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/avutil.h"
28 #include "libavutil/bswap.h"
29 #include "libavutil/cpu.h"
30 #include "libavutil/intreadwrite.h"
31 #include "libavutil/mathematics.h"
32 #include "libavutil/pixdesc.h"
36 #include "swscale_internal.h"
38 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
39 { 1, 3, 1, 3, 1, 3, 1, 3, },
40 { 2, 0, 2, 0, 2, 0, 2, 0, },
43 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
44 { 6, 2, 6, 2, 6, 2, 6, 2, },
45 { 0, 4, 0, 4, 0, 4, 0, 4, },
48 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
49 { 8, 4, 11, 7, 8, 4, 11, 7, },
50 { 2, 14, 1, 13, 2, 14, 1, 13, },
51 { 10, 6, 9, 5, 10, 6, 9, 5, },
52 { 0, 12, 3, 15, 0, 12, 3, 15, },
55 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
56 { 17, 9, 23, 15, 16, 8, 22, 14, },
57 { 5, 29, 3, 27, 4, 28, 2, 26, },
58 { 21, 13, 19, 11, 20, 12, 18, 10, },
59 { 0, 24, 6, 30, 1, 25, 7, 31, },
60 { 16, 8, 22, 14, 17, 9, 23, 15, },
61 { 4, 28, 2, 26, 5, 29, 3, 27, },
62 { 20, 12, 18, 10, 21, 13, 19, 11, },
63 { 1, 25, 7, 31, 0, 24, 6, 30, },
66 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
67 { 0, 55, 14, 68, 3, 58, 17, 72, },
68 { 37, 18, 50, 32, 40, 22, 54, 35, },
69 { 9, 64, 5, 59, 13, 67, 8, 63, },
70 { 46, 27, 41, 23, 49, 31, 44, 26, },
71 { 2, 57, 16, 71, 1, 56, 15, 70, },
72 { 39, 21, 52, 34, 38, 19, 51, 33, },
73 { 11, 66, 7, 62, 10, 65, 6, 60, },
74 { 48, 30, 43, 25, 47, 29, 42, 24, },
78 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
79 {117, 62, 158, 103, 113, 58, 155, 100, },
80 { 34, 199, 21, 186, 31, 196, 17, 182, },
81 {144, 89, 131, 76, 141, 86, 127, 72, },
82 { 0, 165, 41, 206, 10, 175, 52, 217, },
83 {110, 55, 151, 96, 120, 65, 162, 107, },
84 { 28, 193, 14, 179, 38, 203, 24, 189, },
85 {138, 83, 124, 69, 148, 93, 134, 79, },
86 { 7, 172, 48, 213, 3, 168, 45, 210, },
89 // tries to correct a gamma of 1.5
90 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
91 { 0, 143, 18, 200, 2, 156, 25, 215, },
92 { 78, 28, 125, 64, 89, 36, 138, 74, },
93 { 10, 180, 3, 161, 16, 195, 8, 175, },
94 {109, 51, 93, 38, 121, 60, 105, 47, },
95 { 1, 152, 23, 210, 0, 147, 20, 205, },
96 { 85, 33, 134, 71, 81, 30, 130, 67, },
97 { 14, 190, 6, 171, 12, 185, 5, 166, },
98 {117, 57, 101, 44, 113, 54, 97, 41, },
101 // tries to correct a gamma of 2.0
102 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
103 { 0, 124, 8, 193, 0, 140, 12, 213, },
104 { 55, 14, 104, 42, 66, 19, 119, 52, },
105 { 3, 168, 1, 145, 6, 187, 3, 162, },
106 { 86, 31, 70, 21, 99, 39, 82, 28, },
107 { 0, 134, 11, 206, 0, 129, 9, 200, },
108 { 62, 17, 114, 48, 58, 16, 109, 45, },
109 { 5, 181, 2, 157, 4, 175, 1, 151, },
110 { 95, 36, 78, 26, 90, 34, 74, 24, },
113 // tries to correct a gamma of 2.5
114 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
115 { 0, 107, 3, 187, 0, 125, 6, 212, },
116 { 39, 7, 86, 28, 49, 11, 102, 36, },
117 { 1, 158, 0, 131, 3, 180, 1, 151, },
118 { 68, 19, 52, 12, 81, 25, 64, 17, },
119 { 0, 119, 5, 203, 0, 113, 4, 195, },
120 { 45, 9, 96, 33, 42, 8, 91, 30, },
121 { 2, 172, 1, 144, 2, 165, 0, 137, },
122 { 77, 23, 60, 15, 72, 21, 56, 14, },
126 #define output_pixel(pos, val, bias, signedness) \
128 AV_WB16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
130 AV_WL16(pos, bias + av_clip_ ## signedness ## 16(val >> shift)); \
133 static av_always_inline void
134 yuv2plane1_16_c_template(const int32_t *src, uint16_t *dest, int dstW,
135 int big_endian, int output_bits)
138 int shift = 19 - output_bits;
140 for (i = 0; i < dstW; i++) {
141 int val = src[i] + (1 << (shift - 1));
142 output_pixel(&dest[i], val, 0, uint);
146 static av_always_inline void
147 yuv2planeX_16_c_template(const int16_t *filter, int filterSize,
148 const int32_t **src, uint16_t *dest, int dstW,
149 int big_endian, int output_bits)
152 int shift = 15 + 16 - output_bits;
154 for (i = 0; i < dstW; i++) {
155 int val = 1 << (30-output_bits);
158 /* range of val is [0,0x7FFFFFFF], so 31 bits, but with lanczos/spline
159 * filters (or anything with negative coeffs, the range can be slightly
160 * wider in both directions. To account for this overflow, we subtract
161 * a constant so it always fits in the signed range (assuming a
162 * reasonable filterSize), and re-add that at the end. */
164 for (j = 0; j < filterSize; j++)
165 val += src[j][i] * filter[j];
167 output_pixel(&dest[i], val, 0x8000, int);
173 #define output_pixel(pos, val) \
175 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
177 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
180 static av_always_inline void
181 yuv2plane1_10_c_template(const int16_t *src, uint16_t *dest, int dstW,
182 int big_endian, int output_bits)
185 int shift = 15 - output_bits;
187 for (i = 0; i < dstW; i++) {
188 int val = src[i] + (1 << (shift - 1));
189 output_pixel(&dest[i], val);
193 static av_always_inline void
194 yuv2planeX_10_c_template(const int16_t *filter, int filterSize,
195 const int16_t **src, uint16_t *dest, int dstW,
196 int big_endian, int output_bits)
199 int shift = 11 + 16 - output_bits;
201 for (i = 0; i < dstW; i++) {
202 int val = 1 << (26-output_bits);
205 for (j = 0; j < filterSize; j++)
206 val += src[j][i] * filter[j];
208 output_pixel(&dest[i], val);
214 #define yuv2NBPS(bits, BE_LE, is_be, template_size, typeX_t) \
215 static void yuv2plane1_ ## bits ## BE_LE ## _c(const int16_t *src, \
216 uint8_t *dest, int dstW, \
217 const uint8_t *dither, int offset)\
219 yuv2plane1_ ## template_size ## _c_template((const typeX_t *) src, \
220 (uint16_t *) dest, dstW, is_be, bits); \
222 static void yuv2planeX_ ## bits ## BE_LE ## _c(const int16_t *filter, int filterSize, \
223 const int16_t **src, uint8_t *dest, int dstW, \
224 const uint8_t *dither, int offset)\
226 yuv2planeX_## template_size ## _c_template(filter, \
227 filterSize, (const typeX_t **) src, \
228 (uint16_t *) dest, dstW, is_be, bits); \
230 yuv2NBPS( 9, BE, 1, 10, int16_t)
231 yuv2NBPS( 9, LE, 0, 10, int16_t)
232 yuv2NBPS(10, BE, 1, 10, int16_t)
233 yuv2NBPS(10, LE, 0, 10, int16_t)
234 yuv2NBPS(16, BE, 1, 16, int32_t)
235 yuv2NBPS(16, LE, 0, 16, int32_t)
237 static void yuv2planeX_8_c(const int16_t *filter, int filterSize,
238 const int16_t **src, uint8_t *dest, int dstW,
239 const uint8_t *dither, int offset)
242 for (i=0; i<dstW; i++) {
243 int val = dither[(i + offset) & 7] << 12;
245 for (j=0; j<filterSize; j++)
246 val += src[j][i] * filter[j];
248 dest[i]= av_clip_uint8(val>>19);
252 static void yuv2plane1_8_c(const int16_t *src, uint8_t *dest, int dstW,
253 const uint8_t *dither, int offset)
256 for (i=0; i<dstW; i++) {
257 int val = (src[i] + dither[(i + offset) & 7]) >> 7;
258 dest[i]= av_clip_uint8(val);
262 static void yuv2nv12cX_c(SwsContext *c, const int16_t *chrFilter, int chrFilterSize,
263 const int16_t **chrUSrc, const int16_t **chrVSrc,
264 uint8_t *dest, int chrDstW)
266 enum PixelFormat dstFormat = c->dstFormat;
267 const uint8_t *chrDither = c->chrDither8;
270 if (dstFormat == PIX_FMT_NV12)
271 for (i=0; i<chrDstW; i++) {
272 int u = chrDither[i & 7] << 12;
273 int v = chrDither[(i + 3) & 7] << 12;
275 for (j=0; j<chrFilterSize; j++) {
276 u += chrUSrc[j][i] * chrFilter[j];
277 v += chrVSrc[j][i] * chrFilter[j];
280 dest[2*i]= av_clip_uint8(u>>19);
281 dest[2*i+1]= av_clip_uint8(v>>19);
284 for (i=0; i<chrDstW; i++) {
285 int u = chrDither[i & 7] << 12;
286 int v = chrDither[(i + 3) & 7] << 12;
288 for (j=0; j<chrFilterSize; j++) {
289 u += chrUSrc[j][i] * chrFilter[j];
290 v += chrVSrc[j][i] * chrFilter[j];
293 dest[2*i]= av_clip_uint8(v>>19);
294 dest[2*i+1]= av_clip_uint8(u>>19);
298 #define output_pixel(pos, val) \
299 if (target == PIX_FMT_GRAY16BE) { \
305 static av_always_inline void
306 yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter,
307 const int32_t **lumSrc, int lumFilterSize,
308 const int16_t *chrFilter, const int32_t **chrUSrc,
309 const int32_t **chrVSrc, int chrFilterSize,
310 const int32_t **alpSrc, uint16_t *dest, int dstW,
311 int y, enum PixelFormat target)
315 for (i = 0; i < (dstW >> 1); i++) {
317 int Y1 = (1 << 14) - 0x40000000;
318 int Y2 = (1 << 14) - 0x40000000;
320 for (j = 0; j < lumFilterSize; j++) {
321 Y1 += lumSrc[j][i * 2] * lumFilter[j];
322 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
326 Y1 = av_clip_int16(Y1);
327 Y2 = av_clip_int16(Y2);
328 output_pixel(&dest[i * 2 + 0], 0x8000 + Y1);
329 output_pixel(&dest[i * 2 + 1], 0x8000 + Y2);
333 static av_always_inline void
334 yuv2gray16_2_c_template(SwsContext *c, const int32_t *buf[2],
335 const int32_t *ubuf[2], const int32_t *vbuf[2],
336 const int32_t *abuf[2], uint16_t *dest, int dstW,
337 int yalpha, int uvalpha, int y,
338 enum PixelFormat target)
340 int yalpha1 = 4095 - yalpha;
342 const int32_t *buf0 = buf[0], *buf1 = buf[1];
344 for (i = 0; i < (dstW >> 1); i++) {
345 int Y1 = (buf0[i * 2 ] * yalpha1 + buf1[i * 2 ] * yalpha) >> 15;
346 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 15;
348 output_pixel(&dest[i * 2 + 0], Y1);
349 output_pixel(&dest[i * 2 + 1], Y2);
353 static av_always_inline void
354 yuv2gray16_1_c_template(SwsContext *c, const int32_t *buf0,
355 const int32_t *ubuf[2], const int32_t *vbuf[2],
356 const int32_t *abuf0, uint16_t *dest, int dstW,
357 int uvalpha, int y, enum PixelFormat target)
361 for (i = 0; i < (dstW >> 1); i++) {
362 int Y1 = buf0[i * 2 ] << 1;
363 int Y2 = buf0[i * 2 + 1] << 1;
365 output_pixel(&dest[i * 2 + 0], Y1);
366 output_pixel(&dest[i * 2 + 1], Y2);
372 #define YUV2PACKED16WRAPPER(name, base, ext, fmt) \
373 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
374 const int16_t **_lumSrc, int lumFilterSize, \
375 const int16_t *chrFilter, const int16_t **_chrUSrc, \
376 const int16_t **_chrVSrc, int chrFilterSize, \
377 const int16_t **_alpSrc, uint8_t *_dest, int dstW, \
380 const int32_t **lumSrc = (const int32_t **) _lumSrc, \
381 **chrUSrc = (const int32_t **) _chrUSrc, \
382 **chrVSrc = (const int32_t **) _chrVSrc, \
383 **alpSrc = (const int32_t **) _alpSrc; \
384 uint16_t *dest = (uint16_t *) _dest; \
385 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
386 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
387 alpSrc, dest, dstW, y, fmt); \
390 static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
391 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
392 const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
393 int yalpha, int uvalpha, int y) \
395 const int32_t **buf = (const int32_t **) _buf, \
396 **ubuf = (const int32_t **) _ubuf, \
397 **vbuf = (const int32_t **) _vbuf, \
398 **abuf = (const int32_t **) _abuf; \
399 uint16_t *dest = (uint16_t *) _dest; \
400 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
401 dest, dstW, yalpha, uvalpha, y, fmt); \
404 static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
405 const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
406 const int16_t *_abuf0, uint8_t *_dest, int dstW, \
407 int uvalpha, int y) \
409 const int32_t *buf0 = (const int32_t *) _buf0, \
410 **ubuf = (const int32_t **) _ubuf, \
411 **vbuf = (const int32_t **) _vbuf, \
412 *abuf0 = (const int32_t *) _abuf0; \
413 uint16_t *dest = (uint16_t *) _dest; \
414 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
415 dstW, uvalpha, y, fmt); \
418 YUV2PACKED16WRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE)
419 YUV2PACKED16WRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE)
421 #define output_pixel(pos, acc) \
422 if (target == PIX_FMT_MONOBLACK) { \
428 static av_always_inline void
429 yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
430 const int16_t **lumSrc, int lumFilterSize,
431 const int16_t *chrFilter, const int16_t **chrUSrc,
432 const int16_t **chrVSrc, int chrFilterSize,
433 const int16_t **alpSrc, uint8_t *dest, int dstW,
434 int y, enum PixelFormat target)
436 const uint8_t * const d128=dither_8x8_220[y&7];
437 uint8_t *g = c->table_gU[128] + c->table_gV[128];
441 for (i = 0; i < dstW - 1; i += 2) {
446 for (j = 0; j < lumFilterSize; j++) {
447 Y1 += lumSrc[j][i] * lumFilter[j];
448 Y2 += lumSrc[j][i+1] * lumFilter[j];
452 if ((Y1 | Y2) & 0x100) {
453 Y1 = av_clip_uint8(Y1);
454 Y2 = av_clip_uint8(Y2);
456 acc += acc + g[Y1 + d128[(i + 0) & 7]];
457 acc += acc + g[Y2 + d128[(i + 1) & 7]];
459 output_pixel(*dest++, acc);
464 static av_always_inline void
465 yuv2mono_2_c_template(SwsContext *c, const int16_t *buf[2],
466 const int16_t *ubuf[2], const int16_t *vbuf[2],
467 const int16_t *abuf[2], uint8_t *dest, int dstW,
468 int yalpha, int uvalpha, int y,
469 enum PixelFormat target)
471 const int16_t *buf0 = buf[0], *buf1 = buf[1];
472 const uint8_t * const d128 = dither_8x8_220[y & 7];
473 uint8_t *g = c->table_gU[128] + c->table_gV[128];
474 int yalpha1 = 4095 - yalpha;
477 for (i = 0; i < dstW - 7; i += 8) {
478 int acc = g[((buf0[i ] * yalpha1 + buf1[i ] * yalpha) >> 19) + d128[0]];
479 acc += acc + g[((buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19) + d128[1]];
480 acc += acc + g[((buf0[i + 2] * yalpha1 + buf1[i + 2] * yalpha) >> 19) + d128[2]];
481 acc += acc + g[((buf0[i + 3] * yalpha1 + buf1[i + 3] * yalpha) >> 19) + d128[3]];
482 acc += acc + g[((buf0[i + 4] * yalpha1 + buf1[i + 4] * yalpha) >> 19) + d128[4]];
483 acc += acc + g[((buf0[i + 5] * yalpha1 + buf1[i + 5] * yalpha) >> 19) + d128[5]];
484 acc += acc + g[((buf0[i + 6] * yalpha1 + buf1[i + 6] * yalpha) >> 19) + d128[6]];
485 acc += acc + g[((buf0[i + 7] * yalpha1 + buf1[i + 7] * yalpha) >> 19) + d128[7]];
486 output_pixel(*dest++, acc);
490 static av_always_inline void
491 yuv2mono_1_c_template(SwsContext *c, const int16_t *buf0,
492 const int16_t *ubuf[2], const int16_t *vbuf[2],
493 const int16_t *abuf0, uint8_t *dest, int dstW,
494 int uvalpha, int y, enum PixelFormat target)
496 const uint8_t * const d128 = dither_8x8_220[y & 7];
497 uint8_t *g = c->table_gU[128] + c->table_gV[128];
500 for (i = 0; i < dstW - 7; i += 8) {
501 int acc = g[(buf0[i ] >> 7) + d128[0]];
502 acc += acc + g[(buf0[i + 1] >> 7) + d128[1]];
503 acc += acc + g[(buf0[i + 2] >> 7) + d128[2]];
504 acc += acc + g[(buf0[i + 3] >> 7) + d128[3]];
505 acc += acc + g[(buf0[i + 4] >> 7) + d128[4]];
506 acc += acc + g[(buf0[i + 5] >> 7) + d128[5]];
507 acc += acc + g[(buf0[i + 6] >> 7) + d128[6]];
508 acc += acc + g[(buf0[i + 7] >> 7) + d128[7]];
509 output_pixel(*dest++, acc);
515 #define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
516 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
517 const int16_t **lumSrc, int lumFilterSize, \
518 const int16_t *chrFilter, const int16_t **chrUSrc, \
519 const int16_t **chrVSrc, int chrFilterSize, \
520 const int16_t **alpSrc, uint8_t *dest, int dstW, \
523 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
524 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
525 alpSrc, dest, dstW, y, fmt); \
528 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
529 const int16_t *ubuf[2], const int16_t *vbuf[2], \
530 const int16_t *abuf[2], uint8_t *dest, int dstW, \
531 int yalpha, int uvalpha, int y) \
533 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
534 dest, dstW, yalpha, uvalpha, y, fmt); \
537 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
538 const int16_t *ubuf[2], const int16_t *vbuf[2], \
539 const int16_t *abuf0, uint8_t *dest, int dstW, \
540 int uvalpha, int y) \
542 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, \
543 abuf0, dest, dstW, uvalpha, \
547 YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE)
548 YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK)
550 #define output_pixels(pos, Y1, U, Y2, V) \
551 if (target == PIX_FMT_YUYV422) { \
552 dest[pos + 0] = Y1; \
554 dest[pos + 2] = Y2; \
558 dest[pos + 1] = Y1; \
560 dest[pos + 3] = Y2; \
563 static av_always_inline void
564 yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
565 const int16_t **lumSrc, int lumFilterSize,
566 const int16_t *chrFilter, const int16_t **chrUSrc,
567 const int16_t **chrVSrc, int chrFilterSize,
568 const int16_t **alpSrc, uint8_t *dest, int dstW,
569 int y, enum PixelFormat target)
573 for (i = 0; i < (dstW >> 1); i++) {
580 for (j = 0; j < lumFilterSize; j++) {
581 Y1 += lumSrc[j][i * 2] * lumFilter[j];
582 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
584 for (j = 0; j < chrFilterSize; j++) {
585 U += chrUSrc[j][i] * chrFilter[j];
586 V += chrVSrc[j][i] * chrFilter[j];
592 if ((Y1 | Y2 | U | V) & 0x100) {
593 Y1 = av_clip_uint8(Y1);
594 Y2 = av_clip_uint8(Y2);
595 U = av_clip_uint8(U);
596 V = av_clip_uint8(V);
598 output_pixels(4*i, Y1, U, Y2, V);
602 static av_always_inline void
603 yuv2422_2_c_template(SwsContext *c, const int16_t *buf[2],
604 const int16_t *ubuf[2], const int16_t *vbuf[2],
605 const int16_t *abuf[2], uint8_t *dest, int dstW,
606 int yalpha, int uvalpha, int y,
607 enum PixelFormat target)
609 const int16_t *buf0 = buf[0], *buf1 = buf[1],
610 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
611 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
612 int yalpha1 = 4095 - yalpha;
613 int uvalpha1 = 4095 - uvalpha;
616 for (i = 0; i < (dstW >> 1); i++) {
617 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
618 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
619 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
620 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
622 output_pixels(i * 4, Y1, U, Y2, V);
626 static av_always_inline void
627 yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
628 const int16_t *ubuf[2], const int16_t *vbuf[2],
629 const int16_t *abuf0, uint8_t *dest, int dstW,
630 int uvalpha, int y, enum PixelFormat target)
632 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
635 if (uvalpha < 2048) {
636 for (i = 0; i < (dstW >> 1); i++) {
637 int Y1 = buf0[i * 2] >> 7;
638 int Y2 = buf0[i * 2 + 1] >> 7;
639 int U = ubuf0[i] >> 7;
640 int V = vbuf0[i] >> 7;
642 output_pixels(i * 4, Y1, U, Y2, V);
645 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
646 for (i = 0; i < (dstW >> 1); i++) {
647 int Y1 = buf0[i * 2] >> 7;
648 int Y2 = buf0[i * 2 + 1] >> 7;
649 int U = (ubuf0[i] + ubuf1[i]) >> 8;
650 int V = (vbuf0[i] + vbuf1[i]) >> 8;
652 output_pixels(i * 4, Y1, U, Y2, V);
659 YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422)
660 YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422)
662 #define R_B ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? R : B)
663 #define B_R ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? B : R)
664 #define output_pixel(pos, val) \
665 if (isBE(target)) { \
671 static av_always_inline void
672 yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
673 const int32_t **lumSrc, int lumFilterSize,
674 const int16_t *chrFilter, const int32_t **chrUSrc,
675 const int32_t **chrVSrc, int chrFilterSize,
676 const int32_t **alpSrc, uint16_t *dest, int dstW,
677 int y, enum PixelFormat target)
681 for (i = 0; i < (dstW >> 1); i++) {
683 int Y1 = -0x40000000;
684 int Y2 = -0x40000000;
685 int U = -128 << 23; // 19
689 for (j = 0; j < lumFilterSize; j++) {
690 Y1 += lumSrc[j][i * 2] * lumFilter[j];
691 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
693 for (j = 0; j < chrFilterSize; j++) {
694 U += chrUSrc[j][i] * chrFilter[j];
695 V += chrVSrc[j][i] * chrFilter[j];
698 // 8bit: 12+15=27; 16-bit: 12+19=31
706 // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
707 Y1 -= c->yuv2rgb_y_offset;
708 Y2 -= c->yuv2rgb_y_offset;
709 Y1 *= c->yuv2rgb_y_coeff;
710 Y2 *= c->yuv2rgb_y_coeff;
713 // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
715 R = V * c->yuv2rgb_v2r_coeff;
716 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
717 B = U * c->yuv2rgb_u2b_coeff;
719 // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
720 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
721 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
722 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
723 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
724 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
725 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
730 static av_always_inline void
731 yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
732 const int32_t *ubuf[2], const int32_t *vbuf[2],
733 const int32_t *abuf[2], uint16_t *dest, int dstW,
734 int yalpha, int uvalpha, int y,
735 enum PixelFormat target)
737 const int32_t *buf0 = buf[0], *buf1 = buf[1],
738 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
739 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
740 int yalpha1 = 4095 - yalpha;
741 int uvalpha1 = 4095 - uvalpha;
744 for (i = 0; i < (dstW >> 1); i++) {
745 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 14;
746 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 14;
747 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha + (-128 << 23)) >> 14;
748 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha + (-128 << 23)) >> 14;
751 Y1 -= c->yuv2rgb_y_offset;
752 Y2 -= c->yuv2rgb_y_offset;
753 Y1 *= c->yuv2rgb_y_coeff;
754 Y2 *= c->yuv2rgb_y_coeff;
758 R = V * c->yuv2rgb_v2r_coeff;
759 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
760 B = U * c->yuv2rgb_u2b_coeff;
762 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
763 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
764 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
765 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
766 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
767 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
772 static av_always_inline void
773 yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
774 const int32_t *ubuf[2], const int32_t *vbuf[2],
775 const int32_t *abuf0, uint16_t *dest, int dstW,
776 int uvalpha, int y, enum PixelFormat target)
778 const int32_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
781 if (uvalpha < 2048) {
782 for (i = 0; i < (dstW >> 1); i++) {
783 int Y1 = (buf0[i * 2] ) >> 2;
784 int Y2 = (buf0[i * 2 + 1]) >> 2;
785 int U = (ubuf0[i] + (-128 << 11)) >> 2;
786 int V = (vbuf0[i] + (-128 << 11)) >> 2;
789 Y1 -= c->yuv2rgb_y_offset;
790 Y2 -= c->yuv2rgb_y_offset;
791 Y1 *= c->yuv2rgb_y_coeff;
792 Y2 *= c->yuv2rgb_y_coeff;
796 R = V * c->yuv2rgb_v2r_coeff;
797 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
798 B = U * c->yuv2rgb_u2b_coeff;
800 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
801 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
802 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
803 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
804 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
805 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
809 const int32_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
810 for (i = 0; i < (dstW >> 1); i++) {
811 int Y1 = (buf0[i * 2] ) >> 2;
812 int Y2 = (buf0[i * 2 + 1]) >> 2;
813 int U = (ubuf0[i] + ubuf1[i] + (-128 << 12)) >> 3;
814 int V = (vbuf0[i] + vbuf1[i] + (-128 << 12)) >> 3;
817 Y1 -= c->yuv2rgb_y_offset;
818 Y2 -= c->yuv2rgb_y_offset;
819 Y1 *= c->yuv2rgb_y_coeff;
820 Y2 *= c->yuv2rgb_y_coeff;
824 R = V * c->yuv2rgb_v2r_coeff;
825 G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
826 B = U * c->yuv2rgb_u2b_coeff;
828 output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
829 output_pixel(&dest[1], av_clip_uintp2( G + Y1, 30) >> 14);
830 output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
831 output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
832 output_pixel(&dest[4], av_clip_uintp2( G + Y2, 30) >> 14);
833 output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
843 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE)
844 YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE)
845 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE)
846 YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE)
849 * Write out 2 RGB pixels in the target pixel format. This function takes a
850 * R/G/B LUT as generated by ff_yuv2rgb_c_init_tables(), which takes care of
851 * things like endianness conversion and shifting. The caller takes care of
852 * setting the correct offset in these tables from the chroma (U/V) values.
853 * This function then uses the luminance (Y1/Y2) values to write out the
854 * correct RGB values into the destination buffer.
856 static av_always_inline void
857 yuv2rgb_write(uint8_t *_dest, int i, unsigned Y1, unsigned Y2,
858 unsigned A1, unsigned A2,
859 const void *_r, const void *_g, const void *_b, int y,
860 enum PixelFormat target, int hasAlpha)
862 if (target == PIX_FMT_ARGB || target == PIX_FMT_RGBA ||
863 target == PIX_FMT_ABGR || target == PIX_FMT_BGRA) {
864 uint32_t *dest = (uint32_t *) _dest;
865 const uint32_t *r = (const uint32_t *) _r;
866 const uint32_t *g = (const uint32_t *) _g;
867 const uint32_t *b = (const uint32_t *) _b;
870 int sh = hasAlpha ? ((target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24) : 0;
872 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
873 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
876 int sh = (target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24;
878 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
879 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
881 dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
882 dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
885 } else if (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) {
886 uint8_t *dest = (uint8_t *) _dest;
887 const uint8_t *r = (const uint8_t *) _r;
888 const uint8_t *g = (const uint8_t *) _g;
889 const uint8_t *b = (const uint8_t *) _b;
891 #define r_b ((target == PIX_FMT_RGB24) ? r : b)
892 #define b_r ((target == PIX_FMT_RGB24) ? b : r)
893 dest[i * 6 + 0] = r_b[Y1];
894 dest[i * 6 + 1] = g[Y1];
895 dest[i * 6 + 2] = b_r[Y1];
896 dest[i * 6 + 3] = r_b[Y2];
897 dest[i * 6 + 4] = g[Y2];
898 dest[i * 6 + 5] = b_r[Y2];
901 } else if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565 ||
902 target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555 ||
903 target == PIX_FMT_RGB444 || target == PIX_FMT_BGR444) {
904 uint16_t *dest = (uint16_t *) _dest;
905 const uint16_t *r = (const uint16_t *) _r;
906 const uint16_t *g = (const uint16_t *) _g;
907 const uint16_t *b = (const uint16_t *) _b;
908 int dr1, dg1, db1, dr2, dg2, db2;
910 if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565) {
911 dr1 = dither_2x2_8[ y & 1 ][0];
912 dg1 = dither_2x2_4[ y & 1 ][0];
913 db1 = dither_2x2_8[(y & 1) ^ 1][0];
914 dr2 = dither_2x2_8[ y & 1 ][1];
915 dg2 = dither_2x2_4[ y & 1 ][1];
916 db2 = dither_2x2_8[(y & 1) ^ 1][1];
917 } else if (target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555) {
918 dr1 = dither_2x2_8[ y & 1 ][0];
919 dg1 = dither_2x2_8[ y & 1 ][1];
920 db1 = dither_2x2_8[(y & 1) ^ 1][0];
921 dr2 = dither_2x2_8[ y & 1 ][1];
922 dg2 = dither_2x2_8[ y & 1 ][0];
923 db2 = dither_2x2_8[(y & 1) ^ 1][1];
925 dr1 = dither_4x4_16[ y & 3 ][0];
926 dg1 = dither_4x4_16[ y & 3 ][1];
927 db1 = dither_4x4_16[(y & 3) ^ 3][0];
928 dr2 = dither_4x4_16[ y & 3 ][1];
929 dg2 = dither_4x4_16[ y & 3 ][0];
930 db2 = dither_4x4_16[(y & 3) ^ 3][1];
933 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
934 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
935 } else /* 8/4-bit */ {
936 uint8_t *dest = (uint8_t *) _dest;
937 const uint8_t *r = (const uint8_t *) _r;
938 const uint8_t *g = (const uint8_t *) _g;
939 const uint8_t *b = (const uint8_t *) _b;
940 int dr1, dg1, db1, dr2, dg2, db2;
942 if (target == PIX_FMT_RGB8 || target == PIX_FMT_BGR8) {
943 const uint8_t * const d64 = dither_8x8_73[y & 7];
944 const uint8_t * const d32 = dither_8x8_32[y & 7];
945 dr1 = dg1 = d32[(i * 2 + 0) & 7];
946 db1 = d64[(i * 2 + 0) & 7];
947 dr2 = dg2 = d32[(i * 2 + 1) & 7];
948 db2 = d64[(i * 2 + 1) & 7];
950 const uint8_t * const d64 = dither_8x8_73 [y & 7];
951 const uint8_t * const d128 = dither_8x8_220[y & 7];
952 dr1 = db1 = d128[(i * 2 + 0) & 7];
953 dg1 = d64[(i * 2 + 0) & 7];
954 dr2 = db2 = d128[(i * 2 + 1) & 7];
955 dg2 = d64[(i * 2 + 1) & 7];
958 if (target == PIX_FMT_RGB4 || target == PIX_FMT_BGR4) {
959 dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
960 ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
962 dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
963 dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
968 static av_always_inline void
969 yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
970 const int16_t **lumSrc, int lumFilterSize,
971 const int16_t *chrFilter, const int16_t **chrUSrc,
972 const int16_t **chrVSrc, int chrFilterSize,
973 const int16_t **alpSrc, uint8_t *dest, int dstW,
974 int y, enum PixelFormat target, int hasAlpha)
978 for (i = 0; i < (dstW >> 1); i++) {
984 int av_unused A1, A2;
985 const void *r, *g, *b;
987 for (j = 0; j < lumFilterSize; j++) {
988 Y1 += lumSrc[j][i * 2] * lumFilter[j];
989 Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
991 for (j = 0; j < chrFilterSize; j++) {
992 U += chrUSrc[j][i] * chrFilter[j];
993 V += chrVSrc[j][i] * chrFilter[j];
999 if ((Y1 | Y2 | U | V) & 0x100) {
1000 Y1 = av_clip_uint8(Y1);
1001 Y2 = av_clip_uint8(Y2);
1002 U = av_clip_uint8(U);
1003 V = av_clip_uint8(V);
1008 for (j = 0; j < lumFilterSize; j++) {
1009 A1 += alpSrc[j][i * 2 ] * lumFilter[j];
1010 A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
1014 if ((A1 | A2) & 0x100) {
1015 A1 = av_clip_uint8(A1);
1016 A2 = av_clip_uint8(A2);
1020 /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
1022 g = (c->table_gU[U] + c->table_gV[V]);
1025 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1026 r, g, b, y, target, hasAlpha);
1030 static av_always_inline void
1031 yuv2rgb_2_c_template(SwsContext *c, const int16_t *buf[2],
1032 const int16_t *ubuf[2], const int16_t *vbuf[2],
1033 const int16_t *abuf[2], uint8_t *dest, int dstW,
1034 int yalpha, int uvalpha, int y,
1035 enum PixelFormat target, int hasAlpha)
1037 const int16_t *buf0 = buf[0], *buf1 = buf[1],
1038 *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
1039 *vbuf0 = vbuf[0], *vbuf1 = vbuf[1],
1040 *abuf0 = hasAlpha ? abuf[0] : NULL,
1041 *abuf1 = hasAlpha ? abuf[1] : NULL;
1042 int yalpha1 = 4095 - yalpha;
1043 int uvalpha1 = 4095 - uvalpha;
1046 for (i = 0; i < (dstW >> 1); i++) {
1047 int Y1 = (buf0[i * 2] * yalpha1 + buf1[i * 2] * yalpha) >> 19;
1048 int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 19;
1049 int U = (ubuf0[i] * uvalpha1 + ubuf1[i] * uvalpha) >> 19;
1050 int V = (vbuf0[i] * uvalpha1 + vbuf1[i] * uvalpha) >> 19;
1052 const void *r = c->table_rV[V],
1053 *g = (c->table_gU[U] + c->table_gV[V]),
1054 *b = c->table_bU[U];
1057 A1 = (abuf0[i * 2 ] * yalpha1 + abuf1[i * 2 ] * yalpha) >> 19;
1058 A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
1061 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1062 r, g, b, y, target, hasAlpha);
1066 static av_always_inline void
1067 yuv2rgb_1_c_template(SwsContext *c, const int16_t *buf0,
1068 const int16_t *ubuf[2], const int16_t *vbuf[2],
1069 const int16_t *abuf0, uint8_t *dest, int dstW,
1070 int uvalpha, int y, enum PixelFormat target,
1073 const int16_t *ubuf0 = ubuf[0], *vbuf0 = vbuf[0];
1076 if (uvalpha < 2048) {
1077 for (i = 0; i < (dstW >> 1); i++) {
1078 int Y1 = buf0[i * 2] >> 7;
1079 int Y2 = buf0[i * 2 + 1] >> 7;
1080 int U = ubuf0[i] >> 7;
1081 int V = vbuf0[i] >> 7;
1083 const void *r = c->table_rV[V],
1084 *g = (c->table_gU[U] + c->table_gV[V]),
1085 *b = c->table_bU[U];
1088 A1 = abuf0[i * 2 ] >> 7;
1089 A2 = abuf0[i * 2 + 1] >> 7;
1092 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1093 r, g, b, y, target, hasAlpha);
1096 const int16_t *ubuf1 = ubuf[1], *vbuf1 = vbuf[1];
1097 for (i = 0; i < (dstW >> 1); i++) {
1098 int Y1 = buf0[i * 2] >> 7;
1099 int Y2 = buf0[i * 2 + 1] >> 7;
1100 int U = (ubuf0[i] + ubuf1[i]) >> 8;
1101 int V = (vbuf0[i] + vbuf1[i]) >> 8;
1103 const void *r = c->table_rV[V],
1104 *g = (c->table_gU[U] + c->table_gV[V]),
1105 *b = c->table_bU[U];
1108 A1 = abuf0[i * 2 ] >> 7;
1109 A2 = abuf0[i * 2 + 1] >> 7;
1112 yuv2rgb_write(dest, i, Y1, Y2, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
1113 r, g, b, y, target, hasAlpha);
1118 #define YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1119 static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
1120 const int16_t **lumSrc, int lumFilterSize, \
1121 const int16_t *chrFilter, const int16_t **chrUSrc, \
1122 const int16_t **chrVSrc, int chrFilterSize, \
1123 const int16_t **alpSrc, uint8_t *dest, int dstW, \
1126 name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
1127 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
1128 alpSrc, dest, dstW, y, fmt, hasAlpha); \
1130 #define YUV2RGBWRAPPER(name, base, ext, fmt, hasAlpha) \
1131 YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
1132 static void name ## ext ## _2_c(SwsContext *c, const int16_t *buf[2], \
1133 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1134 const int16_t *abuf[2], uint8_t *dest, int dstW, \
1135 int yalpha, int uvalpha, int y) \
1137 name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
1138 dest, dstW, yalpha, uvalpha, y, fmt, hasAlpha); \
1141 static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
1142 const int16_t *ubuf[2], const int16_t *vbuf[2], \
1143 const int16_t *abuf0, uint8_t *dest, int dstW, \
1144 int uvalpha, int y) \
1146 name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
1147 dstW, uvalpha, y, fmt, hasAlpha); \
1151 YUV2RGBWRAPPER(yuv2rgb,, 32_1, PIX_FMT_RGB32_1, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1152 YUV2RGBWRAPPER(yuv2rgb,, 32, PIX_FMT_RGB32, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1154 #if CONFIG_SWSCALE_ALPHA
1155 YUV2RGBWRAPPER(yuv2rgb,, a32_1, PIX_FMT_RGB32_1, 1)
1156 YUV2RGBWRAPPER(yuv2rgb,, a32, PIX_FMT_RGB32, 1)
1158 YUV2RGBWRAPPER(yuv2rgb,, x32_1, PIX_FMT_RGB32_1, 0)
1159 YUV2RGBWRAPPER(yuv2rgb,, x32, PIX_FMT_RGB32, 0)
1161 YUV2RGBWRAPPER(yuv2, rgb, rgb24, PIX_FMT_RGB24, 0)
1162 YUV2RGBWRAPPER(yuv2, rgb, bgr24, PIX_FMT_BGR24, 0)
1163 YUV2RGBWRAPPER(yuv2rgb,, 16, PIX_FMT_RGB565, 0)
1164 YUV2RGBWRAPPER(yuv2rgb,, 15, PIX_FMT_RGB555, 0)
1165 YUV2RGBWRAPPER(yuv2rgb,, 12, PIX_FMT_RGB444, 0)
1166 YUV2RGBWRAPPER(yuv2rgb,, 8, PIX_FMT_RGB8, 0)
1167 YUV2RGBWRAPPER(yuv2rgb,, 4, PIX_FMT_RGB4, 0)
1168 YUV2RGBWRAPPER(yuv2rgb,, 4b, PIX_FMT_RGB4_BYTE, 0)
1170 static av_always_inline void
1171 yuv2rgb_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
1172 const int16_t **lumSrc, int lumFilterSize,
1173 const int16_t *chrFilter, const int16_t **chrUSrc,
1174 const int16_t **chrVSrc, int chrFilterSize,
1175 const int16_t **alpSrc, uint8_t *dest,
1176 int dstW, int y, enum PixelFormat target, int hasAlpha)
1179 int step = (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) ? 3 : 4;
1181 for (i = 0; i < dstW; i++) {
1189 for (j = 0; j < lumFilterSize; j++) {
1190 Y += lumSrc[j][i] * lumFilter[j];
1192 for (j = 0; j < chrFilterSize; j++) {
1193 U += chrUSrc[j][i] * chrFilter[j];
1194 V += chrVSrc[j][i] * chrFilter[j];
1201 for (j = 0; j < lumFilterSize; j++) {
1202 A += alpSrc[j][i] * lumFilter[j];
1206 A = av_clip_uint8(A);
1208 Y -= c->yuv2rgb_y_offset;
1209 Y *= c->yuv2rgb_y_coeff;
1211 R = Y + V*c->yuv2rgb_v2r_coeff;
1212 G = Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;
1213 B = Y + U*c->yuv2rgb_u2b_coeff;
1214 if ((R | G | B) & 0xC0000000) {
1215 R = av_clip_uintp2(R, 30);
1216 G = av_clip_uintp2(G, 30);
1217 B = av_clip_uintp2(B, 30);
1222 dest[0] = hasAlpha ? A : 255;
1236 dest[3] = hasAlpha ? A : 255;
1239 dest[0] = hasAlpha ? A : 255;
1254 dest[3] = hasAlpha ? A : 255;
1262 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1263 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1264 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1265 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
1267 #if CONFIG_SWSCALE_ALPHA
1268 YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA, 1)
1269 YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR, 1)
1270 YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA, 1)
1271 YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB, 1)
1273 YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, PIX_FMT_BGRA, 0)
1274 YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, PIX_FMT_ABGR, 0)
1275 YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, PIX_FMT_RGBA, 0)
1276 YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, PIX_FMT_ARGB, 0)
1278 YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full, PIX_FMT_BGR24, 0)
1279 YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full, PIX_FMT_RGB24, 0)
1281 void ff_sws_init_output_funcs(SwsContext *c,
1282 yuv2planar1_fn *yuv2plane1,
1283 yuv2planarX_fn *yuv2planeX,
1284 yuv2interleavedX_fn *yuv2nv12cX,
1285 yuv2packed1_fn *yuv2packed1,
1286 yuv2packed2_fn *yuv2packed2,
1287 yuv2packedX_fn *yuv2packedX)
1289 enum PixelFormat dstFormat = c->dstFormat;
1291 if (is16BPS(dstFormat)) {
1292 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_16BE_c : yuv2planeX_16LE_c;
1293 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_16BE_c : yuv2plane1_16LE_c;
1294 } else if (is9_OR_10BPS(dstFormat)) {
1295 if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
1296 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_9BE_c : yuv2planeX_9LE_c;
1297 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_9BE_c : yuv2plane1_9LE_c;
1299 *yuv2planeX = isBE(dstFormat) ? yuv2planeX_10BE_c : yuv2planeX_10LE_c;
1300 *yuv2plane1 = isBE(dstFormat) ? yuv2plane1_10BE_c : yuv2plane1_10LE_c;
1303 *yuv2plane1 = yuv2plane1_8_c;
1304 *yuv2planeX = yuv2planeX_8_c;
1305 if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)
1306 *yuv2nv12cX = yuv2nv12cX_c;
1309 if(c->flags & SWS_FULL_CHR_H_INT) {
1310 switch (dstFormat) {
1313 *yuv2packedX = yuv2rgba32_full_X_c;
1315 #if CONFIG_SWSCALE_ALPHA
1317 *yuv2packedX = yuv2rgba32_full_X_c;
1319 #endif /* CONFIG_SWSCALE_ALPHA */
1321 *yuv2packedX = yuv2rgbx32_full_X_c;
1323 #endif /* !CONFIG_SMALL */
1327 *yuv2packedX = yuv2argb32_full_X_c;
1329 #if CONFIG_SWSCALE_ALPHA
1331 *yuv2packedX = yuv2argb32_full_X_c;
1333 #endif /* CONFIG_SWSCALE_ALPHA */
1335 *yuv2packedX = yuv2xrgb32_full_X_c;
1337 #endif /* !CONFIG_SMALL */
1341 *yuv2packedX = yuv2bgra32_full_X_c;
1343 #if CONFIG_SWSCALE_ALPHA
1345 *yuv2packedX = yuv2bgra32_full_X_c;
1347 #endif /* CONFIG_SWSCALE_ALPHA */
1349 *yuv2packedX = yuv2bgrx32_full_X_c;
1351 #endif /* !CONFIG_SMALL */
1355 *yuv2packedX = yuv2abgr32_full_X_c;
1357 #if CONFIG_SWSCALE_ALPHA
1359 *yuv2packedX = yuv2abgr32_full_X_c;
1361 #endif /* CONFIG_SWSCALE_ALPHA */
1363 *yuv2packedX = yuv2xbgr32_full_X_c;
1365 #endif /* !CONFIG_SMALL */
1368 *yuv2packedX = yuv2rgb24_full_X_c;
1371 *yuv2packedX = yuv2bgr24_full_X_c;
1375 switch (dstFormat) {
1376 case PIX_FMT_RGB48LE:
1377 *yuv2packed1 = yuv2rgb48le_1_c;
1378 *yuv2packed2 = yuv2rgb48le_2_c;
1379 *yuv2packedX = yuv2rgb48le_X_c;
1381 case PIX_FMT_RGB48BE:
1382 *yuv2packed1 = yuv2rgb48be_1_c;
1383 *yuv2packed2 = yuv2rgb48be_2_c;
1384 *yuv2packedX = yuv2rgb48be_X_c;
1386 case PIX_FMT_BGR48LE:
1387 *yuv2packed1 = yuv2bgr48le_1_c;
1388 *yuv2packed2 = yuv2bgr48le_2_c;
1389 *yuv2packedX = yuv2bgr48le_X_c;
1391 case PIX_FMT_BGR48BE:
1392 *yuv2packed1 = yuv2bgr48be_1_c;
1393 *yuv2packed2 = yuv2bgr48be_2_c;
1394 *yuv2packedX = yuv2bgr48be_X_c;
1399 *yuv2packed1 = yuv2rgb32_1_c;
1400 *yuv2packed2 = yuv2rgb32_2_c;
1401 *yuv2packedX = yuv2rgb32_X_c;
1403 #if CONFIG_SWSCALE_ALPHA
1405 *yuv2packed1 = yuv2rgba32_1_c;
1406 *yuv2packed2 = yuv2rgba32_2_c;
1407 *yuv2packedX = yuv2rgba32_X_c;
1409 #endif /* CONFIG_SWSCALE_ALPHA */
1411 *yuv2packed1 = yuv2rgbx32_1_c;
1412 *yuv2packed2 = yuv2rgbx32_2_c;
1413 *yuv2packedX = yuv2rgbx32_X_c;
1415 #endif /* !CONFIG_SMALL */
1417 case PIX_FMT_RGB32_1:
1418 case PIX_FMT_BGR32_1:
1420 *yuv2packed1 = yuv2rgb32_1_1_c;
1421 *yuv2packed2 = yuv2rgb32_1_2_c;
1422 *yuv2packedX = yuv2rgb32_1_X_c;
1424 #if CONFIG_SWSCALE_ALPHA
1426 *yuv2packed1 = yuv2rgba32_1_1_c;
1427 *yuv2packed2 = yuv2rgba32_1_2_c;
1428 *yuv2packedX = yuv2rgba32_1_X_c;
1430 #endif /* CONFIG_SWSCALE_ALPHA */
1432 *yuv2packed1 = yuv2rgbx32_1_1_c;
1433 *yuv2packed2 = yuv2rgbx32_1_2_c;
1434 *yuv2packedX = yuv2rgbx32_1_X_c;
1436 #endif /* !CONFIG_SMALL */
1439 *yuv2packed1 = yuv2rgb24_1_c;
1440 *yuv2packed2 = yuv2rgb24_2_c;
1441 *yuv2packedX = yuv2rgb24_X_c;
1444 *yuv2packed1 = yuv2bgr24_1_c;
1445 *yuv2packed2 = yuv2bgr24_2_c;
1446 *yuv2packedX = yuv2bgr24_X_c;
1448 case PIX_FMT_RGB565LE:
1449 case PIX_FMT_RGB565BE:
1450 case PIX_FMT_BGR565LE:
1451 case PIX_FMT_BGR565BE:
1452 *yuv2packed1 = yuv2rgb16_1_c;
1453 *yuv2packed2 = yuv2rgb16_2_c;
1454 *yuv2packedX = yuv2rgb16_X_c;
1456 case PIX_FMT_RGB555LE:
1457 case PIX_FMT_RGB555BE:
1458 case PIX_FMT_BGR555LE:
1459 case PIX_FMT_BGR555BE:
1460 *yuv2packed1 = yuv2rgb15_1_c;
1461 *yuv2packed2 = yuv2rgb15_2_c;
1462 *yuv2packedX = yuv2rgb15_X_c;
1464 case PIX_FMT_RGB444LE:
1465 case PIX_FMT_RGB444BE:
1466 case PIX_FMT_BGR444LE:
1467 case PIX_FMT_BGR444BE:
1468 *yuv2packed1 = yuv2rgb12_1_c;
1469 *yuv2packed2 = yuv2rgb12_2_c;
1470 *yuv2packedX = yuv2rgb12_X_c;
1474 *yuv2packed1 = yuv2rgb8_1_c;
1475 *yuv2packed2 = yuv2rgb8_2_c;
1476 *yuv2packedX = yuv2rgb8_X_c;
1480 *yuv2packed1 = yuv2rgb4_1_c;
1481 *yuv2packed2 = yuv2rgb4_2_c;
1482 *yuv2packedX = yuv2rgb4_X_c;
1484 case PIX_FMT_RGB4_BYTE:
1485 case PIX_FMT_BGR4_BYTE:
1486 *yuv2packed1 = yuv2rgb4b_1_c;
1487 *yuv2packed2 = yuv2rgb4b_2_c;
1488 *yuv2packedX = yuv2rgb4b_X_c;
1492 switch (dstFormat) {
1493 case PIX_FMT_GRAY16BE:
1494 *yuv2packed1 = yuv2gray16BE_1_c;
1495 *yuv2packed2 = yuv2gray16BE_2_c;
1496 *yuv2packedX = yuv2gray16BE_X_c;
1498 case PIX_FMT_GRAY16LE:
1499 *yuv2packed1 = yuv2gray16LE_1_c;
1500 *yuv2packed2 = yuv2gray16LE_2_c;
1501 *yuv2packedX = yuv2gray16LE_X_c;
1503 case PIX_FMT_MONOWHITE:
1504 *yuv2packed1 = yuv2monowhite_1_c;
1505 *yuv2packed2 = yuv2monowhite_2_c;
1506 *yuv2packedX = yuv2monowhite_X_c;
1508 case PIX_FMT_MONOBLACK:
1509 *yuv2packed1 = yuv2monoblack_1_c;
1510 *yuv2packed2 = yuv2monoblack_2_c;
1511 *yuv2packedX = yuv2monoblack_X_c;
1513 case PIX_FMT_YUYV422:
1514 *yuv2packed1 = yuv2yuyv422_1_c;
1515 *yuv2packed2 = yuv2yuyv422_2_c;
1516 *yuv2packedX = yuv2yuyv422_X_c;
1518 case PIX_FMT_UYVY422:
1519 *yuv2packed1 = yuv2uyvy422_1_c;
1520 *yuv2packed2 = yuv2uyvy422_2_c;
1521 *yuv2packedX = yuv2uyvy422_X_c;