2 * Copyright (C) 2001-2012 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
27 #include "libavutil/avutil.h"
28 #include "libavutil/avassert.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, 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, 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, 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, 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, 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, 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, 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)
140 av_assert0(output_bits == 16);
142 for (i = 0; i < dstW; i++) {
143 int val = src[i] + (1 << (shift - 1));
144 output_pixel(&dest[i], val, 0, uint);
148 static av_always_inline void
149 yuv2planeX_16_c_template(const int16_t *filter, int filterSize,
150 const int32_t **src, uint16_t *dest, int dstW,
151 int big_endian, int output_bits)
155 av_assert0(output_bits == 16);
157 for (i = 0; i < dstW; i++) {
158 int val = 1 << (shift - 1);
161 /* range of val is [0,0x7FFFFFFF], so 31 bits, but with lanczos/spline
162 * filters (or anything with negative coeffs, the range can be slightly
163 * wider in both directions. To account for this overflow, we subtract
164 * a constant so it always fits in the signed range (assuming a
165 * reasonable filterSize), and re-add that at the end. */
167 for (j = 0; j < filterSize; j++)
168 val += src[j][i] * filter[j];
170 output_pixel(&dest[i], val, 0x8000, int);
176 #define output_pixel(pos, val) \
178 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
180 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
183 static av_always_inline void
184 yuv2plane1_10_c_template(const int16_t *src, uint16_t *dest, int dstW,
185 int big_endian, int output_bits)
188 int shift = 15 - output_bits;
190 for (i = 0; i < dstW; i++) {
191 int val = src[i] + (1 << (shift - 1));
192 output_pixel(&dest[i], val);
196 static av_always_inline void
197 yuv2planeX_10_c_template(const int16_t *filter, int filterSize,
198 const int16_t **src, uint16_t *dest, int dstW,
199 int big_endian, int output_bits)
202 int shift = 11 + 16 - output_bits;
204 for (i = 0; i < dstW; i++) {
205 int val = 1 << (shift - 1);
208 for (j = 0; j < filterSize; j++)
209 val += src[j][i] * filter[j];
211 output_pixel(&dest[i], val);
217 #define yuv2NBPS(bits, BE_LE, is_be, template_size, typeX_t) \
218 static void yuv2plane1_ ## bits ## BE_LE ## _c(const int16_t *src, \
219 uint8_t *dest, int dstW, \
220 const uint8_t *dither, int offset)\
222 yuv2plane1_ ## template_size ## _c_template((const typeX_t *) src, \
223 (uint16_t *) dest, dstW, is_be, bits); \
225 static void yuv2planeX_ ## bits ## BE_LE ## _c(const int16_t *filter, int filterSize, \
226 const int16_t **src, uint8_t *dest, int dstW, \
227 const uint8_t *dither, int offset)\
229 yuv2planeX_## template_size ## _c_template(filter, \
230 filterSize, (const typeX_t **) src, \
231 (uint16_t *) dest, dstW, is_be, bits); \
233 yuv2NBPS( 9, BE, 1, 10, int16_t)
234 yuv2NBPS( 9, LE, 0, 10, int16_t)
235 yuv2NBPS(10, BE, 1, 10, int16_t)
236 yuv2NBPS(10, LE, 0, 10, int16_t)
237 yuv2NBPS(16, BE, 1, 16, int32_t)
238 yuv2NBPS(16, LE, 0, 16, int32_t)
240 static void yuv2planeX_8_c(const int16_t *filter, int filterSize,
241 const int16_t **src, uint8_t *dest, int dstW,
242 const uint8_t *dither, int offset)
245 for (i=0; i<dstW; i++) {
246 int val = dither[(i + offset) & 7] << 12;
248 for (j=0; j<filterSize; j++)
249 val += src[j][i] * filter[j];
251 dest[i]= av_clip_uint8(val>>19);
255 static void yuv2plane1_8_c(const int16_t *src, uint8_t *dest, int dstW,
256 const uint8_t *dither, int offset)
259 for (i=0; i<dstW; i++) {
260 int val = (src[i] + dither[(i + offset) & 7]) >> 7;
261 dest[i]= av_clip_uint8(val);
265 static void yuv2nv12cX_c(SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
266 const int16_t **chrUSrc, const int16_t **chrVSrc,
267 uint8_t *dest, int chrDstW)
269 enum PixelFormat dstFormat = c->dstFormat;
270 const uint8_t *chrDither = c->chrDither8;
273 if (dstFormat == PIX_FMT_NV12)
274 for (i=0; i<chrDstW; i++) {
275 int u = chrDither[i & 7] << 12;
276 int v = chrDither[(i + 3) & 7] << 12;
278 for (j=0; j<chrFilterSize; j++) {
279 u += chrUSrc[j][i] * chrFilter[j];
280 v += chrVSrc[j][i] * chrFilter[j];
283 dest[2*i]= av_clip_uint8(u>>19);
284 dest[2*i+1]= av_clip_uint8(v>>19);
287 for (i=0; i<chrDstW; i++) {
288 int u = chrDither[i & 7] << 12;
289 int v = chrDither[(i + 3) & 7] << 12;
291 for (j=0; j<chrFilterSize; j++) {
292 u += chrUSrc[j][i] * chrFilter[j];
293 v += chrVSrc[j][i] * chrFilter[j];
296 dest[2*i]= av_clip_uint8(v>>19);
297 dest[2*i+1]= av_clip_uint8(u>>19);
301 #define accumulate_bit(acc, val) \
303 acc |= (val) >= (128 + 110)
304 #define output_pixel(pos, acc) \
305 if (target == PIX_FMT_MONOBLACK) { \
311 static av_always_inline void
312 yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
313 const int16_t **lumSrc, int lumFilterSize,
314 const int16_t *chrFilter, const int16_t **chrUSrc,
315 const int16_t **chrVSrc, int chrFilterSize,
316 const int16_t **alpSrc, uint8_t *dest, int dstW,
317 int y, enum PixelFormat target)
319 const uint8_t * const d128=dither_8x8_220[y&7];
323 for (i = 0; i < dstW; i += 2) {
328 for (j = 0; j < lumFilterSize; j++) {
329 Y1 += lumSrc[j][i] * lumFilter[j];
330 Y2 += lumSrc[j][i+1] * lumFilter[j];
334 if ((Y1 | Y2) & 0x100) {
335 Y1 = av_clip_uint8(Y1);
336 Y2 = av_clip_uint8(Y2);
338 accumulate_bit(acc, Y1 + d128[(i + 0) & 7]);
339 accumulate_bit(acc, Y2 + d128[(i + 1) & 7]);
341 output_pixel(*dest++, acc);
346 output_pixel(*dest, acc);
350 static av_always_inline void
351 yuv2mono_2_c_template(SwsContext *c, const int16_t *buf[2],
352 const int16_t *ubuf[2], const int16_t *vbuf[2],
353 const int16_t *abuf[2], uint8_t *dest, int dstW,
354 int yalpha, int uvalpha, int y,
355 enum PixelFormat target)
357 const int16_t *buf0 = buf[0], *buf1 = buf[1];
358 const uint8_t * const d128 = dither_8x8_220[y & 7];
359 int yalpha1 = 4095 - yalpha;
362 for (i = 0; i < dstW; i += 8) {
365 Y = (buf0[i + 0] * yalpha1 + buf1[i + 0] * yalpha) >> 19;
366 accumulate_bit(acc, Y + d128[0]);
367 Y = (buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19;
368 accumulate_bit(acc, Y + d128[1]);
369 Y = (buf0[i + 2] * yalpha1 + buf1[i + 2] * yalpha) >> 19;
370 accumulate_bit(acc, Y + d128[2]);
371 Y = (buf0[i + 3] * yalpha1 + buf1[i + 3] * yalpha) >> 19;
372 accumulate_bit(acc, Y + d128[3]);
373 Y = (buf0[i + 4] * yalpha1 + buf1[i + 4] * yalpha) >> 19;
374 accumulate_bit(acc, Y + d128[4]);
375 Y = (buf0[i + 5] * yalpha1 + buf1[i + 5] * yalpha) >> 19;
376 accumulate_bit(acc, Y + d128[5]);
377 Y = (buf0[i + 6] * yalpha1 + buf1[i + 6] * yalpha) >> 19;
378 accumulate_bit(acc, Y + d128[6]);
379 Y = (buf0[i + 7] * yalpha1 + buf1[i + 7] * yalpha) >> 19;
380 accumulate_bit(acc, Y + d128[7]);
382 output_pixel(*dest++, acc);
386 static av_always_inline void
387 yuv2mono_1_c_template(SwsContext *c, const int16_t *buf0,
388 const int16_t *ubuf[2], const int16_t *vbuf[2],
389 const int16_t *abuf0, uint8_t *dest, int dstW,
390 int uvalpha, int y, enum PixelFormat target)
392 const uint8_t * const d128 = dither_8x8_220[y & 7];
395 for (i = 0; i < dstW; i += 8) {
398 accumulate_bit(acc, ((buf0[i + 0] + 64) >> 7) + d128[0]);
399 accumulate_bit(acc, ((buf0[i + 1] + 64) >> 7) + d128[1]);
400 accumulate_bit(acc, ((buf0[i + 2] + 64) >> 7) + d128[2]);
401 accumulate_bit(acc, ((buf0[i + 3] + 64) >> 7) + d128[3]);
402 accumulate_bit(acc, ((buf0[i + 4] + 64) >> 7) + d128[4]);
403 accumulate_bit(acc, ((buf0[i + 5] + 64) >> 7) + d128[5]);
404 accumulate_bit(acc, ((buf0[i + 6] + 64) >> 7) + d128[6]);
405 accumulate_bit(acc, ((buf0[i + 7] + 64) >> 7) + d128[7]);
407 output_pixel(*dest++, acc);
412 #undef accumulate_bit
414 #define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
415 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
416 const int16_t **lumSrc, int lumFilterSize, \
417 const int16_t *chrFilter, const int16_t **chrUSrc, \
418 const int16_t **chrVSrc, int chrFilterSize, \
419 const int16_t **alpSrc, uint8_t *dest, int dstW, \
422 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
423 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
424 alpSrc, dest, dstW, y, fmt); \
427 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
428 const int16_t *ubuf[2], const int16_t *vbuf[2], \
429 const int16_t *abuf[2], uint8_t *dest, int dstW, \
430 int yalpha, int uvalpha, int y) \
432 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
433 dest, dstW, yalpha, uvalpha, y, fmt); \
436 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
437 const int16_t *ubuf[2], const int16_t *vbuf[2], \
438 const int16_t *abuf0, uint8_t *dest, int dstW, \
439 int uvalpha, int y) \
441 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, \
442 abuf0, dest, dstW, uvalpha, \
446 YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE)
447 YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK)
449 #define output_pixels(pos, Y1, U, Y2, V) \
450 if (target == PIX_FMT_YUYV422) { \
451 dest[pos + 0] = Y1; \
453 dest[pos + 2] = Y2; \
457 dest[pos + 1] = Y1; \
459 dest[pos + 3] = Y2; \
462 static av_always_inline void
463 yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
464 const int16_t **lumSrc, int lumFilterSize,
465 const int16_t *chrFilter, const int16_t **chrUSrc,
466 const int16_t **chrVSrc, int chrFilterSize,
467 const int16_t **alpSrc, uint8_t *dest, int dstW,
468 int y, enum PixelFormat target)
472 for (i = 0; i < (dstW >> 1); i++) {
479 for (j = 0; j < lumFilterSize; j++) {
480 Y1 += lumSrc[j][i * 2] * lumFilter[j];
481 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
483 for (j = 0; j < chrFilterSize; j++) {
484 U += chrUSrc[j][i] * chrFilter[j];
485 V += chrVSrc[j][i] * chrFilter[j];
491 if ((Y1 | Y2 | U | V) & 0x100) {
492 Y1 = av_clip_uint8(Y1);
493 Y2 = av_clip_uint8(Y2);
494 U = av_clip_uint8(U);
495 V = av_clip_uint8(V);
497 output_pixels(4*i, Y1, U, Y2, V);
501 static av_always_inline void
502 yuv2422_2_c_template(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 enum PixelFormat target)
508 const int16_t *buf0 = buf[0], *buf1 = buf[1],
509 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
510 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
511 int yalpha1 = 4095 - yalpha;
512 int uvalpha1 = 4095 - uvalpha;
515 for (i = 0; i < (dstW >> 1); i++) {
516 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
517 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
518 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
519 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
521 if ((Y1 | Y2 | U | V) & 0x100) {
522 Y1 = av_clip_uint8(Y1);
523 Y2 = av_clip_uint8(Y2);
524 U = av_clip_uint8(U);
525 V = av_clip_uint8(V);
528 output_pixels(i * 4, Y1, U, Y2, V);
532 static av_always_inline void
533 yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
534 const int16_t *ubuf[2], const int16_t *vbuf[2],
535 const int16_t *abuf0, uint8_t *dest, int dstW,
536 int uvalpha, int y, enum PixelFormat target)
538 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
541 if (uvalpha < 2048) {
542 for (i = 0; i < (dstW >> 1); i++) {
543 int Y1 = (buf0[i * 2 ]+64) >> 7;
544 int Y2 = (buf0[i * 2 + 1]+64) >> 7;
545 int U = (ubuf0[i] +64) >> 7;
546 int V = (vbuf0[i] +64) >> 7;
548 if ((Y1 | Y2 | U | V) & 0x100) {
549 Y1 = av_clip_uint8(Y1);
550 Y2 = av_clip_uint8(Y2);
551 U = av_clip_uint8(U);
552 V = av_clip_uint8(V);
555 Y1 = av_clip_uint8(Y1);
556 Y2 = av_clip_uint8(Y2);
557 U = av_clip_uint8(U);
558 V = av_clip_uint8(V);
560 output_pixels(i * 4, Y1, U, Y2, V);
563 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
564 for (i = 0; i < (dstW >> 1); i++) {
565 int Y1 = (buf0[i * 2 ] + 64) >> 7;
566 int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
567 int U = (ubuf0[i] + ubuf1[i]+128) >> 8;
568 int V = (vbuf0[i] + vbuf1[i]+128) >> 8;
570 if ((Y1 | Y2 | U | V) & 0x100) {
571 Y1 = av_clip_uint8(Y1);
572 Y2 = av_clip_uint8(Y2);
573 U = av_clip_uint8(U);
574 V = av_clip_uint8(V);
577 Y1 = av_clip_uint8(Y1);
578 Y2 = av_clip_uint8(Y2);
579 U = av_clip_uint8(U);
580 V = av_clip_uint8(V);
582 output_pixels(i * 4, Y1, U, Y2, V);
589 YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422)
590 YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422)
592 #define R_B ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? R : B)
593 #define B_R ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? B : R)
594 #define output_pixel(pos, val) \
595 if (isBE(target)) { \
601 static av_always_inline void
602 yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
603 const int32_t **lumSrc, int lumFilterSize,
604 const int16_t *chrFilter, const int32_t **chrUSrc,
605 const int32_t **chrVSrc, int chrFilterSize,
606 const int32_t **alpSrc, uint16_t *dest, int dstW,
607 int y, enum PixelFormat target)
611 for (i = 0; i < (dstW >> 1); i++) {
613 int Y1 = -0x40000000;
614 int Y2 = -0x40000000;
615 int U = -128 << 23; // 19
619 for (j = 0; j < lumFilterSize; j++) {
620 Y1 += lumSrc[j][i * 2] * lumFilter[j];
621 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
623 for (j = 0; j < chrFilterSize; j++) {
624 U += chrUSrc[j][i] * chrFilter[j];
625 V += chrVSrc[j][i] * chrFilter[j];
628 // 8bit: 12+15=27; 16-bit: 12+19=31
636 // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
637 Y1 -= c->yuv2rgb_y_offset;
638 Y2 -= c->yuv2rgb_y_offset;
639 Y1 *= c->yuv2rgb_y_coeff;
640 Y2 *= c->yuv2rgb_y_coeff;
643 // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
645 R = V * c->yuv2rgb_v2r_coeff;
646 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
647 B = U * c->yuv2rgb_u2b_coeff;
649 // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
650 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
651 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
652 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
653 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
654 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
655 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
660 static av_always_inline void
661 yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
662 const int32_t *ubuf[2], const int32_t *vbuf[2],
663 const int32_t *abuf[2], uint16_t *dest, int dstW,
664 int yalpha, int uvalpha, int y,
665 enum PixelFormat target)
667 const int32_t *buf0 = buf[0], *buf1 = buf[1],
668 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
669 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
670 int yalpha1 = 4095 - yalpha;
671 int uvalpha1 = 4095 - uvalpha;
674 for (i = 0; i < (dstW >> 1); i++) {
675 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 14;
676 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 14;
677 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha + (-128 << 23)) >> 14;
678 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha + (-128 << 23)) >> 14;
681 Y1 -= c->yuv2rgb_y_offset;
682 Y2 -= c->yuv2rgb_y_offset;
683 Y1 *= c->yuv2rgb_y_coeff;
684 Y2 *= c->yuv2rgb_y_coeff;
688 R = V * c->yuv2rgb_v2r_coeff;
689 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
690 B = U * c->yuv2rgb_u2b_coeff;
692 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
693 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
694 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
695 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
696 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
697 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
702 static av_always_inline void
703 yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
704 const int32_t *ubuf[2], const int32_t *vbuf[2],
705 const int32_t *abuf0, uint16_t *dest, int dstW,
706 int uvalpha, int y, enum PixelFormat target)
708 const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
711 if (uvalpha < 2048) {
712 for (i = 0; i < (dstW >> 1); i++) {
713 int Y1 = (buf0[i * 2] ) >> 2;
714 int Y2 = (buf0[i * 2 + 1]) >> 2;
715 int U = (ubuf0[i] + (-128 << 11)) >> 2;
716 int V = (vbuf0[i] + (-128 << 11)) >> 2;
719 Y1 -= c->yuv2rgb_y_offset;
720 Y2 -= c->yuv2rgb_y_offset;
721 Y1 *= c->yuv2rgb_y_coeff;
722 Y2 *= c->yuv2rgb_y_coeff;
726 R = V * c->yuv2rgb_v2r_coeff;
727 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
728 B = U * c->yuv2rgb_u2b_coeff;
730 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
731 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
732 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
733 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
734 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
735 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
739 const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
740 for (i = 0; i < (dstW >> 1); i++) {
741 int Y1 = (buf0[i * 2] ) >> 2;
742 int Y2 = (buf0[i * 2 + 1]) >> 2;
743 int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
744 int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
747 Y1 -= c->yuv2rgb_y_offset;
748 Y2 -= c->yuv2rgb_y_offset;
749 Y1 *= c->yuv2rgb_y_coeff;
750 Y2 *= c->yuv2rgb_y_coeff;
754 R = V * c->yuv2rgb_v2r_coeff;
755 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
756 B = U * c->yuv2rgb_u2b_coeff;
758 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
759 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
760 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
761 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
762 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
763 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
773 #define YUV2PACKED16WRAPPER(name, base, ext, fmt) \
774 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
775 const int16_t **_lumSrc, int lumFilterSize, \
776 const int16_t *chrFilter, const int16_t **_chrUSrc, \
777 const int16_t **_chrVSrc, int chrFilterSize, \
778 const int16_t **_alpSrc, uint8_t *_dest, int dstW, \
781 const int32_t **lumSrc = (const int32_t **) _lumSrc, \
782 **chrUSrc = (const int32_t **) _chrUSrc, \
783 **chrVSrc = (const int32_t **) _chrVSrc, \
784 **alpSrc = (const int32_t **) _alpSrc; \
785 uint16_t *dest = (uint16_t *) _dest; \
786 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
787 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
788 alpSrc, dest, dstW, y, fmt); \
791 static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
792 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
793 const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
794 int yalpha, int uvalpha, int y) \
796 const int32_t **buf = (const int32_t **) _buf, \
797 **ubuf = (const int32_t **) _ubuf, \
798 **vbuf = (const int32_t **) _vbuf, \
799 **abuf = (const int32_t **) _abuf; \
800 uint16_t *dest = (uint16_t *) _dest; \
801 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
802 dest, dstW, yalpha, uvalpha, y, fmt); \
805 static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
806 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
807 const int16_t *_abuf0, uint8_t *_dest, int dstW, \
808 int uvalpha, int y) \
810 const int32_t *buf0 = (const int32_t *) _buf0, \
811 **ubuf = (const int32_t **) _ubuf, \
812 **vbuf = (const int32_t **) _vbuf, \
813 *abuf0 = (const int32_t *) _abuf0; \
814 uint16_t *dest = (uint16_t *) _dest; \
815 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
816 dstW, uvalpha, y, fmt); \
819 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE)
820 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE)
821 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE)
822 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE)
825 * Write out 2 RGB pixels in the target pixel format. This function takes a
826 * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
827 * things like endianness conversion and shifting. The caller takes care of
828 * setting the correct offset in these tables from the chroma (U/V) values.
829 * This function then uses the luminance (Y1/Y2) values to write out the
830 * correct RGB values into the destination buffer.
832 static av_always_inline void
833 yuv2rgb_write(uint8_t *_dest, int i, int Y1, int Y2,
834 unsigned A1, unsigned A2,
835 const void *_r, const void *_g, const void *_b, int y,
836 enum PixelFormat target, int hasAlpha)
838 if (target == PIX_FMT_ARGB || target == PIX_FMT_RGBA ||
839 target == PIX_FMT_ABGR || target == PIX_FMT_BGRA) {
840 uint32_t *dest = (uint32_t *) _dest;
841 const uint32_t *r = (const uint32_t *) _r;
842 const uint32_t *g = (const uint32_t *) _g;
843 const uint32_t *b = (const uint32_t *) _b;
846 int sh = hasAlpha ? ((target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24) : 0;
848 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
849 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
852 int sh = (target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24;
854 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
855 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
857 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
858 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
861 } else if (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) {
862 uint8_t *dest = (uint8_t *) _dest;
863 const uint8_t *r = (const uint8_t *) _r;
864 const uint8_t *g = (const uint8_t *) _g;
865 const uint8_t *b = (const uint8_t *) _b;
867 #define r_b ((target == PIX_FMT_RGB24) ? r : b)
868 #define b_r ((target == PIX_FMT_RGB24) ? b : r)
870 dest[i * 6 + 0] = r_b[Y1];
871 dest[i * 6 + 1] = g[Y1];
872 dest[i * 6 + 2] = b_r[Y1];
873 dest[i * 6 + 3] = r_b[Y2];
874 dest[i * 6 + 4] = g[Y2];
875 dest[i * 6 + 5] = b_r[Y2];
878 } else if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565 ||
879 target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555 ||
880 target == PIX_FMT_RGB444 || target == PIX_FMT_BGR444) {
881 uint16_t *dest = (uint16_t *) _dest;
882 const uint16_t *r = (const uint16_t *) _r;
883 const uint16_t *g = (const uint16_t *) _g;
884 const uint16_t *b = (const uint16_t *) _b;
885 int dr1, dg1, db1, dr2, dg2, db2;
887 if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565) {
888 dr1 = dither_2x2_8[ y & 1 ][0];
889 dg1 = dither_2x2_4[ y & 1 ][0];
890 db1 = dither_2x2_8[(y & 1) ^ 1][0];
891 dr2 = dither_2x2_8[ y & 1 ][1];
892 dg2 = dither_2x2_4[ y & 1 ][1];
893 db2 = dither_2x2_8[(y & 1) ^ 1][1];
894 } else if (target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555) {
895 dr1 = dither_2x2_8[ y & 1 ][0];
896 dg1 = dither_2x2_8[ y & 1 ][1];
897 db1 = dither_2x2_8[(y & 1) ^ 1][0];
898 dr2 = dither_2x2_8[ y & 1 ][1];
899 dg2 = dither_2x2_8[ y & 1 ][0];
900 db2 = dither_2x2_8[(y & 1) ^ 1][1];
902 dr1 = dither_4x4_16[ y & 3 ][0];
903 dg1 = dither_4x4_16[ y & 3 ][1];
904 db1 = dither_4x4_16[(y & 3) ^ 3][0];
905 dr2 = dither_4x4_16[ y & 3 ][1];
906 dg2 = dither_4x4_16[ y & 3 ][0];
907 db2 = dither_4x4_16[(y & 3) ^ 3][1];
910 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
911 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
912 } else /* 8/4-bit */ {
913 uint8_t *dest = (uint8_t *) _dest;
914 const uint8_t *r = (const uint8_t *) _r;
915 const uint8_t *g = (const uint8_t *) _g;
916 const uint8_t *b = (const uint8_t *) _b;
917 int dr1, dg1, db1, dr2, dg2, db2;
919 if (target == PIX_FMT_RGB8 || target == PIX_FMT_BGR8) {
920 const uint8_t * const d64 = dither_8x8_73[y & 7];
921 const uint8_t * const d32 = dither_8x8_32[y & 7];
922 dr1 = dg1 = d32[(i * 2 + 0) & 7];
923 db1 = d64[(i * 2 + 0) & 7];
924 dr2 = dg2 = d32[(i * 2 + 1) & 7];
925 db2 = d64[(i * 2 + 1) & 7];
927 const uint8_t * const d64 = dither_8x8_73 [y & 7];
928 const uint8_t * const d128 = dither_8x8_220[y & 7];
929 dr1 = db1 = d128[(i * 2 + 0) & 7];
930 dg1 = d64[(i * 2 + 0) & 7];
931 dr2 = db2 = d128[(i * 2 + 1) & 7];
932 dg2 = d64[(i * 2 + 1) & 7];
935 if (target == PIX_FMT_RGB4 || target == PIX_FMT_BGR4) {
936 dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
937 ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
939 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
940 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
945 static av_always_inline void
946 yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
947 const int16_t **lumSrc, int lumFilterSize,
948 const int16_t *chrFilter, const int16_t **chrUSrc,
949 const int16_t **chrVSrc, int chrFilterSize,
950 const int16_t **alpSrc, uint8_t *dest, int dstW,
951 int y, enum PixelFormat target, int hasAlpha)
955 for (i = 0; i < (dstW >> 1); i++) {
961 const void *r, *g, *b;
963 for (j = 0; j < lumFilterSize; j++) {
964 Y1 += lumSrc[j][i * 2] * lumFilter[j];
965 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
967 for (j = 0; j < chrFilterSize; j++) {
968 U += chrUSrc[j][i] * chrFilter[j];
969 V += chrVSrc[j][i] * chrFilter[j];
978 for (j = 0; j < lumFilterSize; j++) {
979 A1 += alpSrc[j][i * 2 ] * lumFilter[j];
980 A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
984 if ((A1 | A2) & 0x100) {
985 A1 = av_clip_uint8(A1);
986 A2 = av_clip_uint8(A2);
990 r = c->table_rV[V + YUVRGB_TABLE_HEADROOM];
991 g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]);
992 b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
994 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
995 r, g, b, y, target, hasAlpha);
999 static av_always_inline void
1000 yuv2rgb_2_c_template(SwsContext *c, const int16_t *buf[2],
1001 const int16_t *ubuf[2], const int16_t *vbuf[2],
1002 const int16_t *abuf[2], uint8_t *dest, int dstW,
1003 int yalpha, int uvalpha, int y,
1004 enum PixelFormat target, int hasAlpha)
1006 const int16_t *buf0 = buf[0], *buf1 = buf[1],
1007 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
1008 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
1009 *abuf0 = hasAlpha ? abuf[0] : NULL,
1010 *abuf1 = hasAlpha ? abuf[1] : NULL;
1011 int yalpha1 = 4095 - yalpha;
1012 int uvalpha1 = 4095 - uvalpha;
1015 for (i = 0; i < (dstW >> 1); i++) {
1016 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
1017 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
1018 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
1019 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
1021 const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1022 *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1023 *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1025 Y1 = av_clip_uint8(Y1);
1026 Y2 = av_clip_uint8(Y2);
1027 U = av_clip_uint8(U);
1028 V = av_clip_uint8(V);
1031 A1 = (abuf0[i * 2 ] * yalpha1 + abuf1[i * 2 ] * yalpha) >> 19;
1032 A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
1033 A1 = av_clip_uint8(A1);
1034 A2 = av_clip_uint8(A2);
1037 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1038 r, g, b, y, target, hasAlpha);
1042 static av_always_inline void
1043 yuv2rgb_1_c_template(SwsContext *c, const int16_t *buf0,
1044 const int16_t *ubuf[2], const int16_t *vbuf[2],
1045 const int16_t *abuf0, uint8_t *dest, int dstW,
1046 int uvalpha, int y, enum PixelFormat target,
1049 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
1052 if (uvalpha < 2048) {
1053 for (i = 0; i < (dstW >> 1); i++) {
1054 int Y1 = (buf0[i * 2 ] + 64) >> 7;
1055 int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
1056 int U = (ubuf0[i] + 64) >> 7;
1057 int V = (vbuf0[i] + 64) >> 7;
1059 const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1060 *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1061 *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1063 Y1 = av_clip_uint8(Y1);
1064 Y2 = av_clip_uint8(Y2);
1065 U = av_clip_uint8(U);
1066 V = av_clip_uint8(V);
1069 A1 = abuf0[i * 2 ] * 255 + 16384 >> 15;
1070 A2 = abuf0[i * 2 + 1] * 255 + 16384 >> 15;
1071 A1 = av_clip_uint8(A1);
1072 A2 = av_clip_uint8(A2);
1075 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1076 r, g, b, y, target, hasAlpha);
1079 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
1080 for (i = 0; i < (dstW >> 1); i++) {
1081 int Y1 = (buf0[i * 2 ] + 64) >> 7;
1082 int Y2 = (buf0[i * 2 + 1] + 64) >> 7;
1083 int U = (ubuf0[i] + ubuf1[i] + 128) >> 8;
1084 int V = (vbuf0[i] + vbuf1[i] + 128) >> 8;
1086 const void *r = c->table_rV[V + YUVRGB_TABLE_HEADROOM],
1087 *g = (c->table_gU[U + YUVRGB_TABLE_HEADROOM] + c->table_gV[V + YUVRGB_TABLE_HEADROOM]),
1088 *b = c->table_bU[U + YUVRGB_TABLE_HEADROOM];
1090 Y1 = av_clip_uint8(Y1);
1091 Y2 = av_clip_uint8(Y2);
1092 U = av_clip_uint8(U);
1093 V = av_clip_uint8(V);
1096 A1 = (abuf0[i * 2 ] + 64) >> 7;
1097 A2 = (abuf0[i * 2 + 1] + 64) >> 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, PIX_FMT_RGB32_1, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1142 YUV2RGBWRAPPER(yuv2rgb,, 32, PIX_FMT_RGB32, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1144 #if CONFIG_SWSCALE_ALPHA
1145 YUV2RGBWRAPPER(yuv2rgb,, a32_1, PIX_FMT_RGB32_1, 1)
1146 YUV2RGBWRAPPER(yuv2rgb,, a32, PIX_FMT_RGB32, 1)
1148 YUV2RGBWRAPPER(yuv2rgb,, x32_1, PIX_FMT_RGB32_1, 0)
1149 YUV2RGBWRAPPER(yuv2rgb,, x32, PIX_FMT_RGB32, 0)
1151 YUV2RGBWRAPPER(yuv2, rgb, rgb24, PIX_FMT_RGB24, 0)
1152 YUV2RGBWRAPPER(yuv2, rgb, bgr24, PIX_FMT_BGR24, 0)
1153 YUV2RGBWRAPPER(yuv2rgb,, 16, PIX_FMT_RGB565, 0)
1154 YUV2RGBWRAPPER(yuv2rgb,, 15, PIX_FMT_RGB555, 0)
1155 YUV2RGBWRAPPER(yuv2rgb,, 12, PIX_FMT_RGB444, 0)
1156 YUV2RGBWRAPPER(yuv2rgb,, 8, PIX_FMT_RGB8, 0)
1157 YUV2RGBWRAPPER(yuv2rgb,, 4, PIX_FMT_RGB4, 0)
1158 YUV2RGBWRAPPER(yuv2rgb,, 4b, 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 PixelFormat target, int hasAlpha)
1169 int step = (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) ? 3 : 4;
1171 for (i = 0; i < dstW; i++) {
1174 int U = (1<<9)-(128 << 19);
1175 int V = (1<<9)-(128 << 19);
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);
1211 dest[0] = hasAlpha ? A : 255;
1225 dest[3] = hasAlpha ? A : 255;
1228 dest[0] = hasAlpha ? A : 255;
1242 dest[3] = hasAlpha ? A : 255;
1250 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1251 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1252 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1253 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1255 #if CONFIG_SWSCALE_ALPHA
1256 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, 1)
1257 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, 1)
1258 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, 1)
1259 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, 1)
1261 YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, PIX_FMT_BGRA, 0)
1262 YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, PIX_FMT_ABGR, 0)
1263 YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, PIX_FMT_RGBA, 0)
1264 YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, PIX_FMT_ARGB, 0)
1266 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full, PIX_FMT_BGR24, 0)
1267 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full, PIX_FMT_RGB24, 0)
1269 void ff_sws_init_output_funcs(SwsContext *c,
1270 yuv2planar1_fn *yuv2plane1,
1271 yuv2planarX_fn *yuv2planeX,
1272 yuv2interleavedX_fn *yuv2nv12cX,
1273 yuv2packed1_fn *yuv2packed1,
1274 yuv2packed2_fn *yuv2packed2,
1275 yuv2packedX_fn *yuv2packedX)
1277 enum PixelFormat dstFormat = c->dstFormat;
1279 if (is16BPS(dstFormat)) {
1280 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_16BE_c : yuv2planeX_16LE_c;
1281 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_16BE_c : yuv2plane1_16LE_c;
1282 } else if (is9_OR_10BPS(dstFormat)) {
1283 if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
1284 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_9BE_c : yuv2planeX_9LE_c;
1285 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_9BE_c : yuv2plane1_9LE_c;
1287 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_10BE_c : yuv2planeX_10LE_c;
1288 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_10BE_c : yuv2plane1_10LE_c;
1291 *yuv2plane1 = yuv2plane1_8_c;
1292 *yuv2planeX = yuv2planeX_8_c;
1293 if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)
1294 *yuv2nv12cX = yuv2nv12cX_c;
1297 if(c->flags & SWS_FULL_CHR_H_INT) {
1298 switch (dstFormat) {
1301 *yuv2packedX = yuv2rgba32_full_X_c;
1303 #if CONFIG_SWSCALE_ALPHA
1305 *yuv2packedX = yuv2rgba32_full_X_c;
1307 #endif /* CONFIG_SWSCALE_ALPHA */
1309 *yuv2packedX = yuv2rgbx32_full_X_c;
1311 #endif /* !CONFIG_SMALL */
1315 *yuv2packedX = yuv2argb32_full_X_c;
1317 #if CONFIG_SWSCALE_ALPHA
1319 *yuv2packedX = yuv2argb32_full_X_c;
1321 #endif /* CONFIG_SWSCALE_ALPHA */
1323 *yuv2packedX = yuv2xrgb32_full_X_c;
1325 #endif /* !CONFIG_SMALL */
1329 *yuv2packedX = yuv2bgra32_full_X_c;
1331 #if CONFIG_SWSCALE_ALPHA
1333 *yuv2packedX = yuv2bgra32_full_X_c;
1335 #endif /* CONFIG_SWSCALE_ALPHA */
1337 *yuv2packedX = yuv2bgrx32_full_X_c;
1339 #endif /* !CONFIG_SMALL */
1343 *yuv2packedX = yuv2abgr32_full_X_c;
1345 #if CONFIG_SWSCALE_ALPHA
1347 *yuv2packedX = yuv2abgr32_full_X_c;
1349 #endif /* CONFIG_SWSCALE_ALPHA */
1351 *yuv2packedX = yuv2xbgr32_full_X_c;
1353 #endif /* !CONFIG_SMALL */
1356 *yuv2packedX = yuv2rgb24_full_X_c;
1359 *yuv2packedX = yuv2bgr24_full_X_c;
1366 switch (dstFormat) {
1367 case PIX_FMT_RGB48LE:
1368 *yuv2packed1 = yuv2rgb48le_1_c;
1369 *yuv2packed2 = yuv2rgb48le_2_c;
1370 *yuv2packedX = yuv2rgb48le_X_c;
1372 case PIX_FMT_RGB48BE:
1373 *yuv2packed1 = yuv2rgb48be_1_c;
1374 *yuv2packed2 = yuv2rgb48be_2_c;
1375 *yuv2packedX = yuv2rgb48be_X_c;
1377 case PIX_FMT_BGR48LE:
1378 *yuv2packed1 = yuv2bgr48le_1_c;
1379 *yuv2packed2 = yuv2bgr48le_2_c;
1380 *yuv2packedX = yuv2bgr48le_X_c;
1382 case PIX_FMT_BGR48BE:
1383 *yuv2packed1 = yuv2bgr48be_1_c;
1384 *yuv2packed2 = yuv2bgr48be_2_c;
1385 *yuv2packedX = yuv2bgr48be_X_c;
1390 *yuv2packed1 = yuv2rgb32_1_c;
1391 *yuv2packed2 = yuv2rgb32_2_c;
1392 *yuv2packedX = yuv2rgb32_X_c;
1394 #if CONFIG_SWSCALE_ALPHA
1396 *yuv2packed1 = yuv2rgba32_1_c;
1397 *yuv2packed2 = yuv2rgba32_2_c;
1398 *yuv2packedX = yuv2rgba32_X_c;
1400 #endif /* CONFIG_SWSCALE_ALPHA */
1402 *yuv2packed1 = yuv2rgbx32_1_c;
1403 *yuv2packed2 = yuv2rgbx32_2_c;
1404 *yuv2packedX = yuv2rgbx32_X_c;
1406 #endif /* !CONFIG_SMALL */
1408 case PIX_FMT_RGB32_1:
1409 case PIX_FMT_BGR32_1:
1411 *yuv2packed1 = yuv2rgb32_1_1_c;
1412 *yuv2packed2 = yuv2rgb32_1_2_c;
1413 *yuv2packedX = yuv2rgb32_1_X_c;
1415 #if CONFIG_SWSCALE_ALPHA
1417 *yuv2packed1 = yuv2rgba32_1_1_c;
1418 *yuv2packed2 = yuv2rgba32_1_2_c;
1419 *yuv2packedX = yuv2rgba32_1_X_c;
1421 #endif /* CONFIG_SWSCALE_ALPHA */
1423 *yuv2packed1 = yuv2rgbx32_1_1_c;
1424 *yuv2packed2 = yuv2rgbx32_1_2_c;
1425 *yuv2packedX = yuv2rgbx32_1_X_c;
1427 #endif /* !CONFIG_SMALL */
1430 *yuv2packed1 = yuv2rgb24_1_c;
1431 *yuv2packed2 = yuv2rgb24_2_c;
1432 *yuv2packedX = yuv2rgb24_X_c;
1435 *yuv2packed1 = yuv2bgr24_1_c;
1436 *yuv2packed2 = yuv2bgr24_2_c;
1437 *yuv2packedX = yuv2bgr24_X_c;
1439 case PIX_FMT_RGB565LE:
1440 case PIX_FMT_RGB565BE:
1441 case PIX_FMT_BGR565LE:
1442 case PIX_FMT_BGR565BE:
1443 *yuv2packed1 = yuv2rgb16_1_c;
1444 *yuv2packed2 = yuv2rgb16_2_c;
1445 *yuv2packedX = yuv2rgb16_X_c;
1447 case PIX_FMT_RGB555LE:
1448 case PIX_FMT_RGB555BE:
1449 case PIX_FMT_BGR555LE:
1450 case PIX_FMT_BGR555BE:
1451 *yuv2packed1 = yuv2rgb15_1_c;
1452 *yuv2packed2 = yuv2rgb15_2_c;
1453 *yuv2packedX = yuv2rgb15_X_c;
1455 case PIX_FMT_RGB444LE:
1456 case PIX_FMT_RGB444BE:
1457 case PIX_FMT_BGR444LE:
1458 case PIX_FMT_BGR444BE:
1459 *yuv2packed1 = yuv2rgb12_1_c;
1460 *yuv2packed2 = yuv2rgb12_2_c;
1461 *yuv2packedX = yuv2rgb12_X_c;
1465 *yuv2packed1 = yuv2rgb8_1_c;
1466 *yuv2packed2 = yuv2rgb8_2_c;
1467 *yuv2packedX = yuv2rgb8_X_c;
1471 *yuv2packed1 = yuv2rgb4_1_c;
1472 *yuv2packed2 = yuv2rgb4_2_c;
1473 *yuv2packedX = yuv2rgb4_X_c;
1475 case PIX_FMT_RGB4_BYTE:
1476 case PIX_FMT_BGR4_BYTE:
1477 *yuv2packed1 = yuv2rgb4b_1_c;
1478 *yuv2packed2 = yuv2rgb4b_2_c;
1479 *yuv2packedX = yuv2rgb4b_X_c;
1483 switch (dstFormat) {
1484 case PIX_FMT_MONOWHITE:
1485 *yuv2packed1 = yuv2monowhite_1_c;
1486 *yuv2packed2 = yuv2monowhite_2_c;
1487 *yuv2packedX = yuv2monowhite_X_c;
1489 case PIX_FMT_MONOBLACK:
1490 *yuv2packed1 = yuv2monoblack_1_c;
1491 *yuv2packed2 = yuv2monoblack_2_c;
1492 *yuv2packedX = yuv2monoblack_X_c;
1494 case PIX_FMT_YUYV422:
1495 *yuv2packed1 = yuv2yuyv422_1_c;
1496 *yuv2packed2 = yuv2yuyv422_2_c;
1497 *yuv2packedX = yuv2yuyv422_X_c;
1499 case PIX_FMT_UYVY422:
1500 *yuv2packed1 = yuv2uyvy422_1_c;
1501 *yuv2packed2 = yuv2uyvy422_2_c;
1502 *yuv2packedX = yuv2uyvy422_X_c;