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, 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)
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 PixelFormat dstFormat = c->dstFormat;
268 const uint8_t *chrDither = c->chrDither8;
271 if (dstFormat == 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 == 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 PixelFormat target)
317 const uint8_t * const d128=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 PixelFormat target)
355 const int16_t *buf0 = buf[0], *buf1 = buf[1];
356 const uint8_t * const d128 = 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 PixelFormat target)
390 const uint8_t * const d128 = 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, PIX_FMT_MONOWHITE)
445 YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK)
447 #define output_pixels(pos, Y1, U, Y2, V) \
448 if (target == PIX_FMT_YUYV422) { \
449 dest[pos + 0] = Y1; \
451 dest[pos + 2] = Y2; \
455 dest[pos + 1] = Y1; \
457 dest[pos + 3] = Y2; \
460 static av_always_inline void
461 yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
462 const int16_t **lumSrc, int lumFilterSize,
463 const int16_t *chrFilter, const int16_t **chrUSrc,
464 const int16_t **chrVSrc, int chrFilterSize,
465 const int16_t **alpSrc, uint8_t *dest, int dstW,
466 int y, enum PixelFormat target)
470 for (i = 0; i < ((dstW + 1) >> 1); i++) {
477 for (j = 0; j < lumFilterSize; j++) {
478 Y1 += lumSrc[j][i * 2] * lumFilter[j];
479 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
481 for (j = 0; j < chrFilterSize; j++) {
482 U += chrUSrc[j][i] * chrFilter[j];
483 V += chrVSrc[j][i] * chrFilter[j];
489 if ((Y1 | Y2 | U | V) & 0x100) {
490 Y1 = av_clip_uint8(Y1);
491 Y2 = av_clip_uint8(Y2);
492 U = av_clip_uint8(U);
493 V = av_clip_uint8(V);
495 output_pixels(4*i, Y1, U, Y2, V);
499 static av_always_inline void
500 yuv2422_2_c_template(SwsContext *c, const int16_t *buf[2],
501 const int16_t *ubuf[2], const int16_t *vbuf[2],
502 const int16_t *abuf[2], uint8_t *dest, int dstW,
503 int yalpha, int uvalpha, int y,
504 enum PixelFormat target)
506 const int16_t *buf0 = buf[0], *buf1 = buf[1],
507 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
508 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
509 int yalpha1 = 4096 - yalpha;
510 int uvalpha1 = 4096 - uvalpha;
513 for (i = 0; i < ((dstW + 1) >> 1); i++) {
514 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
515 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
516 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
517 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
519 Y1 = av_clip_uint8(Y1);
520 Y2 = av_clip_uint8(Y2);
521 U = av_clip_uint8(U);
522 V = av_clip_uint8(V);
524 output_pixels(i * 4, Y1, U, Y2, V);
528 static av_always_inline void
529 yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
530 const int16_t *ubuf[2], const int16_t *vbuf[2],
531 const int16_t *abuf0, uint8_t *dest, int dstW,
532 int uvalpha, int y, enum PixelFormat target)
534 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
537 if (uvalpha < 2048) {
538 for (i = 0; i < ((dstW + 1) >> 1); i++) {
539 int Y1 = buf0[i * 2] >> 7;
540 int Y2 = buf0[i * 2 + 1] >> 7;
541 int U = ubuf0[i] >> 7;
542 int V = vbuf0[i] >> 7;
544 Y1 = av_clip_uint8(Y1);
545 Y2 = av_clip_uint8(Y2);
546 U = av_clip_uint8(U);
547 V = av_clip_uint8(V);
549 output_pixels(i * 4, Y1, U, Y2, V);
552 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
553 for (i = 0; i < ((dstW + 1) >> 1); i++) {
554 int Y1 = buf0[i * 2] >> 7;
555 int Y2 = buf0[i * 2 + 1] >> 7;
556 int U = (ubuf0[i] + ubuf1[i]) >> 8;
557 int V = (vbuf0[i] + vbuf1[i]) >> 8;
559 Y1 = av_clip_uint8(Y1);
560 Y2 = av_clip_uint8(Y2);
561 U = av_clip_uint8(U);
562 V = av_clip_uint8(V);
564 output_pixels(i * 4, Y1, U, Y2, V);
571 YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422)
572 YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422)
574 #define R_B ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? R : B)
575 #define B_R ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? B : R)
576 #define output_pixel(pos, val) \
577 if (isBE(target)) { \
583 static av_always_inline void
584 yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
585 const int32_t **lumSrc, int lumFilterSize,
586 const int16_t *chrFilter, const int32_t **chrUSrc,
587 const int32_t **chrVSrc, int chrFilterSize,
588 const int32_t **alpSrc, uint16_t *dest, int dstW,
589 int y, enum PixelFormat target)
593 for (i = 0; i < ((dstW + 1) >> 1); i++) {
595 int Y1 = -0x40000000;
596 int Y2 = -0x40000000;
597 int U = -128 << 23; // 19
601 for (j = 0; j < lumFilterSize; j++) {
602 Y1 += lumSrc[j][i * 2] * lumFilter[j];
603 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
605 for (j = 0; j < chrFilterSize; j++) {
606 U += chrUSrc[j][i] * chrFilter[j];
607 V += chrVSrc[j][i] * chrFilter[j];
610 // 8bit: 12+15=27; 16-bit: 12+19=31
618 // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
619 Y1 -= c->yuv2rgb_y_offset;
620 Y2 -= c->yuv2rgb_y_offset;
621 Y1 *= c->yuv2rgb_y_coeff;
622 Y2 *= c->yuv2rgb_y_coeff;
625 // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
627 R = V * c->yuv2rgb_v2r_coeff;
628 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
629 B = U * c->yuv2rgb_u2b_coeff;
631 // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
632 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
633 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
634 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
635 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
636 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
637 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
642 static av_always_inline void
643 yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
644 const int32_t *ubuf[2], const int32_t *vbuf[2],
645 const int32_t *abuf[2], uint16_t *dest, int dstW,
646 int yalpha, int uvalpha, int y,
647 enum PixelFormat target)
649 const int32_t *buf0 = buf[0], *buf1 = buf[1],
650 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
651 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
652 int yalpha1 = 4096 - yalpha;
653 int uvalpha1 = 4096 - uvalpha;
656 for (i = 0; i < ((dstW + 1) >> 1); i++) {
657 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 14;
658 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 14;
659 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha + (-128 << 23)) >> 14;
660 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha + (-128 << 23)) >> 14;
663 Y1 -= c->yuv2rgb_y_offset;
664 Y2 -= c->yuv2rgb_y_offset;
665 Y1 *= c->yuv2rgb_y_coeff;
666 Y2 *= c->yuv2rgb_y_coeff;
670 R = V * c->yuv2rgb_v2r_coeff;
671 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
672 B = U * c->yuv2rgb_u2b_coeff;
674 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
675 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
676 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
677 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
678 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
679 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
684 static av_always_inline void
685 yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
686 const int32_t *ubuf[2], const int32_t *vbuf[2],
687 const int32_t *abuf0, uint16_t *dest, int dstW,
688 int uvalpha, int y, enum PixelFormat target)
690 const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
693 if (uvalpha < 2048) {
694 for (i = 0; i < ((dstW + 1) >> 1); i++) {
695 int Y1 = (buf0[i * 2] ) >> 2;
696 int Y2 = (buf0[i * 2 + 1]) >> 2;
697 int U = (ubuf0[i] + (-128 << 11)) >> 2;
698 int V = (vbuf0[i] + (-128 << 11)) >> 2;
701 Y1 -= c->yuv2rgb_y_offset;
702 Y2 -= c->yuv2rgb_y_offset;
703 Y1 *= c->yuv2rgb_y_coeff;
704 Y2 *= c->yuv2rgb_y_coeff;
708 R = V * c->yuv2rgb_v2r_coeff;
709 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
710 B = U * c->yuv2rgb_u2b_coeff;
712 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
713 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
714 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
715 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
716 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
717 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
721 const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
722 for (i = 0; i < ((dstW + 1) >> 1); i++) {
723 int Y1 = (buf0[i * 2] ) >> 2;
724 int Y2 = (buf0[i * 2 + 1]) >> 2;
725 int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
726 int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
729 Y1 -= c->yuv2rgb_y_offset;
730 Y2 -= c->yuv2rgb_y_offset;
731 Y1 *= c->yuv2rgb_y_coeff;
732 Y2 *= c->yuv2rgb_y_coeff;
736 R = V * c->yuv2rgb_v2r_coeff;
737 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
738 B = U * c->yuv2rgb_u2b_coeff;
740 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
741 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
742 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
743 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
744 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
745 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
755 #define YUV2PACKED16WRAPPER(name, base, ext, fmt) \
756 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
757 const int16_t **_lumSrc, int lumFilterSize, \
758 const int16_t *chrFilter, const int16_t **_chrUSrc, \
759 const int16_t **_chrVSrc, int chrFilterSize, \
760 const int16_t **_alpSrc, uint8_t *_dest, int dstW, \
763 const int32_t **lumSrc = (const int32_t **) _lumSrc, \
764 **chrUSrc = (const int32_t **) _chrUSrc, \
765 **chrVSrc = (const int32_t **) _chrVSrc, \
766 **alpSrc = (const int32_t **) _alpSrc; \
767 uint16_t *dest = (uint16_t *) _dest; \
768 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
769 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
770 alpSrc, dest, dstW, y, fmt); \
773 static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
774 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
775 const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
776 int yalpha, int uvalpha, int y) \
778 const int32_t **buf = (const int32_t **) _buf, \
779 **ubuf = (const int32_t **) _ubuf, \
780 **vbuf = (const int32_t **) _vbuf, \
781 **abuf = (const int32_t **) _abuf; \
782 uint16_t *dest = (uint16_t *) _dest; \
783 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
784 dest, dstW, yalpha, uvalpha, y, fmt); \
787 static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
788 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
789 const int16_t *_abuf0, uint8_t *_dest, int dstW, \
790 int uvalpha, int y) \
792 const int32_t *buf0 = (const int32_t *) _buf0, \
793 **ubuf = (const int32_t **) _ubuf, \
794 **vbuf = (const int32_t **) _vbuf, \
795 *abuf0 = (const int32_t *) _abuf0; \
796 uint16_t *dest = (uint16_t *) _dest; \
797 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
798 dstW, uvalpha, y, fmt); \
801 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE)
802 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE)
803 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE)
804 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE)
807 * Write out 2 RGB pixels in the target pixel format. This function takes a
808 * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
809 * things like endianness conversion and shifting. The caller takes care of
810 * setting the correct offset in these tables from the chroma (U/V) values.
811 * This function then uses the luminance (Y1/Y2) values to write out the
812 * correct RGB values into the destination buffer.
814 static av_always_inline void
815 yuv2rgb_write(uint8_t *_dest, int i, unsigned Y1, unsigned Y2,
816 unsigned A1, unsigned A2,
817 const void *_r, const void *_g, const void *_b, int y,
818 enum PixelFormat target, int hasAlpha)
820 if (target == PIX_FMT_ARGB || target == PIX_FMT_RGBA ||
821 target == PIX_FMT_ABGR || target == PIX_FMT_BGRA) {
822 uint32_t *dest = (uint32_t *) _dest;
823 const uint32_t *r = (const uint32_t *) _r;
824 const uint32_t *g = (const uint32_t *) _g;
825 const uint32_t *b = (const uint32_t *) _b;
828 int sh = hasAlpha ? ((target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24) : 0;
830 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
831 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
834 int sh = (target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24;
836 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
837 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
839 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
840 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
843 } else if (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) {
844 uint8_t *dest = (uint8_t *) _dest;
845 const uint8_t *r = (const uint8_t *) _r;
846 const uint8_t *g = (const uint8_t *) _g;
847 const uint8_t *b = (const uint8_t *) _b;
849 #define r_b ((target == PIX_FMT_RGB24) ? r : b)
850 #define b_r ((target == PIX_FMT_RGB24) ? b : r)
851 dest[i * 6 + 0] = r_b[Y1];
852 dest[i * 6 + 1] = g[Y1];
853 dest[i * 6 + 2] = b_r[Y1];
854 dest[i * 6 + 3] = r_b[Y2];
855 dest[i * 6 + 4] = g[Y2];
856 dest[i * 6 + 5] = b_r[Y2];
859 } else if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565 ||
860 target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555 ||
861 target == PIX_FMT_RGB444 || target == PIX_FMT_BGR444) {
862 uint16_t *dest = (uint16_t *) _dest;
863 const uint16_t *r = (const uint16_t *) _r;
864 const uint16_t *g = (const uint16_t *) _g;
865 const uint16_t *b = (const uint16_t *) _b;
866 int dr1, dg1, db1, dr2, dg2, db2;
868 if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565) {
869 dr1 = dither_2x2_8[ y & 1 ][0];
870 dg1 = dither_2x2_4[ y & 1 ][0];
871 db1 = dither_2x2_8[(y & 1) ^ 1][0];
872 dr2 = dither_2x2_8[ y & 1 ][1];
873 dg2 = dither_2x2_4[ y & 1 ][1];
874 db2 = dither_2x2_8[(y & 1) ^ 1][1];
875 } else if (target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555) {
876 dr1 = dither_2x2_8[ y & 1 ][0];
877 dg1 = dither_2x2_8[ y & 1 ][1];
878 db1 = dither_2x2_8[(y & 1) ^ 1][0];
879 dr2 = dither_2x2_8[ y & 1 ][1];
880 dg2 = dither_2x2_8[ y & 1 ][0];
881 db2 = dither_2x2_8[(y & 1) ^ 1][1];
883 dr1 = dither_4x4_16[ y & 3 ][0];
884 dg1 = dither_4x4_16[ y & 3 ][1];
885 db1 = dither_4x4_16[(y & 3) ^ 3][0];
886 dr2 = dither_4x4_16[ y & 3 ][1];
887 dg2 = dither_4x4_16[ y & 3 ][0];
888 db2 = dither_4x4_16[(y & 3) ^ 3][1];
891 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
892 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
893 } else /* 8/4-bit */ {
894 uint8_t *dest = (uint8_t *) _dest;
895 const uint8_t *r = (const uint8_t *) _r;
896 const uint8_t *g = (const uint8_t *) _g;
897 const uint8_t *b = (const uint8_t *) _b;
898 int dr1, dg1, db1, dr2, dg2, db2;
900 if (target == PIX_FMT_RGB8 || target == PIX_FMT_BGR8) {
901 const uint8_t * const d64 = dither_8x8_73[y & 7];
902 const uint8_t * const d32 = dither_8x8_32[y & 7];
903 dr1 = dg1 = d32[(i * 2 + 0) & 7];
904 db1 = d64[(i * 2 + 0) & 7];
905 dr2 = dg2 = d32[(i * 2 + 1) & 7];
906 db2 = d64[(i * 2 + 1) & 7];
908 const uint8_t * const d64 = dither_8x8_73 [y & 7];
909 const uint8_t * const d128 = dither_8x8_220[y & 7];
910 dr1 = db1 = d128[(i * 2 + 0) & 7];
911 dg1 = d64[(i * 2 + 0) & 7];
912 dr2 = db2 = d128[(i * 2 + 1) & 7];
913 dg2 = d64[(i * 2 + 1) & 7];
916 if (target == PIX_FMT_RGB4 || target == PIX_FMT_BGR4) {
917 dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
918 ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
920 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
921 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
926 static av_always_inline void
927 yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
928 const int16_t **lumSrc, int lumFilterSize,
929 const int16_t *chrFilter, const int16_t **chrUSrc,
930 const int16_t **chrVSrc, int chrFilterSize,
931 const int16_t **alpSrc, uint8_t *dest, int dstW,
932 int y, enum PixelFormat target, int hasAlpha)
936 for (i = 0; i < ((dstW + 1) >> 1); i++) {
942 const void *r, *g, *b;
944 for (j = 0; j < lumFilterSize; j++) {
945 Y1 += lumSrc[j][i * 2] * lumFilter[j];
946 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
948 for (j = 0; j < chrFilterSize; j++) {
949 U += chrUSrc[j][i] * chrFilter[j];
950 V += chrVSrc[j][i] * chrFilter[j];
956 if ((Y1 | Y2 | U | V) & 0x100) {
957 Y1 = av_clip_uint8(Y1);
958 Y2 = av_clip_uint8(Y2);
959 U = av_clip_uint8(U);
960 V = av_clip_uint8(V);
965 for (j = 0; j < lumFilterSize; j++) {
966 A1 += alpSrc[j][i * 2 ] * lumFilter[j];
967 A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
971 if ((A1 | A2) & 0x100) {
972 A1 = av_clip_uint8(A1);
973 A2 = av_clip_uint8(A2);
977 /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
979 g = (c->table_gU[U] + c->table_gV[V]);
982 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
983 r, g, b, y, target, hasAlpha);
987 static av_always_inline void
988 yuv2rgb_2_c_template(SwsContext *c, const int16_t *buf[2],
989 const int16_t *ubuf[2], const int16_t *vbuf[2],
990 const int16_t *abuf[2], uint8_t *dest, int dstW,
991 int yalpha, int uvalpha, int y,
992 enum PixelFormat target, int hasAlpha)
994 const int16_t *buf0 = buf[0], *buf1 = buf[1],
995 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
996 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
997 *abuf0 = hasAlpha ? abuf[0] : NULL,
998 *abuf1 = hasAlpha ? abuf[1] : NULL;
999 int yalpha1 = 4096 - yalpha;
1000 int uvalpha1 = 4096 - uvalpha;
1003 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1004 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
1005 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
1006 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
1007 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
1009 const void *r, *g, *b;
1011 Y1 = av_clip_uint8(Y1);
1012 Y2 = av_clip_uint8(Y2);
1013 U = av_clip_uint8(U);
1014 V = av_clip_uint8(V);
1017 g = (c->table_gU[U] + c->table_gV[V]);
1021 A1 = (abuf0[i * 2 ] * yalpha1 + abuf1[i * 2 ] * yalpha) >> 19;
1022 A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
1023 A1 = av_clip_uint8(A1);
1024 A2 = av_clip_uint8(A2);
1027 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1028 r, g, b, y, target, hasAlpha);
1032 static av_always_inline void
1033 yuv2rgb_1_c_template(SwsContext *c, const int16_t *buf0,
1034 const int16_t *ubuf[2], const int16_t *vbuf[2],
1035 const int16_t *abuf0, uint8_t *dest, int dstW,
1036 int uvalpha, int y, enum PixelFormat target,
1039 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
1042 if (uvalpha < 2048) {
1043 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1044 int Y1 = buf0[i * 2] >> 7;
1045 int Y2 = buf0[i * 2 + 1] >> 7;
1046 int U = ubuf0[i] >> 7;
1047 int V = vbuf0[i] >> 7;
1049 const void *r, *g, *b;
1051 Y1 = av_clip_uint8(Y1);
1052 Y2 = av_clip_uint8(Y2);
1053 U = av_clip_uint8(U);
1054 V = av_clip_uint8(V);
1057 g = (c->table_gU[U] + c->table_gV[V]);
1061 A1 = abuf0[i * 2 ] >> 7;
1062 A2 = abuf0[i * 2 + 1] >> 7;
1063 A1 = av_clip_uint8(A1);
1064 A2 = av_clip_uint8(A2);
1067 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1068 r, g, b, y, target, hasAlpha);
1071 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
1072 for (i = 0; i < ((dstW + 1) >> 1); i++) {
1073 int Y1 = buf0[i * 2] >> 7;
1074 int Y2 = buf0[i * 2 + 1] >> 7;
1075 int U = (ubuf0[i] + ubuf1[i]) >> 8;
1076 int V = (vbuf0[i] + vbuf1[i]) >> 8;
1078 const void *r, *g, *b;
1080 Y1 = av_clip_uint8(Y1);
1081 Y2 = av_clip_uint8(Y2);
1082 U = av_clip_uint8(U);
1083 V = av_clip_uint8(V);
1086 g = (c->table_gU[U] + c->table_gV[V]);
1090 A1 = abuf0[i * 2 ] >> 7;
1091 A2 = abuf0[i * 2 + 1] >> 7;
1092 A1 = av_clip_uint8(A1);
1093 A2 = av_clip_uint8(A2);
1096 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1097 r, g, b, y, target, hasAlpha);
1102 #define YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1103 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
1104 const int16_t **lumSrc, int lumFilterSize, \
1105 const int16_t *chrFilter, const int16_t **chrUSrc, \
1106 const int16_t **chrVSrc, int chrFilterSize, \
1107 const int16_t **alpSrc, uint8_t *dest, int dstW, \
1110 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
1111 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
1112 alpSrc, dest, dstW, y, fmt, hasAlpha); \
1114 #define YUV2RGBWRAPPER(name, base, ext, fmt, hasAlpha) \
1115 YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1116 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
1117 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1118 const int16_t *abuf[2], uint8_t *dest, int dstW, \
1119 int yalpha, int uvalpha, int y) \
1121 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
1122 dest, dstW, yalpha, uvalpha, y, fmt, hasAlpha); \
1125 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
1126 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1127 const int16_t *abuf0, uint8_t *dest, int dstW, \
1128 int uvalpha, int y) \
1130 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
1131 dstW, uvalpha, y, fmt, hasAlpha); \
1135 YUV2RGBWRAPPER(yuv2rgb,, 32_1, PIX_FMT_RGB32_1, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1136 YUV2RGBWRAPPER(yuv2rgb,, 32, PIX_FMT_RGB32, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1138 #if CONFIG_SWSCALE_ALPHA
1139 YUV2RGBWRAPPER(yuv2rgb,, a32_1, PIX_FMT_RGB32_1, 1)
1140 YUV2RGBWRAPPER(yuv2rgb,, a32, PIX_FMT_RGB32, 1)
1142 YUV2RGBWRAPPER(yuv2rgb,, x32_1, PIX_FMT_RGB32_1, 0)
1143 YUV2RGBWRAPPER(yuv2rgb,, x32, PIX_FMT_RGB32, 0)
1145 YUV2RGBWRAPPER(yuv2, rgb, rgb24, PIX_FMT_RGB24, 0)
1146 YUV2RGBWRAPPER(yuv2, rgb, bgr24, PIX_FMT_BGR24, 0)
1147 YUV2RGBWRAPPER(yuv2rgb,, 16, PIX_FMT_RGB565, 0)
1148 YUV2RGBWRAPPER(yuv2rgb,, 15, PIX_FMT_RGB555, 0)
1149 YUV2RGBWRAPPER(yuv2rgb,, 12, PIX_FMT_RGB444, 0)
1150 YUV2RGBWRAPPER(yuv2rgb,, 8, PIX_FMT_RGB8, 0)
1151 YUV2RGBWRAPPER(yuv2rgb,, 4, PIX_FMT_RGB4, 0)
1152 YUV2RGBWRAPPER(yuv2rgb,, 4b, PIX_FMT_RGB4_BYTE, 0)
1154 static av_always_inline void
1155 yuv2rgb_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
1156 const int16_t **lumSrc, int lumFilterSize,
1157 const int16_t *chrFilter, const int16_t **chrUSrc,
1158 const int16_t **chrVSrc, int chrFilterSize,
1159 const int16_t **alpSrc, uint8_t *dest,
1160 int dstW, int y, enum PixelFormat target, int hasAlpha)
1163 int step = (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) ? 3 : 4;
1165 for (i = 0; i < dstW; i++) {
1172 for (j = 0; j < lumFilterSize; j++) {
1173 Y += lumSrc[j][i] * lumFilter[j];
1175 for (j = 0; j < chrFilterSize; j++) {
1176 U += chrUSrc[j][i] * chrFilter[j];
1177 V += chrVSrc[j][i] * chrFilter[j];
1184 for (j = 0; j < lumFilterSize; j++) {
1185 A += alpSrc[j][i] * lumFilter[j];
1189 A = av_clip_uint8(A);
1191 Y -= c->yuv2rgb_y_offset;
1192 Y *= c->yuv2rgb_y_coeff;
1194 R = Y + V*c->yuv2rgb_v2r_coeff;
1195 G = Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;
1196 B = Y + U*c->yuv2rgb_u2b_coeff;
1197 if ((R | G | B) & 0xC0000000) {
1198 R = av_clip_uintp2(R, 30);
1199 G = av_clip_uintp2(G, 30);
1200 B = av_clip_uintp2(B, 30);
1205 dest[0] = hasAlpha ? A : 255;
1219 dest[3] = hasAlpha ? A : 255;
1222 dest[0] = hasAlpha ? A : 255;
1237 dest[3] = hasAlpha ? A : 255;
1245 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1246 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1247 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1248 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1250 #if CONFIG_SWSCALE_ALPHA
1251 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, 1)
1252 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, 1)
1253 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, 1)
1254 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, 1)
1256 YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, PIX_FMT_BGRA, 0)
1257 YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, PIX_FMT_ABGR, 0)
1258 YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, PIX_FMT_RGBA, 0)
1259 YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, PIX_FMT_ARGB, 0)
1261 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full, PIX_FMT_BGR24, 0)
1262 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full, PIX_FMT_RGB24, 0)
1264 av_cold void ff_sws_init_output_funcs(SwsContext *c,
1265 yuv2planar1_fn *yuv2plane1,
1266 yuv2planarX_fn *yuv2planeX,
1267 yuv2interleavedX_fn *yuv2nv12cX,
1268 yuv2packed1_fn *yuv2packed1,
1269 yuv2packed2_fn *yuv2packed2,
1270 yuv2packedX_fn *yuv2packedX)
1272 enum PixelFormat dstFormat = c->dstFormat;
1274 if (is16BPS(dstFormat)) {
1275 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_16BE_c : yuv2planeX_16LE_c;
1276 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_16BE_c : yuv2plane1_16LE_c;
1277 } else if (is9_OR_10BPS(dstFormat)) {
1278 if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
1279 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_9BE_c : yuv2planeX_9LE_c;
1280 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_9BE_c : yuv2plane1_9LE_c;
1282 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_10BE_c : yuv2planeX_10LE_c;
1283 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_10BE_c : yuv2plane1_10LE_c;
1286 *yuv2plane1 = yuv2plane1_8_c;
1287 *yuv2planeX = yuv2planeX_8_c;
1288 if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)
1289 *yuv2nv12cX = yuv2nv12cX_c;
1292 if(c->flags & SWS_FULL_CHR_H_INT) {
1293 switch (dstFormat) {
1296 *yuv2packedX = yuv2rgba32_full_X_c;
1298 #if CONFIG_SWSCALE_ALPHA
1300 *yuv2packedX = yuv2rgba32_full_X_c;
1302 #endif /* CONFIG_SWSCALE_ALPHA */
1304 *yuv2packedX = yuv2rgbx32_full_X_c;
1306 #endif /* !CONFIG_SMALL */
1310 *yuv2packedX = yuv2argb32_full_X_c;
1312 #if CONFIG_SWSCALE_ALPHA
1314 *yuv2packedX = yuv2argb32_full_X_c;
1316 #endif /* CONFIG_SWSCALE_ALPHA */
1318 *yuv2packedX = yuv2xrgb32_full_X_c;
1320 #endif /* !CONFIG_SMALL */
1324 *yuv2packedX = yuv2bgra32_full_X_c;
1326 #if CONFIG_SWSCALE_ALPHA
1328 *yuv2packedX = yuv2bgra32_full_X_c;
1330 #endif /* CONFIG_SWSCALE_ALPHA */
1332 *yuv2packedX = yuv2bgrx32_full_X_c;
1334 #endif /* !CONFIG_SMALL */
1338 *yuv2packedX = yuv2abgr32_full_X_c;
1340 #if CONFIG_SWSCALE_ALPHA
1342 *yuv2packedX = yuv2abgr32_full_X_c;
1344 #endif /* CONFIG_SWSCALE_ALPHA */
1346 *yuv2packedX = yuv2xbgr32_full_X_c;
1348 #endif /* !CONFIG_SMALL */
1351 *yuv2packedX = yuv2rgb24_full_X_c;
1354 *yuv2packedX = yuv2bgr24_full_X_c;
1358 switch (dstFormat) {
1359 case PIX_FMT_RGB48LE:
1360 *yuv2packed1 = yuv2rgb48le_1_c;
1361 *yuv2packed2 = yuv2rgb48le_2_c;
1362 *yuv2packedX = yuv2rgb48le_X_c;
1364 case PIX_FMT_RGB48BE:
1365 *yuv2packed1 = yuv2rgb48be_1_c;
1366 *yuv2packed2 = yuv2rgb48be_2_c;
1367 *yuv2packedX = yuv2rgb48be_X_c;
1369 case PIX_FMT_BGR48LE:
1370 *yuv2packed1 = yuv2bgr48le_1_c;
1371 *yuv2packed2 = yuv2bgr48le_2_c;
1372 *yuv2packedX = yuv2bgr48le_X_c;
1374 case PIX_FMT_BGR48BE:
1375 *yuv2packed1 = yuv2bgr48be_1_c;
1376 *yuv2packed2 = yuv2bgr48be_2_c;
1377 *yuv2packedX = yuv2bgr48be_X_c;
1382 *yuv2packed1 = yuv2rgb32_1_c;
1383 *yuv2packed2 = yuv2rgb32_2_c;
1384 *yuv2packedX = yuv2rgb32_X_c;
1386 #if CONFIG_SWSCALE_ALPHA
1388 *yuv2packed1 = yuv2rgba32_1_c;
1389 *yuv2packed2 = yuv2rgba32_2_c;
1390 *yuv2packedX = yuv2rgba32_X_c;
1392 #endif /* CONFIG_SWSCALE_ALPHA */
1394 *yuv2packed1 = yuv2rgbx32_1_c;
1395 *yuv2packed2 = yuv2rgbx32_2_c;
1396 *yuv2packedX = yuv2rgbx32_X_c;
1398 #endif /* !CONFIG_SMALL */
1400 case PIX_FMT_RGB32_1:
1401 case PIX_FMT_BGR32_1:
1403 *yuv2packed1 = yuv2rgb32_1_1_c;
1404 *yuv2packed2 = yuv2rgb32_1_2_c;
1405 *yuv2packedX = yuv2rgb32_1_X_c;
1407 #if CONFIG_SWSCALE_ALPHA
1409 *yuv2packed1 = yuv2rgba32_1_1_c;
1410 *yuv2packed2 = yuv2rgba32_1_2_c;
1411 *yuv2packedX = yuv2rgba32_1_X_c;
1413 #endif /* CONFIG_SWSCALE_ALPHA */
1415 *yuv2packed1 = yuv2rgbx32_1_1_c;
1416 *yuv2packed2 = yuv2rgbx32_1_2_c;
1417 *yuv2packedX = yuv2rgbx32_1_X_c;
1419 #endif /* !CONFIG_SMALL */
1422 *yuv2packed1 = yuv2rgb24_1_c;
1423 *yuv2packed2 = yuv2rgb24_2_c;
1424 *yuv2packedX = yuv2rgb24_X_c;
1427 *yuv2packed1 = yuv2bgr24_1_c;
1428 *yuv2packed2 = yuv2bgr24_2_c;
1429 *yuv2packedX = yuv2bgr24_X_c;
1431 case PIX_FMT_RGB565LE:
1432 case PIX_FMT_RGB565BE:
1433 case PIX_FMT_BGR565LE:
1434 case PIX_FMT_BGR565BE:
1435 *yuv2packed1 = yuv2rgb16_1_c;
1436 *yuv2packed2 = yuv2rgb16_2_c;
1437 *yuv2packedX = yuv2rgb16_X_c;
1439 case PIX_FMT_RGB555LE:
1440 case PIX_FMT_RGB555BE:
1441 case PIX_FMT_BGR555LE:
1442 case PIX_FMT_BGR555BE:
1443 *yuv2packed1 = yuv2rgb15_1_c;
1444 *yuv2packed2 = yuv2rgb15_2_c;
1445 *yuv2packedX = yuv2rgb15_X_c;
1447 case PIX_FMT_RGB444LE:
1448 case PIX_FMT_RGB444BE:
1449 case PIX_FMT_BGR444LE:
1450 case PIX_FMT_BGR444BE:
1451 *yuv2packed1 = yuv2rgb12_1_c;
1452 *yuv2packed2 = yuv2rgb12_2_c;
1453 *yuv2packedX = yuv2rgb12_X_c;
1457 *yuv2packed1 = yuv2rgb8_1_c;
1458 *yuv2packed2 = yuv2rgb8_2_c;
1459 *yuv2packedX = yuv2rgb8_X_c;
1463 *yuv2packed1 = yuv2rgb4_1_c;
1464 *yuv2packed2 = yuv2rgb4_2_c;
1465 *yuv2packedX = yuv2rgb4_X_c;
1467 case PIX_FMT_RGB4_BYTE:
1468 case PIX_FMT_BGR4_BYTE:
1469 *yuv2packed1 = yuv2rgb4b_1_c;
1470 *yuv2packed2 = yuv2rgb4b_2_c;
1471 *yuv2packedX = yuv2rgb4b_X_c;
1475 switch (dstFormat) {
1476 case PIX_FMT_MONOWHITE:
1477 *yuv2packed1 = yuv2monowhite_1_c;
1478 *yuv2packed2 = yuv2monowhite_2_c;
1479 *yuv2packedX = yuv2monowhite_X_c;
1481 case PIX_FMT_MONOBLACK:
1482 *yuv2packed1 = yuv2monoblack_1_c;
1483 *yuv2packed2 = yuv2monoblack_2_c;
1484 *yuv2packedX = yuv2monoblack_X_c;
1486 case PIX_FMT_YUYV422:
1487 *yuv2packed1 = yuv2yuyv422_1_c;
1488 *yuv2packed2 = yuv2yuyv422_2_c;
1489 *yuv2packedX = yuv2yuyv422_X_c;
1491 case PIX_FMT_UYVY422:
1492 *yuv2packed1 = yuv2uyvy422_1_c;
1493 *yuv2packed2 = yuv2uyvy422_2_c;
1494 *yuv2packedX = yuv2uyvy422_X_c;