2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of Libav.
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 #include "libavutil/attributes.h"
28 #include "libavutil/avutil.h"
29 #include "libavutil/bswap.h"
30 #include "libavutil/cpu.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavutil/mathematics.h"
33 #include "libavutil/pixdesc.h"
37 #include "swscale_internal.h"
39 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
40 { 1, 3, 1, 3, 1, 3, 1, 3, },
41 { 2, 0, 2, 0, 2, 0, 2, 0, },
44 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
45 { 6, 2, 6, 2, 6, 2, 6, 2, },
46 { 0, 4, 0, 4, 0, 4, 0, 4, },
49 DECLARE_ALIGNED(8, const uint8_t, ff_dither_4x4_16)[4][8] = {
50 { 8, 4, 11, 7, 8, 4, 11, 7, },
51 { 2, 14, 1, 13, 2, 14, 1, 13, },
52 { 10, 6, 9, 5, 10, 6, 9, 5, },
53 { 0, 12, 3, 15, 0, 12, 3, 15, },
56 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_32)[8][8] = {
57 { 17, 9, 23, 15, 16, 8, 22, 14, },
58 { 5, 29, 3, 27, 4, 28, 2, 26, },
59 { 21, 13, 19, 11, 20, 12, 18, 10, },
60 { 0, 24, 6, 30, 1, 25, 7, 31, },
61 { 16, 8, 22, 14, 17, 9, 23, 15, },
62 { 4, 28, 2, 26, 5, 29, 3, 27, },
63 { 20, 12, 18, 10, 21, 13, 19, 11, },
64 { 1, 25, 7, 31, 0, 24, 6, 30, },
67 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_73)[8][8] = {
68 { 0, 55, 14, 68, 3, 58, 17, 72, },
69 { 37, 18, 50, 32, 40, 22, 54, 35, },
70 { 9, 64, 5, 59, 13, 67, 8, 63, },
71 { 46, 27, 41, 23, 49, 31, 44, 26, },
72 { 2, 57, 16, 71, 1, 56, 15, 70, },
73 { 39, 21, 52, 34, 38, 19, 51, 33, },
74 { 11, 66, 7, 62, 10, 65, 6, 60, },
75 { 48, 30, 43, 25, 47, 29, 42, 24, },
79 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_220)[8][8] = {
80 {117, 62, 158, 103, 113, 58, 155, 100, },
81 { 34, 199, 21, 186, 31, 196, 17, 182, },
82 {144, 89, 131, 76, 141, 86, 127, 72, },
83 { 0, 165, 41, 206, 10, 175, 52, 217, },
84 {110, 55, 151, 96, 120, 65, 162, 107, },
85 { 28, 193, 14, 179, 38, 203, 24, 189, },
86 {138, 83, 124, 69, 148, 93, 134, 79, },
87 { 7, 172, 48, 213, 3, 168, 45, 210, },
90 // tries to correct a gamma of 1.5
91 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_220)[8][8] = {
92 { 0, 143, 18, 200, 2, 156, 25, 215, },
93 { 78, 28, 125, 64, 89, 36, 138, 74, },
94 { 10, 180, 3, 161, 16, 195, 8, 175, },
95 {109, 51, 93, 38, 121, 60, 105, 47, },
96 { 1, 152, 23, 210, 0, 147, 20, 205, },
97 { 85, 33, 134, 71, 81, 30, 130, 67, },
98 { 14, 190, 6, 171, 12, 185, 5, 166, },
99 {117, 57, 101, 44, 113, 54, 97, 41, },
102 // tries to correct a gamma of 2.0
103 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_220)[8][8] = {
104 { 0, 124, 8, 193, 0, 140, 12, 213, },
105 { 55, 14, 104, 42, 66, 19, 119, 52, },
106 { 3, 168, 1, 145, 6, 187, 3, 162, },
107 { 86, 31, 70, 21, 99, 39, 82, 28, },
108 { 0, 134, 11, 206, 0, 129, 9, 200, },
109 { 62, 17, 114, 48, 58, 16, 109, 45, },
110 { 5, 181, 2, 157, 4, 175, 1, 151, },
111 { 95, 36, 78, 26, 90, 34, 74, 24, },
114 // tries to correct a gamma of 2.5
115 DECLARE_ALIGNED(8, const uint8_t, ff_dither_8x8_220)[8][8] = {
116 { 0, 107, 3, 187, 0, 125, 6, 212, },
117 { 39, 7, 86, 28, 49, 11, 102, 36, },
118 { 1, 158, 0, 131, 3, 180, 1, 151, },
119 { 68, 19, 52, 12, 81, 25, 64, 17, },
120 { 0, 119, 5, 203, 0, 113, 4, 195, },
121 { 45, 9, 96, 33, 42, 8, 91, 30, },
122 { 2, 172, 1, 144, 2, 165, 0, 137, },
123 { 77, 23, 60, 15, 72, 21, 56, 14, },
127 #define output_pixel(pos, val, bias, signedness) \
129 AV_WB16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
131 AV_WL16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
134 static av_always_inline void
135 yuv2plane1_16_c_template(const int32_t *src, uint16_t *dest, int dstW,
136 int big_endian, int output_bits)
139 int shift = 19 - output_bits;
141 for (i = 0; i < dstW; i++) {
142 int val = src[i] + (1 << (shift - 1));
143 output_pixel(&dest[i], val, 0, uint);
147 static av_always_inline void
148 yuv2planeX_16_c_template(const int16_t *filter, int filterSize,
149 const int32_t **src, uint16_t *dest, int dstW,
150 int big_endian, int output_bits)
153 int shift = 15 + 16 - output_bits;
155 for (i = 0; i < dstW; i++) {
156 int val = 1 << (30-output_bits);
159 /* range of val is [0,0x7FFFFFFF], so 31 bits, but with lanczos/spline
160 * filters (or anything with negative coeffs, the range can be slightly
161 * wider in both directions. To account for this overflow, we subtract
162 * a constant so it always fits in the signed range (assuming a
163 * reasonable filterSize), and re-add that at the end. */
165 for (j = 0; j < filterSize; j++)
166 val += src[j][i] * filter[j];
168 output_pixel(&dest[i], val, 0x8000, int);
174 #define output_pixel(pos, val) \
176 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
178 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
181 static av_always_inline void
182 yuv2plane1_10_c_template(const int16_t *src, uint16_t *dest, int dstW,
183 int big_endian, int output_bits)
186 int shift = 15 - output_bits;
188 for (i = 0; i < dstW; i++) {
189 int val = src[i] + (1 << (shift - 1));
190 output_pixel(&dest[i], val);
194 static av_always_inline void
195 yuv2planeX_10_c_template(const int16_t *filter, int filterSize,
196 const int16_t **src, uint16_t *dest, int dstW,
197 int big_endian, int output_bits)
200 int shift = 11 + 16 - output_bits;
202 for (i = 0; i < dstW; i++) {
203 int val = 1 << (26-output_bits);
206 for (j = 0; j < filterSize; j++)
207 val += src[j][i] * filter[j];
209 output_pixel(&dest[i], val);
215 #define yuv2NBPS(bits, BE_LE, is_be, template_size, typeX_t) \
216 static void yuv2plane1_ ## bits ## BE_LE ## _c(const int16_t *src, \
217 uint8_t *dest, int dstW, \
218 const uint8_t *dither, int offset)\
220 yuv2plane1_ ## template_size ## _c_template((const typeX_t *) src, \
221 (uint16_t *) dest, dstW, is_be, bits); \
223 static void yuv2planeX_ ## bits ## BE_LE ## _c(const int16_t *filter, int filterSize, \
224 const int16_t **src, uint8_t *dest, int dstW, \
225 const uint8_t *dither, int offset)\
227 yuv2planeX_## template_size ## _c_template(filter, \
228 filterSize, (const typeX_t **) src, \
229 (uint16_t *) dest, dstW, is_be, bits); \
231 yuv2NBPS( 9, BE, 1, 10, int16_t)
232 yuv2NBPS( 9, LE, 0, 10, int16_t)
233 yuv2NBPS(10, BE, 1, 10, int16_t)
234 yuv2NBPS(10, LE, 0, 10, int16_t)
235 yuv2NBPS(16, BE, 1, 16, int32_t)
236 yuv2NBPS(16, LE, 0, 16, int32_t)
238 static void yuv2planeX_8_c(const int16_t *filter, int filterSize,
239 const int16_t **src, uint8_t *dest, int dstW,
240 const uint8_t *dither, int offset)
243 for (i=0; i<dstW; i++) {
244 int val = dither[(i + offset) & 7] << 12;
246 for (j=0; j<filterSize; j++)
247 val += src[j][i] * filter[j];
249 dest[i]= av_clip_uint8(val>>19);
253 static void yuv2plane1_8_c(const int16_t *src, uint8_t *dest, int dstW,
254 const uint8_t *dither, int offset)
257 for (i=0; i<dstW; i++) {
258 int val = (src[i] + dither[(i + offset) & 7]) >> 7;
259 dest[i]= av_clip_uint8(val);
263 static void yuv2nv12cX_c(SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
264 const int16_t **chrUSrc, const int16_t **chrVSrc,
265 uint8_t *dest, int chrDstW)
267 enum AVPixelFormat dstFormat = c->dstFormat;
268 const uint8_t *chrDither = c->chrDither8;
271 if (dstFormat == AV_PIX_FMT_NV12)
272 for (i=0; i<chrDstW; i++) {
273 int u = chrDither[i & 7] << 12;
274 int v = chrDither[(i + 3) & 7] << 12;
276 for (j=0; j<chrFilterSize; j++) {
277 u += chrUSrc[j][i] * chrFilter[j];
278 v += chrVSrc[j][i] * chrFilter[j];
281 dest[2*i]= av_clip_uint8(u>>19);
282 dest[2*i+1]= av_clip_uint8(v>>19);
285 for (i=0; i<chrDstW; i++) {
286 int u = chrDither[i & 7] << 12;
287 int v = chrDither[(i + 3) & 7] << 12;
289 for (j=0; j<chrFilterSize; j++) {
290 u += chrUSrc[j][i] * chrFilter[j];
291 v += chrVSrc[j][i] * chrFilter[j];
294 dest[2*i]= av_clip_uint8(v>>19);
295 dest[2*i+1]= av_clip_uint8(u>>19);
299 #define accumulate_bit(acc, val) \
301 acc |= (val) >= (128 + 110)
302 #define output_pixel(pos, acc) \
303 if (target == AV_PIX_FMT_MONOBLACK) { \
309 static av_always_inline void
310 yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
311 const int16_t **lumSrc, int lumFilterSize,
312 const int16_t *chrFilter, const int16_t **chrUSrc,
313 const int16_t **chrVSrc, int chrFilterSize,
314 const int16_t **alpSrc, uint8_t *dest, int dstW,
315 int y, enum AVPixelFormat target)
317 const uint8_t * const d128 = ff_dither_8x8_220[y&7];
321 for (i = 0; i < dstW; i += 2) {
326 for (j = 0; j < lumFilterSize; j++) {
327 Y1 += lumSrc[j][i] * lumFilter[j];
328 Y2 += lumSrc[j][i+1] * lumFilter[j];
332 if ((Y1 | Y2) & 0x100) {
333 Y1 = av_clip_uint8(Y1);
334 Y2 = av_clip_uint8(Y2);
336 accumulate_bit(acc, Y1 + d128[(i + 0) & 7]);
337 accumulate_bit(acc, Y2 + d128[(i + 1) & 7]);
339 output_pixel(*dest++, acc);
344 output_pixel(*dest, acc);
348 static av_always_inline void
349 yuv2mono_2_c_template(SwsContext *c, const int16_t *buf[2],
350 const int16_t *ubuf[2], const int16_t *vbuf[2],
351 const int16_t *abuf[2], uint8_t *dest, int dstW,
352 int yalpha, int uvalpha, int y,
353 enum AVPixelFormat target)
355 const int16_t *buf0 = buf[0], *buf1 = buf[1];
356 const uint8_t * const d128 = ff_dither_8x8_220[y & 7];
357 int yalpha1 = 4096 - yalpha;
360 for (i = 0; i < dstW; i += 8) {
363 Y = (buf0[i + 0] * yalpha1 + buf1[i + 0] * yalpha) >> 19;
364 accumulate_bit(acc, Y + d128[0]);
365 Y = (buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19;
366 accumulate_bit(acc, Y + d128[1]);
367 Y = (buf0[i + 2] * yalpha1 + buf1[i + 2] * yalpha) >> 19;
368 accumulate_bit(acc, Y + d128[2]);
369 Y = (buf0[i + 3] * yalpha1 + buf1[i + 3] * yalpha) >> 19;
370 accumulate_bit(acc, Y + d128[3]);
371 Y = (buf0[i + 4] * yalpha1 + buf1[i + 4] * yalpha) >> 19;
372 accumulate_bit(acc, Y + d128[4]);
373 Y = (buf0[i + 5] * yalpha1 + buf1[i + 5] * yalpha) >> 19;
374 accumulate_bit(acc, Y + d128[5]);
375 Y = (buf0[i + 6] * yalpha1 + buf1[i + 6] * yalpha) >> 19;
376 accumulate_bit(acc, Y + d128[6]);
377 Y = (buf0[i + 7] * yalpha1 + buf1[i + 7] * yalpha) >> 19;
378 accumulate_bit(acc, Y + d128[7]);
380 output_pixel(*dest++, acc);
384 static av_always_inline void
385 yuv2mono_1_c_template(SwsContext *c, const int16_t *buf0,
386 const int16_t *ubuf[2], const int16_t *vbuf[2],
387 const int16_t *abuf0, uint8_t *dest, int dstW,
388 int uvalpha, int y, enum AVPixelFormat target)
390 const uint8_t * const d128 = ff_dither_8x8_220[y & 7];
393 for (i = 0; i < dstW; i += 8) {
396 accumulate_bit(acc, (buf0[i + 0] >> 7) + d128[0]);
397 accumulate_bit(acc, (buf0[i + 1] >> 7) + d128[1]);
398 accumulate_bit(acc, (buf0[i + 2] >> 7) + d128[2]);
399 accumulate_bit(acc, (buf0[i + 3] >> 7) + d128[3]);
400 accumulate_bit(acc, (buf0[i + 4] >> 7) + d128[4]);
401 accumulate_bit(acc, (buf0[i + 5] >> 7) + d128[5]);
402 accumulate_bit(acc, (buf0[i + 6] >> 7) + d128[6]);
403 accumulate_bit(acc, (buf0[i + 7] >> 7) + d128[7]);
405 output_pixel(*dest++, acc);
410 #undef accumulate_bit
412 #define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
413 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
414 const int16_t **lumSrc, int lumFilterSize, \
415 const int16_t *chrFilter, const int16_t **chrUSrc, \
416 const int16_t **chrVSrc, int chrFilterSize, \
417 const int16_t **alpSrc, uint8_t *dest, int dstW, \
420 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
421 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
422 alpSrc, dest, dstW, y, fmt); \
425 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
426 const int16_t *ubuf[2], const int16_t *vbuf[2], \
427 const int16_t *abuf[2], uint8_t *dest, int dstW, \
428 int yalpha, int uvalpha, int y) \
430 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
431 dest, dstW, yalpha, uvalpha, y, fmt); \
434 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
435 const int16_t *ubuf[2], const int16_t *vbuf[2], \
436 const int16_t *abuf0, uint8_t *dest, int dstW, \
437 int uvalpha, int y) \
439 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, \
440 abuf0, dest, dstW, uvalpha, \
444 YUV2PACKEDWRAPPER(yuv2mono,, white, AV_PIX_FMT_MONOWHITE)
445 YUV2PACKEDWRAPPER(yuv2mono,, black, AV_PIX_FMT_MONOBLACK)
447 #define output_pixels(pos, Y1, U, Y2, V) \
448 if (target == AV_PIX_FMT_YUYV422) { \
449 dest[pos + 0] = Y1; \
451 dest[pos + 2] = Y2; \
453 } else if (target == AV_PIX_FMT_YVYU422) { \
454 dest[pos + 0] = Y1; \
456 dest[pos + 2] = Y2; \
458 } else { /* AV_PIX_FMT_UYVY422 */ \
460 dest[pos + 1] = Y1; \
462 dest[pos + 3] = Y2; \
465 static av_always_inline void
466 yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
467 const int16_t **lumSrc, int lumFilterSize,
468 const int16_t *chrFilter, const int16_t **chrUSrc,
469 const int16_t **chrVSrc, int chrFilterSize,
470 const int16_t **alpSrc, uint8_t *dest, int dstW,
471 int y, enum AVPixelFormat target)
475 for (i = 0; i < ((dstW + 1) >> 1); i++) {
482 for (j = 0; j < lumFilterSize; j++) {
483 Y1 += lumSrc[j][i * 2] * lumFilter[j];
484 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
486 for (j = 0; j < chrFilterSize; j++) {
487 U += chrUSrc[j][i] * chrFilter[j];
488 V += chrVSrc[j][i] * chrFilter[j];
494 if ((Y1 | Y2 | U | V) & 0x100) {
495 Y1 = av_clip_uint8(Y1);
496 Y2 = av_clip_uint8(Y2);
497 U = av_clip_uint8(U);
498 V = av_clip_uint8(V);
500 output_pixels(4*i, Y1, U, Y2, V);
504 static av_always_inline void
505 yuv2422_2_c_template(SwsContext *c, const int16_t *buf[2],
506 const int16_t *ubuf[2], const int16_t *vbuf[2],
507 const int16_t *abuf[2], uint8_t *dest, int dstW,
508 int yalpha, int uvalpha, int y,
509 enum AVPixelFormat target)
511 const int16_t *buf0 = buf[0], *buf1 = buf[1],
512 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
513 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
514 int yalpha1 = 4096 - yalpha;
515 int uvalpha1 = 4096 - uvalpha;
518 for (i = 0; i < ((dstW + 1) >> 1); i++) {
519 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
520 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
521 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
522 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
524 Y1 = av_clip_uint8(Y1);
525 Y2 = av_clip_uint8(Y2);
526 U = av_clip_uint8(U);
527 V = av_clip_uint8(V);
529 output_pixels(i * 4, Y1, U, Y2, V);
533 static av_always_inline void
534 yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
535 const int16_t *ubuf[2], const int16_t *vbuf[2],
536 const int16_t *abuf0, uint8_t *dest, int dstW,
537 int uvalpha, int y, enum AVPixelFormat target)
539 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
542 if (uvalpha < 2048) {
543 for (i = 0; i < ((dstW + 1) >> 1); i++) {
544 int Y1 = buf0[i * 2] >> 7;
545 int Y2 = buf0[i * 2 + 1] >> 7;
546 int U = ubuf0[i] >> 7;
547 int V = vbuf0[i] >> 7;
549 Y1 = av_clip_uint8(Y1);
550 Y2 = av_clip_uint8(Y2);
551 U = av_clip_uint8(U);
552 V = av_clip_uint8(V);
554 output_pixels(i * 4, Y1, U, Y2, V);
557 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
558 for (i = 0; i < ((dstW + 1) >> 1); i++) {
559 int Y1 = buf0[i * 2] >> 7;
560 int Y2 = buf0[i * 2 + 1] >> 7;
561 int U = (ubuf0[i] + ubuf1[i]) >> 8;
562 int V = (vbuf0[i] + vbuf1[i]) >> 8;
564 Y1 = av_clip_uint8(Y1);
565 Y2 = av_clip_uint8(Y2);
566 U = av_clip_uint8(U);
567 V = av_clip_uint8(V);
569 output_pixels(i * 4, Y1, U, Y2, V);
576 YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, AV_PIX_FMT_YUYV422)
577 YUV2PACKEDWRAPPER(yuv2, 422, yvyu422, AV_PIX_FMT_YVYU422)
578 YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, AV_PIX_FMT_UYVY422)
580 #define R_B ((target == AV_PIX_FMT_RGB48LE || target == AV_PIX_FMT_RGB48BE) ? R : B)
581 #define B_R ((target == AV_PIX_FMT_RGB48LE || target == AV_PIX_FMT_RGB48BE) ? B : R)
582 #define output_pixel(pos, val) \
583 if (isBE(target)) { \
589 static av_always_inline void
590 yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
591 const int32_t **lumSrc, int lumFilterSize,
592 const int16_t *chrFilter, const int32_t **chrUSrc,
593 const int32_t **chrVSrc, int chrFilterSize,
594 const int32_t **alpSrc, uint16_t *dest, int dstW,
595 int y, enum AVPixelFormat target)
599 for (i = 0; i < ((dstW + 1) >> 1); i++) {
601 int Y1 = -0x40000000;
602 int Y2 = -0x40000000;
603 int U = -128 << 23; // 19
607 for (j = 0; j < lumFilterSize; j++) {
608 Y1 += lumSrc[j][i * 2] * lumFilter[j];
609 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
611 for (j = 0; j < chrFilterSize; j++) {
612 U += chrUSrc[j][i] * chrFilter[j];
613 V += chrVSrc[j][i] * chrFilter[j];
616 // 8bit: 12+15=27; 16-bit: 12+19=31
624 // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
625 Y1 -= c->yuv2rgb_y_offset;
626 Y2 -= c->yuv2rgb_y_offset;
627 Y1 *= c->yuv2rgb_y_coeff;
628 Y2 *= c->yuv2rgb_y_coeff;
631 // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
633 R = V * c->yuv2rgb_v2r_coeff;
634 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
635 B = U * c->yuv2rgb_u2b_coeff;
637 // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
638 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
639 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
640 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
641 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
642 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
643 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
648 static av_always_inline void
649 yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
650 const int32_t *ubuf[2], const int32_t *vbuf[2],
651 const int32_t *abuf[2], uint16_t *dest, int dstW,
652 int yalpha, int uvalpha, int y,
653 enum AVPixelFormat target)
655 const int32_t *buf0 = buf[0], *buf1 = buf[1],
656 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
657 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
658 int yalpha1 = 4096 - yalpha;
659 int uvalpha1 = 4096 - uvalpha;
662 for (i = 0; i < ((dstW + 1) >> 1); i++) {
663 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 14;
664 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 14;
665 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha + (-128 << 23)) >> 14;
666 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha + (-128 << 23)) >> 14;
669 Y1 -= c->yuv2rgb_y_offset;
670 Y2 -= c->yuv2rgb_y_offset;
671 Y1 *= c->yuv2rgb_y_coeff;
672 Y2 *= c->yuv2rgb_y_coeff;
676 R = V * c->yuv2rgb_v2r_coeff;
677 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
678 B = U * c->yuv2rgb_u2b_coeff;
680 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
681 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
682 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
683 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
684 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
685 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
690 static av_always_inline void
691 yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
692 const int32_t *ubuf[2], const int32_t *vbuf[2],
693 const int32_t *abuf0, uint16_t *dest, int dstW,
694 int uvalpha, int y, enum AVPixelFormat target)
696 const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
699 if (uvalpha < 2048) {
700 for (i = 0; i < ((dstW + 1) >> 1); i++) {
701 int Y1 = (buf0[i * 2] ) >> 2;
702 int Y2 = (buf0[i * 2 + 1]) >> 2;
703 int U = (ubuf0[i] + (-128 << 11)) >> 2;
704 int V = (vbuf0[i] + (-128 << 11)) >> 2;
707 Y1 -= c->yuv2rgb_y_offset;
708 Y2 -= c->yuv2rgb_y_offset;
709 Y1 *= c->yuv2rgb_y_coeff;
710 Y2 *= c->yuv2rgb_y_coeff;
714 R = V * c->yuv2rgb_v2r_coeff;
715 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
716 B = U * c->yuv2rgb_u2b_coeff;
718 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
719 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
720 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
721 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
722 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
723 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
727 const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
728 for (i = 0; i < ((dstW + 1) >> 1); i++) {
729 int Y1 = (buf0[i * 2] ) >> 2;
730 int Y2 = (buf0[i * 2 + 1]) >> 2;
731 int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
732 int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
735 Y1 -= c->yuv2rgb_y_offset;
736 Y2 -= c->yuv2rgb_y_offset;
737 Y1 *= c->yuv2rgb_y_coeff;
738 Y2 *= c->yuv2rgb_y_coeff;
742 R = V * c->yuv2rgb_v2r_coeff;
743 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
744 B = U * c->yuv2rgb_u2b_coeff;
746 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
747 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
748 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
749 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
750 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
751 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
761 #define YUV2PACKED16WRAPPER(name, base, ext, fmt) \
762 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
763 const int16_t **_lumSrc, int lumFilterSize, \
764 const int16_t *chrFilter, const int16_t **_chrUSrc, \
765 const int16_t **_chrVSrc, int chrFilterSize, \
766 const int16_t **_alpSrc, uint8_t *_dest, int dstW, \
769 const int32_t **lumSrc = (const int32_t **) _lumSrc, \
770 **chrUSrc = (const int32_t **) _chrUSrc, \
771 **chrVSrc = (const int32_t **) _chrVSrc, \
772 **alpSrc = (const int32_t **) _alpSrc; \
773 uint16_t *dest = (uint16_t *) _dest; \
774 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
775 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
776 alpSrc, dest, dstW, y, fmt); \
779 static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
780 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
781 const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
782 int yalpha, int uvalpha, int y) \
784 const int32_t **buf = (const int32_t **) _buf, \
785 **ubuf = (const int32_t **) _ubuf, \
786 **vbuf = (const int32_t **) _vbuf, \
787 **abuf = (const int32_t **) _abuf; \
788 uint16_t *dest = (uint16_t *) _dest; \
789 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
790 dest, dstW, yalpha, uvalpha, y, fmt); \
793 static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
794 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
795 const int16_t *_abuf0, uint8_t *_dest, int dstW, \
796 int uvalpha, int y) \
798 const int32_t *buf0 = (const int32_t *) _buf0, \
799 **ubuf = (const int32_t **) _ubuf, \
800 **vbuf = (const int32_t **) _vbuf, \
801 *abuf0 = (const int32_t *) _abuf0; \
802 uint16_t *dest = (uint16_t *) _dest; \
803 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
804 dstW, uvalpha, y, fmt); \
807 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, AV_PIX_FMT_RGB48BE)
808 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, AV_PIX_FMT_RGB48LE)
809 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, AV_PIX_FMT_BGR48BE)
810 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, AV_PIX_FMT_BGR48LE)
813 * Write out 2 RGB pixels in the target pixel format. This function takes a
814 * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
815 * things like endianness conversion and shifting. The caller takes care of
816 * setting the correct offset in these tables from the chroma (U/V) values.
817 * This function then uses the luminance (Y1/Y2) values to write out the
818 * correct RGB values into the destination buffer.
820 static av_always_inline void
821 yuv2rgb_write(uint8_t *_dest, int i, unsigned Y1, unsigned Y2,
822 unsigned A1, unsigned A2,
823 const void *_r, const void *_g, const void *_b, int y,
824 enum AVPixelFormat target, int hasAlpha)
826 if (target == AV_PIX_FMT_ARGB || target == AV_PIX_FMT_RGBA ||
827 target == AV_PIX_FMT_ABGR || target == AV_PIX_FMT_BGRA) {
828 uint32_t *dest = (uint32_t *) _dest;
829 const uint32_t *r = (const uint32_t *) _r;
830 const uint32_t *g = (const uint32_t *) _g;
831 const uint32_t *b = (const uint32_t *) _b;
834 int sh = hasAlpha ? ((target == AV_PIX_FMT_RGB32_1 || target == AV_PIX_FMT_BGR32_1) ? 0 : 24) : 0;
836 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
837 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
840 int sh = (target == AV_PIX_FMT_RGB32_1 || target == AV_PIX_FMT_BGR32_1) ? 0 : 24;
842 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
843 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
845 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
846 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
849 } else if (target == AV_PIX_FMT_RGB24 || target == AV_PIX_FMT_BGR24) {
850 uint8_t *dest = (uint8_t *) _dest;
851 const uint8_t *r = (const uint8_t *) _r;
852 const uint8_t *g = (const uint8_t *) _g;
853 const uint8_t *b = (const uint8_t *) _b;
855 #define r_b ((target == AV_PIX_FMT_RGB24) ? r : b)
856 #define b_r ((target == AV_PIX_FMT_RGB24) ? b : r)
857 dest[i * 6 + 0] = r_b[Y1];
858 dest[i * 6 + 1] = g[Y1];
859 dest[i * 6 + 2] = b_r[Y1];
860 dest[i * 6 + 3] = r_b[Y2];
861 dest[i * 6 + 4] = g[Y2];
862 dest[i * 6 + 5] = b_r[Y2];
865 } else if (target == AV_PIX_FMT_RGB565 || target == AV_PIX_FMT_BGR565 ||
866 target == AV_PIX_FMT_RGB555 || target == AV_PIX_FMT_BGR555 ||
867 target == AV_PIX_FMT_RGB444 || target == AV_PIX_FMT_BGR444) {
868 uint16_t *dest = (uint16_t *) _dest;
869 const uint16_t *r = (const uint16_t *) _r;
870 const uint16_t *g = (const uint16_t *) _g;
871 const uint16_t *b = (const uint16_t *) _b;
872 int dr1, dg1, db1, dr2, dg2, db2;
874 if (target == AV_PIX_FMT_RGB565 || target == AV_PIX_FMT_BGR565) {
875 dr1 = dither_2x2_8[ y & 1 ][0];
876 dg1 = dither_2x2_4[ y & 1 ][0];
877 db1 = dither_2x2_8[(y & 1) ^ 1][0];
878 dr2 = dither_2x2_8[ y & 1 ][1];
879 dg2 = dither_2x2_4[ y & 1 ][1];
880 db2 = dither_2x2_8[(y & 1) ^ 1][1];
881 } else if (target == AV_PIX_FMT_RGB555 || target == AV_PIX_FMT_BGR555) {
882 dr1 = dither_2x2_8[ y & 1 ][0];
883 dg1 = dither_2x2_8[ y & 1 ][1];
884 db1 = dither_2x2_8[(y & 1) ^ 1][0];
885 dr2 = dither_2x2_8[ y & 1 ][1];
886 dg2 = dither_2x2_8[ y & 1 ][0];
887 db2 = dither_2x2_8[(y & 1) ^ 1][1];
889 dr1 = ff_dither_4x4_16[ y & 3 ][0];
890 dg1 = ff_dither_4x4_16[ y & 3 ][1];
891 db1 = ff_dither_4x4_16[(y & 3) ^ 3][0];
892 dr2 = ff_dither_4x4_16[ y & 3 ][1];
893 dg2 = ff_dither_4x4_16[ y & 3 ][0];
894 db2 = ff_dither_4x4_16[(y & 3) ^ 3][1];
897 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
898 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
899 } else /* 8/4-bit */ {
900 uint8_t *dest = (uint8_t *) _dest;
901 const uint8_t *r = (const uint8_t *) _r;
902 const uint8_t *g = (const uint8_t *) _g;
903 const uint8_t *b = (const uint8_t *) _b;
904 int dr1, dg1, db1, dr2, dg2, db2;
906 if (target == AV_PIX_FMT_RGB8 || target == AV_PIX_FMT_BGR8) {
907 const uint8_t * const d64 = ff_dither_8x8_73[y & 7];
908 const uint8_t * const d32 = ff_dither_8x8_32[y & 7];
909 dr1 = dg1 = d32[(i * 2 + 0) & 7];
910 db1 = d64[(i * 2 + 0) & 7];
911 dr2 = dg2 = d32[(i * 2 + 1) & 7];
912 db2 = d64[(i * 2 + 1) & 7];
914 const uint8_t * const d64 = ff_dither_8x8_73 [y & 7];
915 const uint8_t * const d128 = ff_dither_8x8_220[y & 7];
916 dr1 = db1 = d128[(i * 2 + 0) & 7];
917 dg1 = d64[(i * 2 + 0) & 7];
918 dr2 = db2 = d128[(i * 2 + 1) & 7];
919 dg2 = d64[(i * 2 + 1) & 7];
922 if (target == AV_PIX_FMT_RGB4 || target == AV_PIX_FMT_BGR4) {
923 dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
924 ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
926 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
927 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
932 static av_always_inline void
933 yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
934 const int16_t **lumSrc, int lumFilterSize,
935 const int16_t *chrFilter, const int16_t **chrUSrc,
936 const int16_t **chrVSrc, int chrFilterSize,
937 const int16_t **alpSrc, uint8_t *dest, int dstW,
938 int y, enum AVPixelFormat target, int hasAlpha)
942 for (i = 0; i < ((dstW + 1) >> 1); i++) {
948 const void *r, *g, *b;
950 for (j = 0; j < lumFilterSize; j++) {
951 Y1 += lumSrc[j][i * 2] * lumFilter[j];
952 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
954 for (j = 0; j < chrFilterSize; j++) {
955 U += chrUSrc[j][i] * chrFilter[j];
956 V += chrVSrc[j][i] * chrFilter[j];
962 if ((Y1 | Y2 | U | V) & 0x100) {
963 Y1 = av_clip_uint8(Y1);
964 Y2 = av_clip_uint8(Y2);
965 U = av_clip_uint8(U);
966 V = av_clip_uint8(V);
971 for (j = 0; j < lumFilterSize; j++) {
972 A1 += alpSrc[j][i * 2 ] * lumFilter[j];
973 A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
977 if ((A1 | A2) & 0x100) {
978 A1 = av_clip_uint8(A1);
979 A2 = av_clip_uint8(A2);
983 /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
985 g = (c->table_gU[U] + c->table_gV[V]);
988 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
989 r, g, b, y, target, hasAlpha);
993 static av_always_inline void
994 yuv2rgb_2_c_template(SwsContext *c, const int16_t *buf[2],
995 const int16_t *ubuf[2], const int16_t *vbuf[2],
996 const int16_t *abuf[2], uint8_t *dest, int dstW,
997 int yalpha, int uvalpha, int y,
998 enum AVPixelFormat target, int hasAlpha)
1000 const int16_t *buf0 = buf[0], *buf1 = buf[1],
1001 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
1002 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
1003 *abuf0 = hasAlpha ? abuf[0] : NULL,
1004 *abuf1 = hasAlpha ? abuf[1] : NULL;
1005 int yalpha1 = 4096 - yalpha;
1006 int uvalpha1 = 4096 - uvalpha;
1009 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1010 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
1011 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
1012 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
1013 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
1015 const void *r, *g, *b;
1017 Y1 = av_clip_uint8(Y1);
1018 Y2 = av_clip_uint8(Y2);
1019 U = av_clip_uint8(U);
1020 V = av_clip_uint8(V);
1023 g = (c->table_gU[U] + c->table_gV[V]);
1027 A1 = (abuf0[i * 2 ] * yalpha1 + abuf1[i * 2 ] * yalpha) >> 19;
1028 A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
1029 A1 = av_clip_uint8(A1);
1030 A2 = av_clip_uint8(A2);
1033 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1034 r, g, b, y, target, hasAlpha);
1038 static av_always_inline void
1039 yuv2rgb_1_c_template(SwsContext *c, const int16_t *buf0,
1040 const int16_t *ubuf[2], const int16_t *vbuf[2],
1041 const int16_t *abuf0, uint8_t *dest, int dstW,
1042 int uvalpha, int y, enum AVPixelFormat target,
1045 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
1048 if (uvalpha < 2048) {
1049 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1050 int Y1 = buf0[i * 2] >> 7;
1051 int Y2 = buf0[i * 2 + 1] >> 7;
1052 int U = ubuf0[i] >> 7;
1053 int V = vbuf0[i] >> 7;
1055 const void *r, *g, *b;
1057 Y1 = av_clip_uint8(Y1);
1058 Y2 = av_clip_uint8(Y2);
1059 U = av_clip_uint8(U);
1060 V = av_clip_uint8(V);
1063 g = (c->table_gU[U] + c->table_gV[V]);
1067 A1 = abuf0[i * 2 ] >> 7;
1068 A2 = abuf0[i * 2 + 1] >> 7;
1069 A1 = av_clip_uint8(A1);
1070 A2 = av_clip_uint8(A2);
1073 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1074 r, g, b, y, target, hasAlpha);
1077 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
1078 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1079 int Y1 = buf0[i * 2] >> 7;
1080 int Y2 = buf0[i * 2 + 1] >> 7;
1081 int U = (ubuf0[i] + ubuf1[i]) >> 8;
1082 int V = (vbuf0[i] + vbuf1[i]) >> 8;
1084 const void *r, *g, *b;
1086 Y1 = av_clip_uint8(Y1);
1087 Y2 = av_clip_uint8(Y2);
1088 U = av_clip_uint8(U);
1089 V = av_clip_uint8(V);
1092 g = (c->table_gU[U] + c->table_gV[V]);
1096 A1 = abuf0[i * 2 ] >> 7;
1097 A2 = abuf0[i * 2 + 1] >> 7;
1098 A1 = av_clip_uint8(A1);
1099 A2 = av_clip_uint8(A2);
1102 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1103 r, g, b, y, target, hasAlpha);
1108 #define YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1109 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
1110 const int16_t **lumSrc, int lumFilterSize, \
1111 const int16_t *chrFilter, const int16_t **chrUSrc, \
1112 const int16_t **chrVSrc, int chrFilterSize, \
1113 const int16_t **alpSrc, uint8_t *dest, int dstW, \
1116 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
1117 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
1118 alpSrc, dest, dstW, y, fmt, hasAlpha); \
1120 #define YUV2RGBWRAPPER(name, base, ext, fmt, hasAlpha) \
1121 YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1122 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
1123 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1124 const int16_t *abuf[2], uint8_t *dest, int dstW, \
1125 int yalpha, int uvalpha, int y) \
1127 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
1128 dest, dstW, yalpha, uvalpha, y, fmt, hasAlpha); \
1131 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
1132 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1133 const int16_t *abuf0, uint8_t *dest, int dstW, \
1134 int uvalpha, int y) \
1136 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
1137 dstW, uvalpha, y, fmt, hasAlpha); \
1141 YUV2RGBWRAPPER(yuv2rgb,, 32_1, AV_PIX_FMT_RGB32_1, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1142 YUV2RGBWRAPPER(yuv2rgb,, 32, AV_PIX_FMT_RGB32, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1144 #if CONFIG_SWSCALE_ALPHA
1145 YUV2RGBWRAPPER(yuv2rgb,, a32_1, AV_PIX_FMT_RGB32_1, 1)
1146 YUV2RGBWRAPPER(yuv2rgb,, a32, AV_PIX_FMT_RGB32, 1)
1148 YUV2RGBWRAPPER(yuv2rgb,, x32_1, AV_PIX_FMT_RGB32_1, 0)
1149 YUV2RGBWRAPPER(yuv2rgb,, x32, AV_PIX_FMT_RGB32, 0)
1151 YUV2RGBWRAPPER(yuv2, rgb, rgb24, AV_PIX_FMT_RGB24, 0)
1152 YUV2RGBWRAPPER(yuv2, rgb, bgr24, AV_PIX_FMT_BGR24, 0)
1153 YUV2RGBWRAPPER(yuv2rgb,, 16, AV_PIX_FMT_RGB565, 0)
1154 YUV2RGBWRAPPER(yuv2rgb,, 15, AV_PIX_FMT_RGB555, 0)
1155 YUV2RGBWRAPPER(yuv2rgb,, 12, AV_PIX_FMT_RGB444, 0)
1156 YUV2RGBWRAPPER(yuv2rgb,, 8, AV_PIX_FMT_RGB8, 0)
1157 YUV2RGBWRAPPER(yuv2rgb,, 4, AV_PIX_FMT_RGB4, 0)
1158 YUV2RGBWRAPPER(yuv2rgb,, 4b, AV_PIX_FMT_RGB4_BYTE, 0)
1160 static av_always_inline void
1161 yuv2rgb_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
1162 const int16_t **lumSrc, int lumFilterSize,
1163 const int16_t *chrFilter, const int16_t **chrUSrc,
1164 const int16_t **chrVSrc, int chrFilterSize,
1165 const int16_t **alpSrc, uint8_t *dest,
1166 int dstW, int y, enum AVPixelFormat target, int hasAlpha)
1169 int step = (target == AV_PIX_FMT_RGB24 || target == AV_PIX_FMT_BGR24) ? 3 : 4;
1171 for (i = 0; i < dstW; i++) {
1178 for (j = 0; j < lumFilterSize; j++) {
1179 Y += lumSrc[j][i] * lumFilter[j];
1181 for (j = 0; j < chrFilterSize; j++) {
1182 U += chrUSrc[j][i] * chrFilter[j];
1183 V += chrVSrc[j][i] * chrFilter[j];
1190 for (j = 0; j < lumFilterSize; j++) {
1191 A += alpSrc[j][i] * lumFilter[j];
1195 A = av_clip_uint8(A);
1197 Y -= c->yuv2rgb_y_offset;
1198 Y *= c->yuv2rgb_y_coeff;
1200 R = Y + V*c->yuv2rgb_v2r_coeff;
1201 G = Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;
1202 B = Y + U*c->yuv2rgb_u2b_coeff;
1203 if ((R | G | B) & 0xC0000000) {
1204 R = av_clip_uintp2(R, 30);
1205 G = av_clip_uintp2(G, 30);
1206 B = av_clip_uintp2(B, 30);
1210 case AV_PIX_FMT_ARGB:
1211 dest[0] = hasAlpha ? A : 255;
1216 case AV_PIX_FMT_RGB24:
1221 case AV_PIX_FMT_RGBA:
1225 dest[3] = hasAlpha ? A : 255;
1227 case AV_PIX_FMT_ABGR:
1228 dest[0] = hasAlpha ? A : 255;
1234 case AV_PIX_FMT_BGR24:
1239 case AV_PIX_FMT_BGRA:
1243 dest[3] = hasAlpha ? A : 255;
1251 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, AV_PIX_FMT_BGRA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1252 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, AV_PIX_FMT_ABGR, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1253 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, AV_PIX_FMT_RGBA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1254 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, AV_PIX_FMT_ARGB, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1256 #if CONFIG_SWSCALE_ALPHA
1257 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, AV_PIX_FMT_BGRA, 1)
1258 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, AV_PIX_FMT_ABGR, 1)
1259 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, AV_PIX_FMT_RGBA, 1)
1260 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, AV_PIX_FMT_ARGB, 1)
1262 YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, AV_PIX_FMT_BGRA, 0)
1263 YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, AV_PIX_FMT_ABGR, 0)
1264 YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, AV_PIX_FMT_RGBA, 0)
1265 YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, AV_PIX_FMT_ARGB, 0)
1267 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full, AV_PIX_FMT_BGR24, 0)
1268 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full, AV_PIX_FMT_RGB24, 0)
1271 yuv2gbrp_full_X_c(SwsContext *c, const int16_t *lumFilter,
1272 const int16_t **lumSrc, int lumFilterSize,
1273 const int16_t *chrFilter, const int16_t **chrUSrc,
1274 const int16_t **chrVSrc, int chrFilterSize,
1275 const int16_t **alpSrc, uint8_t **dest,
1278 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(c->dstFormat);
1281 uint16_t **dest16 = (uint16_t**)dest;
1282 int SH = 22 + 7 - desc->comp[0].depth_minus1;
1284 for (i = 0; i < dstW; i++) {
1287 int U = (1 << 9) - (128 << 19);
1288 int V = (1 << 9) - (128 << 19);
1291 for (j = 0; j < lumFilterSize; j++)
1292 Y += lumSrc[j][i] * lumFilter[j];
1294 for (j = 0; j < chrFilterSize; j++) {
1295 U += chrUSrc[j][i] * chrFilter[j];
1296 V += chrVSrc[j][i] * chrFilter[j];
1306 for (j = 0; j < lumFilterSize; j++)
1307 A += alpSrc[j][i] * lumFilter[j];
1312 A = av_clip_uint8(A);
1315 Y -= c->yuv2rgb_y_offset;
1316 Y *= c->yuv2rgb_y_coeff;
1318 R = Y + V * c->yuv2rgb_v2r_coeff;
1319 G = Y + V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
1320 B = Y + U * c->yuv2rgb_u2b_coeff;
1322 if ((R | G | B) & 0xC0000000) {
1323 R = av_clip_uintp2(R, 30);
1324 G = av_clip_uintp2(G, 30);
1325 B = av_clip_uintp2(B, 30);
1329 dest16[0][i] = G >> SH;
1330 dest16[1][i] = B >> SH;
1331 dest16[2][i] = R >> SH;
1333 dest[0][i] = G >> 22;
1334 dest[1][i] = B >> 22;
1335 dest[2][i] = R >> 22;
1338 if (SH != 22 && (!isBE(c->dstFormat)) != (!HAVE_BIGENDIAN)) {
1339 for (i = 0; i < dstW; i++) {
1340 dest16[0][i] = av_bswap16(dest16[0][i]);
1341 dest16[1][i] = av_bswap16(dest16[1][i]);
1342 dest16[2][i] = av_bswap16(dest16[2][i]);
1347 av_cold void ff_sws_init_output_funcs(SwsContext *c,
1348 yuv2planar1_fn *yuv2plane1,
1349 yuv2planarX_fn *yuv2planeX,
1350 yuv2interleavedX_fn *yuv2nv12cX,
1351 yuv2packed1_fn *yuv2packed1,
1352 yuv2packed2_fn *yuv2packed2,
1353 yuv2packedX_fn *yuv2packedX,
1354 yuv2anyX_fn *yuv2anyX)
1356 enum AVPixelFormat dstFormat = c->dstFormat;
1357 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(dstFormat);
1359 if (is16BPS(dstFormat)) {
1360 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_16BE_c : yuv2planeX_16LE_c;
1361 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_16BE_c : yuv2plane1_16LE_c;
1362 } else if (is9_OR_10BPS(dstFormat)) {
1363 if (desc->comp[0].depth_minus1 == 8) {
1364 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_9BE_c : yuv2planeX_9LE_c;
1365 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_9BE_c : yuv2plane1_9LE_c;
1367 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_10BE_c : yuv2planeX_10LE_c;
1368 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_10BE_c : yuv2plane1_10LE_c;
1371 *yuv2plane1 = yuv2plane1_8_c;
1372 *yuv2planeX = yuv2planeX_8_c;
1373 if (dstFormat == AV_PIX_FMT_NV12 || dstFormat == AV_PIX_FMT_NV21)
1374 *yuv2nv12cX = yuv2nv12cX_c;
1377 if(c->flags & SWS_FULL_CHR_H_INT) {
1378 switch (dstFormat) {
1379 case AV_PIX_FMT_RGBA:
1381 *yuv2packedX = yuv2rgba32_full_X_c;
1383 #if CONFIG_SWSCALE_ALPHA
1385 *yuv2packedX = yuv2rgba32_full_X_c;
1387 #endif /* CONFIG_SWSCALE_ALPHA */
1389 *yuv2packedX = yuv2rgbx32_full_X_c;
1391 #endif /* !CONFIG_SMALL */
1393 case AV_PIX_FMT_ARGB:
1395 *yuv2packedX = yuv2argb32_full_X_c;
1397 #if CONFIG_SWSCALE_ALPHA
1399 *yuv2packedX = yuv2argb32_full_X_c;
1401 #endif /* CONFIG_SWSCALE_ALPHA */
1403 *yuv2packedX = yuv2xrgb32_full_X_c;
1405 #endif /* !CONFIG_SMALL */
1407 case AV_PIX_FMT_BGRA:
1409 *yuv2packedX = yuv2bgra32_full_X_c;
1411 #if CONFIG_SWSCALE_ALPHA
1413 *yuv2packedX = yuv2bgra32_full_X_c;
1415 #endif /* CONFIG_SWSCALE_ALPHA */
1417 *yuv2packedX = yuv2bgrx32_full_X_c;
1419 #endif /* !CONFIG_SMALL */
1421 case AV_PIX_FMT_ABGR:
1423 *yuv2packedX = yuv2abgr32_full_X_c;
1425 #if CONFIG_SWSCALE_ALPHA
1427 *yuv2packedX = yuv2abgr32_full_X_c;
1429 #endif /* CONFIG_SWSCALE_ALPHA */
1431 *yuv2packedX = yuv2xbgr32_full_X_c;
1433 #endif /* !CONFIG_SMALL */
1435 case AV_PIX_FMT_RGB24:
1436 *yuv2packedX = yuv2rgb24_full_X_c;
1438 case AV_PIX_FMT_BGR24:
1439 *yuv2packedX = yuv2bgr24_full_X_c;
1441 case AV_PIX_FMT_GBRP:
1442 case AV_PIX_FMT_GBRP9BE:
1443 case AV_PIX_FMT_GBRP9LE:
1444 case AV_PIX_FMT_GBRP10BE:
1445 case AV_PIX_FMT_GBRP10LE:
1446 case AV_PIX_FMT_GBRP16BE:
1447 case AV_PIX_FMT_GBRP16LE:
1448 *yuv2anyX = yuv2gbrp_full_X_c;
1452 switch (dstFormat) {
1453 case AV_PIX_FMT_RGB48LE:
1454 *yuv2packed1 = yuv2rgb48le_1_c;
1455 *yuv2packed2 = yuv2rgb48le_2_c;
1456 *yuv2packedX = yuv2rgb48le_X_c;
1458 case AV_PIX_FMT_RGB48BE:
1459 *yuv2packed1 = yuv2rgb48be_1_c;
1460 *yuv2packed2 = yuv2rgb48be_2_c;
1461 *yuv2packedX = yuv2rgb48be_X_c;
1463 case AV_PIX_FMT_BGR48LE:
1464 *yuv2packed1 = yuv2bgr48le_1_c;
1465 *yuv2packed2 = yuv2bgr48le_2_c;
1466 *yuv2packedX = yuv2bgr48le_X_c;
1468 case AV_PIX_FMT_BGR48BE:
1469 *yuv2packed1 = yuv2bgr48be_1_c;
1470 *yuv2packed2 = yuv2bgr48be_2_c;
1471 *yuv2packedX = yuv2bgr48be_X_c;
1473 case AV_PIX_FMT_RGB32:
1474 case AV_PIX_FMT_BGR32:
1476 *yuv2packed1 = yuv2rgb32_1_c;
1477 *yuv2packed2 = yuv2rgb32_2_c;
1478 *yuv2packedX = yuv2rgb32_X_c;
1480 #if CONFIG_SWSCALE_ALPHA
1482 *yuv2packed1 = yuv2rgba32_1_c;
1483 *yuv2packed2 = yuv2rgba32_2_c;
1484 *yuv2packedX = yuv2rgba32_X_c;
1486 #endif /* CONFIG_SWSCALE_ALPHA */
1488 *yuv2packed1 = yuv2rgbx32_1_c;
1489 *yuv2packed2 = yuv2rgbx32_2_c;
1490 *yuv2packedX = yuv2rgbx32_X_c;
1492 #endif /* !CONFIG_SMALL */
1494 case AV_PIX_FMT_RGB32_1:
1495 case AV_PIX_FMT_BGR32_1:
1497 *yuv2packed1 = yuv2rgb32_1_1_c;
1498 *yuv2packed2 = yuv2rgb32_1_2_c;
1499 *yuv2packedX = yuv2rgb32_1_X_c;
1501 #if CONFIG_SWSCALE_ALPHA
1503 *yuv2packed1 = yuv2rgba32_1_1_c;
1504 *yuv2packed2 = yuv2rgba32_1_2_c;
1505 *yuv2packedX = yuv2rgba32_1_X_c;
1507 #endif /* CONFIG_SWSCALE_ALPHA */
1509 *yuv2packed1 = yuv2rgbx32_1_1_c;
1510 *yuv2packed2 = yuv2rgbx32_1_2_c;
1511 *yuv2packedX = yuv2rgbx32_1_X_c;
1513 #endif /* !CONFIG_SMALL */
1515 case AV_PIX_FMT_RGB24:
1516 *yuv2packed1 = yuv2rgb24_1_c;
1517 *yuv2packed2 = yuv2rgb24_2_c;
1518 *yuv2packedX = yuv2rgb24_X_c;
1520 case AV_PIX_FMT_BGR24:
1521 *yuv2packed1 = yuv2bgr24_1_c;
1522 *yuv2packed2 = yuv2bgr24_2_c;
1523 *yuv2packedX = yuv2bgr24_X_c;
1525 case AV_PIX_FMT_RGB565LE:
1526 case AV_PIX_FMT_RGB565BE:
1527 case AV_PIX_FMT_BGR565LE:
1528 case AV_PIX_FMT_BGR565BE:
1529 *yuv2packed1 = yuv2rgb16_1_c;
1530 *yuv2packed2 = yuv2rgb16_2_c;
1531 *yuv2packedX = yuv2rgb16_X_c;
1533 case AV_PIX_FMT_RGB555LE:
1534 case AV_PIX_FMT_RGB555BE:
1535 case AV_PIX_FMT_BGR555LE:
1536 case AV_PIX_FMT_BGR555BE:
1537 *yuv2packed1 = yuv2rgb15_1_c;
1538 *yuv2packed2 = yuv2rgb15_2_c;
1539 *yuv2packedX = yuv2rgb15_X_c;
1541 case AV_PIX_FMT_RGB444LE:
1542 case AV_PIX_FMT_RGB444BE:
1543 case AV_PIX_FMT_BGR444LE:
1544 case AV_PIX_FMT_BGR444BE:
1545 *yuv2packed1 = yuv2rgb12_1_c;
1546 *yuv2packed2 = yuv2rgb12_2_c;
1547 *yuv2packedX = yuv2rgb12_X_c;
1549 case AV_PIX_FMT_RGB8:
1550 case AV_PIX_FMT_BGR8:
1551 *yuv2packed1 = yuv2rgb8_1_c;
1552 *yuv2packed2 = yuv2rgb8_2_c;
1553 *yuv2packedX = yuv2rgb8_X_c;
1555 case AV_PIX_FMT_RGB4:
1556 case AV_PIX_FMT_BGR4:
1557 *yuv2packed1 = yuv2rgb4_1_c;
1558 *yuv2packed2 = yuv2rgb4_2_c;
1559 *yuv2packedX = yuv2rgb4_X_c;
1561 case AV_PIX_FMT_RGB4_BYTE:
1562 case AV_PIX_FMT_BGR4_BYTE:
1563 *yuv2packed1 = yuv2rgb4b_1_c;
1564 *yuv2packed2 = yuv2rgb4b_2_c;
1565 *yuv2packedX = yuv2rgb4b_X_c;
1569 switch (dstFormat) {
1570 case AV_PIX_FMT_MONOWHITE:
1571 *yuv2packed1 = yuv2monowhite_1_c;
1572 *yuv2packed2 = yuv2monowhite_2_c;
1573 *yuv2packedX = yuv2monowhite_X_c;
1575 case AV_PIX_FMT_MONOBLACK:
1576 *yuv2packed1 = yuv2monoblack_1_c;
1577 *yuv2packed2 = yuv2monoblack_2_c;
1578 *yuv2packedX = yuv2monoblack_X_c;
1580 case AV_PIX_FMT_YUYV422:
1581 *yuv2packed1 = yuv2yuyv422_1_c;
1582 *yuv2packed2 = yuv2yuyv422_2_c;
1583 *yuv2packedX = yuv2yuyv422_X_c;
1585 case AV_PIX_FMT_YVYU422:
1586 *yuv2packed1 = yuv2yvyu422_1_c;
1587 *yuv2packed2 = yuv2yvyu422_2_c;
1588 *yuv2packedX = yuv2yvyu422_X_c;
1590 case AV_PIX_FMT_UYVY422:
1591 *yuv2packed1 = yuv2uyvy422_1_c;
1592 *yuv2packed2 = yuv2uyvy422_2_c;
1593 *yuv2packedX = yuv2uyvy422_X_c;