i = 0; i < ((dstW + 1) >> 1); i++) {
1091  int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
1092  int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
1093  int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
1094  int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
1095  int A1, A2;
1096  const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1097  *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1098  *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1099 
1100  if (hasAlpha) {
1101  A1 = (abuf0[i * 2 ] * yalpha1 + abuf1[i * 2 ] * yalpha) >> 19;
1102  A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
1103  A1 = av_clip_uint8(A1);
1104  A2 = av_clip_uint8(A2);
1105  }
1106 
1107  yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1108  r, g, b, y, target, hasAlpha);
1109  }
1110 }
1111 
1112 static av_always_inline void
1113 yuv2rgb_1_c_template(SwsContext *c, const int16_t *buf0,
1114  const int16_t *ubuf[2], const int16_t *vbuf[2],
1115  const int16_t *abuf0, uint8_t *dest, int dstW,
1116  int uvalpha, int y, enum AVPixelFormat target,
1117  int hasAlpha)
1118 {
1119  const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
1120  int i;
1121 
1122  if (uvalpha < 2048) {
1123  for (i = 0; i < ((dstW + 1) >> 1); i++) {
1124  int Y1 = (buf0[i * 2 ] + 64) >> 7;
1125  int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
1126  int U = (ubuf0[i] + 64) >> 7;
1127  int V = (vbuf0[i] + 64) >> 7;
1128  int A1, A2;
1129  const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1130  *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1131  *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1132 
1133  if (hasAlpha) {
1134  A1 = abuf0[i * 2 ] * 255 + 16384 >> 15;
1135  A2 = abuf0[i * 2 + 1] * 255 + 16384 >> 15;
1136  A1 = av_clip_uint8(A1);
1137  A2 = av_clip_uint8(A2);
1138  }
1139 
1140  yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1141  r, g, b, y, target, hasAlpha);
1142  }
1143  } else {
1144  const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
1145  for (i = 0; i < ((dstW + 1) >> 1); i++) {
1146  int Y1 = (buf0[i * 2 ] + 64) >> 7;
1147  int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
1148  int U = (ubuf0[i] + ubuf1[i] + 128) >> 8;
1149  int V = (vbuf0[i] + vbuf1[i] + 128) >> 8;
1150  int A1, A2;
1151  const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1152  *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1153  *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1154 
1155  if (hasAlpha) {
1156  A1 = (abuf0[i * 2 ] + 64) >> 7;
1157  A2 = (abuf0[i * 2 + 1] + 64) >> 7;
1158  A1 = av_clip_uint8(A1);
1159  A2 = av_clip_uint8(A2);
1160  }
1161 
1162  yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1163  r, g, b, y, target, hasAlpha);
1164  }
1165  }
1166 }
1167 
1168 #define YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1169 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
1170  const int16_t **lumSrc, int lumFilterSize, \
1171  const int16_t *chrFilter, const int16_t **chrUSrc, \
1172  const int16_t **chrVSrc, int chrFilterSize, \
1173  const int16_t **alpSrc, uint8_t *dest, int dstW, \
1174  int y) \
1175 { \
1176  name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
1177  chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
1178  alpSrc, dest, dstW, y, fmt, hasAlpha); \
1179 }
1180 #define YUV2RGBWRAPPER(name, base, ext, fmt, hasAlpha) \
1181 YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1182 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
1183  const int16_t *ubuf[2], const int16_t *vbuf[2], \
1184  const int16_t *abuf[2], uint8_t *dest, int dstW, \
1185  int yalpha, int uvalpha, int y) \
1186 { \
1187  name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
1188  dest, dstW, yalpha, uvalpha, y, fmt, hasAlpha); \
1189 } \
1190  \
1191 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
1192  const int16_t *ubuf[2], const int16_t *vbuf[2], \
1193  const int16_t *abuf0, uint8_t *dest, int dstW, \
1194  int uvalpha, int y) \
1195 { \
1196  name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
1197  dstW, uvalpha, y, fmt, hasAlpha); \
1198 }
1199 
1200 #if CONFIG_SMALL
1201 YUV2RGBWRAPPER(yuv2rgb,, 32_1, AV_PIX_FMT_RGB32_1, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1202 YUV2RGBWRAPPER(yuv2rgb,, 32, AV_PIX_FMT_RGB32, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1203 #else
1204 #if CONFIG_SWSCALE_ALPHA
1205 YUV2RGBWRAPPER(yuv2rgb,, a32_1, AV_PIX_FMT_RGB32_1, 1)
1206 YUV2RGBWRAPPER(yuv2rgb,, a32, AV_PIX_FMT_RGB32, 1)
1207 #endif
1208 YUV2RGBWRAPPER(yuv2rgb,, x32_1, AV_PIX_FMT_RGB32_1, 0)
1209 YUV2RGBWRAPPER(yuv2rgb,, x32, AV_PIX_FMT_RGB32, 0)
1210 #endif
1211 YUV2RGBWRAPPER(yuv2, rgb, rgb24, AV_PIX_FMT_RGB24, 0)
1212 YUV2RGBWRAPPER(yuv2, rgb, bgr24, AV_PIX_FMT_BGR24, 0)
1213 YUV2RGBWRAPPER(yuv2rgb,, 16, AV_PIX_FMT_RGB565, 0)
1214 YUV2RGBWRAPPER(yuv2rgb,, 15, AV_PIX_FMT_RGB555, 0)
1215 YUV2RGBWRAPPER(yuv2rgb,, 12, AV_PIX_FMT_RGB444, 0)
1216 YUV2RGBWRAPPER(yuv2rgb,, 8, AV_PIX_FMT_RGB8, 0)
1217 YUV2RGBWRAPPER(yuv2rgb,, 4, AV_PIX_FMT_RGB4, 0)
1218 YUV2RGBWRAPPER(yuv2rgb,, 4b, AV_PIX_FMT_RGB4_BYTE, 0)
1219 
1220 static av_always_inline void
1221 yuv2rgb_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
1222  const int16_t **lumSrc, int lumFilterSize,
1223  const int16_t *chrFilter, const int16_t **chrUSrc,
1224  const int16_t **chrVSrc, int chrFilterSize,
1225  const int16_t **alpSrc, uint8_t *dest,
1226  int dstW, int y, enum AVPixelFormat target, int hasAlpha)
1227 {
1228  int i;
1229  int step = (target == AV_PIX_FMT_RGB24 || target == AV_PIX_FMT_BGR24) ? 3 : 4;
1230  int err[4] = {0};
1231  int isrgb8 = target == AV_PIX_FMT_BGR8 || target == AV_PIX_FMT_RGB8;
1232 
1233  for (i = 0; i < dstW; i++) {
1234  int j;
1235  int Y = 1<<9;
1236  int U = (1<<9)-(128 << 19);
1237  int V = (1<<9)-(128 << 19);
1238  int R, G, B, A;
1239 
1240  for (j = 0; j < lumFilterSize; j++) {
1241  Y += lumSrc[j][i] * lumFilter[j];
1242  }
1243  for (j = 0; j < chrFilterSize; j++) {
1244  U += chrUSrc[j][i] * chrFilter[j];
1245  V += chrVSrc[j][i] * chrFilter[j];
1246  }
1247  Y >>= 10;
1248  U >>= 10;
1249  V >>= 10;
1250  if (hasAlpha) {
1251  A = 1 << 18;
1252  for (j = 0; j < lumFilterSize; j++) {
1253  A += alpSrc[j][i] * lumFilter[j];
1254  }
1255  A >>= 19;
1256  if (A & 0x100)
1257  A = av_clip_uint8(A);
1258  }
1259  Y -= c->yuv2rgb_y_offset;
1260  Y *= c->yuv2rgb_y_coeff;
1261  Y += 1 << 21;
1262  R = Y + V*c->yuv2rgb_v2r_coeff;
1263  G = Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;
1264  B = Y + U*c->yuv2rgb_u2b_coeff;
1265  if ((R | G | B) & 0xC0000000) {
1266  R = av_clip_uintp2(R, 30);
1267  G = av_clip_uintp2(G, 30);
1268  B = av_clip_uintp2(B, 30);
1269  }
1270 
1271  switch(target) {
1272  case AV_PIX_FMT_ARGB:
1273  dest[0] = hasAlpha ? A : 255;
1274  dest[1] = R >> 22;
1275  dest[2] = G >> 22;
1276  dest[3] = B >> 22;
1277  break;
1278  case AV_PIX_FMT_RGB24:
1279  dest[0] = R >> 22;
1280  dest[1] = G >> 22;
1281  dest[2] = B >> 22;
1282  break;
1283  case AV_PIX_FMT_RGBA:
1284  dest[0] = R >> 22;
1285  dest[1] = G >> 22;
1286  dest[2] = B >> 22;
1287  dest[3] = hasAlpha ? A : 255;
1288  break;
1289  case AV_PIX_FMT_ABGR:
1290  dest[0] = hasAlpha ? A : 255;
1291  dest[1] = B >> 22;
1292  dest[2] = G >> 22;
1293  dest[3] = R >> 22;
1294  break;
1295  case AV_PIX_FMT_BGR24:
1296  dest[0] = B >> 22;
1297  dest[1] = G >> 22;
1298  dest[2] = R >> 22;
1299  break;
1300  case AV_PIX_FMT_BGRA:
1301  dest[0] = B >> 22;
1302  dest[1] = G >> 22;
1303  dest[2] = R >> 22;
1304  dest[3] = hasAlpha ? A : 255;
1305  break;
1306  case AV_PIX_FMT_BGR4_BYTE:
1307  case AV_PIX_FMT_RGB4_BYTE:
1308  case AV_PIX_FMT_BGR8:
1309  case AV_PIX_FMT_RGB8:
1310  {
1311  int r,g,b;
1312  R >>= 22;
1313  G >>= 22;
1314  B >>= 22;
1315  R += (7*err[0] + 1*c->dither_error[0][i] + 5*c->dither_error[0][i+1] + 3*c->dither_error[0][i+2])>>4;
1316  G += (7*err[1] + 1*c->dither_error[1][i] + 5*c->dither_error[1][i+1] + 3*c->dither_error[1][i+2])>>4;
1317  B += (7*err[2] + 1*c->dither_error[2][i] + 5*c->dither_error[2][i+1] + 3*c->dither_error[2][i+2])>>4;
1318  c->dither_error[0][i] = err[0];
1319  c->dither_error[1][i] = err[1];
1320  c->dither_error[2][i] = err[2];
1321  r = R >> (isrgb8 ? 5 : 7);
1322  g = G >> (isrgb8 ? 5 : 6);
1323  b = B >> (isrgb8 ? 6 : 7);
1324  r = av_clip(r, 0, isrgb8 ? 7 : 1);
1325  g = av_clip(g, 0, isrgb8 ? 7 : 3);
1326  b = av_clip(b, 0, isrgb8 ? 3 : 1);
1327  err[0] = R - r*(isrgb8 ? 36 : 255);
1328  err[1] = G - g*(isrgb8 ? 36 : 85);
1329  err[2] = B - b*(isrgb8 ? 85 : 255);
1330  if(target == AV_PIX_FMT_BGR4_BYTE) {
1331  dest[0] = r + 2*g + 8*b;
1332  } else if(target == AV_PIX_FMT_RGB4_BYTE) {
1333  dest[0] = b + 2*g + 8*r;
1334  } else if(target == AV_PIX_FMT_BGR8) {
1335  dest[0] = r + 8*g + 64*b;
1336  } else if(target == AV_PIX_FMT_RGB8) {
1337  dest[0] = b + 4*g + 32*r;
1338  } else
1339  av_assert2(0);
1340  step = 1;
1341  break;}
1342  }
1343  dest += step;
1344  }
1345  c->dither_error[0][i] = err[0];
1346  c->dither_error[1][i] = err[1];
1347  c->dither_error[2][i] = err[2];
1348 }
1349 
1350 #if CONFIG_SMALL
1351 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, AV_PIX_FMT_BGRA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1352 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, AV_PIX_FMT_ABGR, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1353 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, AV_PIX_FMT_RGBA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1354 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, AV_PIX_FMT_ARGB, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1355 #else
1356 #if CONFIG_SWSCALE_ALPHA
1357 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, AV_PIX_FMT_BGRA, 1)
1358 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, AV_PIX_FMT_ABGR, 1)
1359 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, AV_PIX_FMT_RGBA, 1)
1360 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, AV_PIX_FMT_ARGB, 1)
1361 #endif
1362 YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, AV_PIX_FMT_BGRA, 0)
1363 YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, AV_PIX_FMT_ABGR, 0)
1364 YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, AV_PIX_FMT_RGBA, 0)
1365 YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, AV_PIX_FMT_ARGB, 0)
1366 #endif
1367 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full, AV_PIX_FMT_BGR24, 0)
1368 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full, AV_PIX_FMT_RGB24, 0)
1369 
1370 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr4_byte_full, AV_PIX_FMT_BGR4_BYTE, 0)
1371 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb4_byte_full, AV_PIX_FMT_RGB4_BYTE, 0)
1372 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr8_full, AV_PIX_FMT_BGR8, 0)
1373 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb8_full, AV_PIX_FMT_RGB8, 0)
1374 
1375 static void
1376 yuv2gbrp_full_X_c(SwsContext *c, const int16_t *lumFilter,
1377  const int16_t **lumSrc, int lumFilterSize,
1378  const int16_t *chrFilter, const int16_t **chrUSrc,
1379  const int16_t **chrVSrc, int chrFilterSize,
1380  const int16_t **alpSrc, uint8_t **dest,
1381  int dstW, int y)
1382 {
1383  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->dstFormat);
1384  int i;
1385  int hasAlpha = 0;
1386  uint16_t **dest16 = (uint16_t**)dest;
1387  int SH = 22 + 7 - desc->comp[0].depth_minus1;
1388 
1389  for (i = 0; i < dstW; i++) {
1390  int j;
1391  int Y = 1 << 9;
1392  int U = (1 << 9) - (128 << 19);
1393  int V = (1 << 9) - (128 << 19);
1394  int R, G, B, A;
1395 
1396  for (j = 0; j < lumFilterSize; j++)
1397  Y += lumSrc[j][i] * lumFilter[j];
1398 
1399  for (j = 0; j < chrFilterSize; j++) {
1400  U += chrUSrc[j][i] * chrFilter[j];
1401  V += chrVSrc[j][i] * chrFilter[j];
1402  }
1403 
1404  Y >>= 10;
1405  U >>= 10;
1406  V >>= 10;
1407 
1408  if (hasAlpha) {
1409  A = 1 << 18;
1410 
1411  for (j = 0; j < lumFilterSize; j++)
1412  A += alpSrc[j][i] * lumFilter[j];
1413 
1414  A >>= 19;
1415 
1416  if (A & 0x100)
1417  A = av_clip_uint8(A);
1418  }
1419 
1420  Y -= c->yuv2rgb_y_offset;
1421  Y *= c->yuv2rgb_y_coeff;
1422  Y += 1 << 21;
1423  R = Y + V * c->yuv2rgb_v2r_coeff;
1424  G = Y + V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
1425  B = Y + U * c->yuv2rgb_u2b_coeff;
1426 
1427  if ((R | G | B) & 0xC0000000) {
1428  R = av_clip_uintp2(R, 30);
1429  G = av_clip_uintp2(G, 30);
1430  B = av_clip_uintp2(B, 30);
1431  }
1432 
1433  if (SH != 22) {
1434  dest16[0][i] = G >> SH;
1435  dest16[1][i] = B >> SH;
1436  dest16[2][i] = R >> SH;
1437  } else {
1438  dest[0][i] = G >> 22;
1439  dest[1][i] = B >> 22;
1440  dest[2][i] = R >> 22;
1441  }
1442  }
1443  if (SH != 22 && (!isBE(c->dstFormat)) != (!HAVE_BIGENDIAN)) {
1444  for (i = 0; i < dstW; i++) {
1445  dest16[0][i] = av_bswap16(dest16[0][i]);
1446  dest16[1][i] = av_bswap16(dest16[1][i]);
1447  dest16[2][i] = av_bswap16(dest16[2][i]);
1448  }
1449  }
1450 }
1451 
1452 av_cold void ff_sws_init_output_funcs(SwsContext *c,
1453  yuv2planar1_fn *yuv2plane1,
1454  yuv2planarX_fn *yuv2planeX,
1455  yuv2interleavedX_fn *yuv2nv12cX,
1456  yuv2packed1_fn *yuv2packed1,
1457  yuv2packed2_fn *yuv2packed2,
1458  yuv2packedX_fn *yuv2packedX,
1459  yuv2anyX_fn *yuv2anyX)
1460 {
1461  enum AVPixelFormat dstFormat = c->dstFormat;
1462  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);
1463 
1464  if (is16BPS(dstFormat)) {
1465  *yuv2planeX = isBE(dstFormat) ? yuv2planeX_16BE_c : yuv2planeX_16LE_c;
1466  *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_16BE_c : yuv2plane1_16LE_c;
1467  } else if (is9_OR_10BPS(dstFormat)) {
1468  if (desc->comp[0].depth_minus1 == 8) {
1469  *yuv2planeX = isBE(dstFormat) ? yuv2planeX_9BE_c : yuv2planeX_9LE_c;
1470  *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_9BE_c : yuv2plane1_9LE_c;
1471  } else if (desc->comp[0].depth_minus1 == 9) {
1472  *yuv2planeX = isBE(dstFormat) ? yuv2planeX_10BE_c : yuv2planeX_10LE_c;
1473  *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_10BE_c : yuv2plane1_10LE_c;
1474  } else if (desc->comp[0].depth_minus1 == 11) {
1475  *yuv2planeX = isBE(dstFormat) ? yuv2planeX_12BE_c : yuv2planeX_12LE_c;
1476  *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_12BE_c : yuv2plane1_12LE_c;
1477  } else if (desc->comp[0].depth_minus1 == 13) {
1478  *yuv2planeX = isBE(dstFormat) ? yuv2planeX_14BE_c : yuv2planeX_14LE_c;
1479  *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_14BE_c : yuv2plane1_14LE_c;
1480  } else
1481  av_assert0(0);
1482  } else {
1483  *yuv2plane1 = yuv2plane1_8_c;
1484  *yuv2planeX = yuv2planeX_8_c;
1485  if (dstFormat == AV_PIX_FMT_NV12 || dstFormat == AV_PIX_FMT_NV21)
1486  *yuv2nv12cX = yuv2nv12cX_c;
1487  }
1488 
1489  if(c->flags & SWS_FULL_CHR_H_INT) {
1490  switch (dstFormat) {
1491  case AV_PIX_FMT_RGBA:
1492 #if CONFIG_SMALL
1493  *yuv2packedX = yuv2rgba32_full_X_c;
1494 #else
1495 #if CONFIG_SWSCALE_ALPHA
1496  if (c->alpPixBuf) {
1497  *yuv2packedX = yuv2rgba32_full_X_c;
1498  } else
1499 #endif /* CONFIG_SWSCALE_ALPHA */
1500  {
1501  *yuv2packedX = yuv2rgbx32_full_X_c;
1502  }
1503 #endif /* !CONFIG_SMALL */
1504  break;
1505  case AV_PIX_FMT_ARGB:
1506 #if CONFIG_SMALL
1507  *yuv2packedX = yuv2argb32_full_X_c;
1508 #else
1509 #if CONFIG_SWSCALE_ALPHA
1510  if (c->alpPixBuf) {
1511  *yuv2packedX = yuv2argb32_full_X_c;
1512  } else
1513 #endif /* CONFIG_SWSCALE_ALPHA */
1514  {
1515  *yuv2packedX = yuv2xrgb32_full_X_c;
1516  }
1517 #endif /* !CONFIG_SMALL */
1518  break;
1519  case AV_PIX_FMT_BGRA:
1520 #if CONFIG_SMALL
1521  *yuv2packedX = yuv2bgra32_full_X_c;
1522 #else
1523 #if CONFIG_SWSCALE_ALPHA
1524  if (c->alpPixBuf) {
1525  *yuv2packedX = yuv2bgra32_full_X_c;
1526  } else
1527 #endif /* CONFIG_SWSCALE_ALPHA */
1528  {
1529  *yuv2packedX = yuv2bgrx32_full_X_c;
1530  }
1531 #endif /* !CONFIG_SMALL */
1532  break;
1533  case AV_PIX_FMT_ABGR:
1534 #if CONFIG_SMALL
1535  *yuv2packedX = yuv2abgr32_full_X_c;
1536 #else
1537 #if CONFIG_SWSCALE_ALPHA
1538  if (c->alpPixBuf) {
1539  *yuv2packedX = yuv2abgr32_full_X_c;
1540  } else
1541 #endif /* CONFIG_SWSCALE_ALPHA */
1542  {
1543  *yuv2packedX = yuv2xbgr32_full_X_c;
1544  }
1545 #endif /* !CONFIG_SMALL */
1546  break;
1547  case AV_PIX_FMT_RGB24:
1548  *yuv2packedX = yuv2rgb24_full_X_c;
1549  break;
1550  case AV_PIX_FMT_BGR24:
1551  *yuv2packedX = yuv2bgr24_full_X_c;
1552  break;
1553  case AV_PIX_FMT_BGR4_BYTE:
1554  *yuv2packedX = yuv2bgr4_byte_full_X_c;
1555  break;
1556  case AV_PIX_FMT_RGB4_BYTE:
1557  *yuv2packedX = yuv2rgb4_byte_full_X_c;
1558  break;
1559  case AV_PIX_FMT_BGR8:
1560  *yuv2packedX = yuv2bgr8_full_X_c;
1561  break;
1562  case AV_PIX_FMT_RGB8:
1563  *yuv2packedX = yuv2rgb8_full_X_c;
1564  break;
1565  case AV_PIX_FMT_GBRP:
1566  case AV_PIX_FMT_GBRP9BE:
1567  case AV_PIX_FMT_GBRP9LE:
1568  case AV_PIX_FMT_GBRP10BE:
1569  case AV_PIX_FMT_GBRP10LE:
1570  case AV_PIX_FMT_GBRP12BE:
1571  case AV_PIX_FMT_GBRP12LE:
1572  case AV_PIX_FMT_GBRP14BE:
1573  case AV_PIX_FMT_GBRP14LE:
1574  case AV_PIX_FMT_GBRP16BE:
1575  case AV_PIX_FMT_GBRP16LE:
1576  *yuv2anyX = yuv2gbrp_full_X_c;
1577  break;
1578  }
1579  if (!*yuv2packedX && !*yuv2anyX)
1580  goto YUV_PACKED;
1581  } else {
1582  YUV_PACKED:
1583  switch (dstFormat) {
1584  case AV_PIX_FMT_RGB48LE:
1585  *yuv2packed1 = yuv2rgb48le_1_c;
1586  *yuv2packed2 = yuv2rgb48le_2_c;
1587  *yuv2packedX = yuv2rgb48le_X_c;
1588  break;
1589  case AV_PIX_FMT_RGB48BE:
1590  *yuv2packed1 = yuv2rgb48be_1_c;
1591  *yuv2packed2 = yuv2rgb48be_2_c;
1592  *yuv2packedX = yuv2rgb48be_X_c;
1593  break;
1594  case AV_PIX_FMT_BGR48LE:
1595  *yuv2packed1 = yuv2bgr48le_1_c;
1596  *yuv2packed2 = yuv2bgr48le_2_c;
1597  *yuv2packedX = yuv2bgr48le_X_c;
1598  break;
1599  case AV_PIX_FMT_BGR48BE:
1600  *yuv2packed1 = yuv2bgr48be_1_c;
1601  *yuv2packed2 = yuv2bgr48be_2_c;
1602  *yuv2packedX = yuv2bgr48be_X_c;
1603  break;
1604  case AV_PIX_FMT_RGB32:
1605  case AV_PIX_FMT_BGR32:
1606 #if CONFIG_SMALL
1607  *yuv2packed1 = yuv2rgb32_1_c;
1608  *yuv2packed2 = yuv2rgb32_2_c;
1609  *yuv2packedX = yuv2rgb32_X_c;
1610 #else
1611 #if CONFIG_SWSCALE_ALPHA
1612  if (c->alpPixBuf) {
1613  *yuv2packed1 = yuv2rgba32_1_c;
1614  *yuv2packed2 = yuv2rgba32_2_c;
1615  *yuv2packedX = yuv2rgba32_X_c;
1616  } else
1617 #endif /* CONFIG_SWSCALE_ALPHA */
1618  {
1619  *yuv2packed1 = yuv2rgbx32_1_c;
1620  *yuv2packed2 = yuv2rgbx32_2_c;
1621  *yuv2packedX = yuv2rgbx32_X_c;
1622  }
1623 #endif /* !CONFIG_SMALL */
1624  break;
1625  case AV_PIX_FMT_RGB32_1:
1626  case AV_PIX_FMT_BGR32_1:
1627 #if CONFIG_SMALL
1628  *yuv2packed1 = yuv2rgb32_1_1_c;
1629  *yuv2packed2 = yuv2rgb32_1_2_c;
1630  *yuv2packedX = yuv2rgb32_1_X_c;
1631 #else
1632 #if CONFIG_SWSCALE_ALPHA
1633  if (c->alpPixBuf) {
1634  *yuv2packed1 = yuv2rgba32_1_1_c;
1635  *yuv2packed2 = yuv2rgba32_1_2_c;
1636  *yuv2packedX = yuv2rgba32_1_X_c;
1637  } else
1638 #endif /* CONFIG_SWSCALE_ALPHA */
1639  {
1640  *yuv2packed1 = yuv2rgbx32_1_1_c;
1641  *yuv2packed2 = yuv2rgbx32_1_2_c;
1642  *yuv2packedX = yuv2rgbx32_1_X_c;
1643  }
1644 #endif /* !CONFIG_SMALL */
1645  break;
1646  case AV_PIX_FMT_RGB24:
1647  *yuv2packed1 = yuv2rgb24_1_c;
1648  *yuv2packed2 = yuv2rgb24_2_c;
1649  *yuv2packedX = yuv2rgb24_X_c;
1650  break;
1651  case AV_PIX_FMT_BGR24:
1652  *yuv2packed1 = yuv2bgr24_1_c;
1653  *yuv2packed2 = yuv2bgr24_2_c;
1654  *yuv2packedX = yuv2bgr24_X_c;
1655  break;
1656  case AV_PIX_FMT_RGB565LE:
1657  case AV_PIX_FMT_RGB565BE:
1658  case AV_PIX_FMT_BGR565LE:
1659  case AV_PIX_FMT_BGR565BE:
1660  *yuv2packed1 = yuv2rgb16_1_c;
1661  *yuv2packed2 = yuv2rgb16_2_c;
1662  *yuv2packedX = yuv2rgb16_X_c;
1663  break;
1664  case AV_PIX_FMT_RGB555LE:
1665  case AV_PIX_FMT_RGB555BE:
1666  case AV_PIX_FMT_BGR555LE:
1667  case AV_PIX_FMT_BGR555BE:
1668  *yuv2packed1 = yuv2rgb15_1_c;
1669  *yuv2packed2 = yuv2rgb15_2_c;
1670  *yuv2packedX = yuv2rgb15_X_c;
1671  break;
1672  case AV_PIX_FMT_RGB444LE:
1673  case AV_PIX_FMT_RGB444BE:
1674  case AV_PIX_FMT_BGR444LE:
1675  case AV_PIX_FMT_BGR444BE:
1676  *yuv2packed1 = yuv2rgb12_1_c;
1677  *yuv2packed2 = yuv2rgb12_2_c;
1678  *yuv2packedX = yuv2rgb12_X_c;
1679  break;
1680  case AV_PIX_FMT_RGB8:
1681  case AV_PIX_FMT_BGR8:
1682  *yuv2packed1 = yuv2rgb8_1_c;
1683  *yuv2packed2 = yuv2rgb8_2_c;
1684  *yuv2packedX = yuv2rgb8_X_c;
1685  break;
1686  case AV_PIX_FMT_RGB4:
1687  case AV_PIX_FMT_BGR4:
1688  *yuv2packed1 = yuv2rgb4_1_c;
1689  *yuv2packed2 = yuv2rgb4_2_c;
1690  *yuv2packedX = yuv2rgb4_X_c;
1691  break;
1692  case AV_PIX_FMT_RGB4_BYTE:
1693  case AV_PIX_FMT_BGR4_BYTE:
1694  *yuv2packed1 = yuv2rgb4b_1_c;
1695  *yuv2packed2 = yuv2rgb4b_2_c;
1696  *yuv2packedX = yuv2rgb4b_X_c;
1697  break;
1698  }
1699  }
1700  switch (dstFormat) {
1701  case AV_PIX_FMT_MONOWHITE:
1702  *yuv2packed1 = yuv2monowhite_1_c;
1703  *yuv2packed2 = yuv2monowhite_2_c;
1704  *yuv2packedX = yuv2monowhite_X_c;
1705  break;
1706  case AV_PIX_FMT_MONOBLACK:
1707  *yuv2packed1 = yuv2monoblack_1_c;
1708  *yuv2packed2 = yuv2monoblack_2_c;
1709  *yuv2packedX = yuv2monoblack_X_c;
1710  break;
1711  case AV_PIX_FMT_YUYV422:
1712  *yuv2packed1 = yuv2yuyv422_1_c;
1713  *yuv2packed2 = yuv2yuyv422_2_c;
1714  *yuv2packedX = yuv2yuyv422_X_c;
1715  break;
1716  case AV_PIX_FMT_UYVY422:
1717  *yuv2packed1 = yuv2uyvy422_1_c;
1718  *yuv2packed2 = yuv2uyvy422_2_c;
1719  *yuv2packedX = yuv2uyvy422_X_c;
1720  break;
1721  }
1722 }
Generated on Sat May 25 2013 04:01:21 for FFmpeg by   doxygen 1.8.2