ff68f344c3bf0df19cc0565cba0cda8621512905
[libav.git] / libswscale / swscale.c
1 /*
2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 *
20 * the C code (not assembly, mmx, ...) of this file can be used
21 * under the LGPL license too
22 */
23
24 /*
25 supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
26 supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
27 {BGR,RGB}{1,4,8,15,16} support dithering
28
29 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
30 YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
31 x -> x
32 YUV9 -> YV12
33 YUV9/YV12 -> Y800
34 Y800 -> YUV9/YV12
35 BGR24 -> BGR32 & RGB24 -> RGB32
36 BGR32 -> BGR24 & RGB32 -> RGB24
37 BGR15 -> BGR16
38 */
39
40 /*
41 tested special converters (most are tested actually, but I did not write it down ...)
42 YV12 -> BGR16
43 YV12 -> YV12
44 BGR15 -> BGR16
45 BGR16 -> BGR16
46 YVU9 -> YV12
47
48 untested special converters
49 YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
50 YV12/I420 -> YV12/I420
51 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
52 BGR24 -> BGR32 & RGB24 -> RGB32
53 BGR32 -> BGR24 & RGB32 -> RGB24
54 BGR24 -> YV12
55 */
56
57 #include <inttypes.h>
58 #include <string.h>
59 #include <math.h>
60 #include <stdio.h>
61 #include "config.h"
62 #include <assert.h>
63 #include "swscale.h"
64 #include "swscale_internal.h"
65 #include "rgb2rgb.h"
66 #include "libavutil/intreadwrite.h"
67 #include "libavutil/x86_cpu.h"
68 #include "libavutil/avutil.h"
69 #include "libavutil/bswap.h"
70 #include "libavutil/pixdesc.h"
71
72 #undef MOVNTQ
73 #undef PAVGB
74
75 //#undef HAVE_MMX2
76 //#define HAVE_AMD3DNOW
77 //#undef HAVE_MMX
78 //#undef ARCH_X86
79 #define DITHER1XBPP
80
81 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
82
83 #ifdef M_PI
84 #define PI M_PI
85 #else
86 #define PI 3.14159265358979323846
87 #endif
88
89 #define isPacked(x) ( \
90 (x)==PIX_FMT_PAL8 \
91 || (x)==PIX_FMT_YUYV422 \
92 || (x)==PIX_FMT_UYVY422 \
93 || isAnyRGB(x) \
94 )
95 #define usePal(x) (av_pix_fmt_descriptors[x].flags & PIX_FMT_PAL)
96
97 #define RGB2YUV_SHIFT 15
98 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
99 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
100 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
101 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
102 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
103 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
104 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
105 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
106 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
107
108 static const double rgb2yuv_table[8][9]={
109 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
110 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
111 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
112 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
113 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
114 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
115 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
116 {0.701 , 0.087 , 0.212 , -0.384, 0.5 -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
117 };
118
119 /*
120 NOTES
121 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
122
123 TODO
124 more intelligent misalignment avoidance for the horizontal scaler
125 write special vertical cubic upscale version
126 optimize C code (YV12 / minmax)
127 add support for packed pixel YUV input & output
128 add support for Y8 output
129 optimize BGR24 & BGR32
130 add BGR4 output support
131 write special BGR->BGR scaler
132 */
133
134 #if ARCH_X86 && CONFIG_GPL
135 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
136 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
137 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
138 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
139 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
140 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
141 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
142 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
143
144 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
145 0x0103010301030103LL,
146 0x0200020002000200LL,};
147
148 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
149 0x0602060206020602LL,
150 0x0004000400040004LL,};
151
152 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
153 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
154 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
155 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
156 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
157 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
158
159 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
160 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
161 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
162
163 #ifdef FAST_BGR2YV12
164 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
165 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
166 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
167 #else
168 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
169 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
170 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
171 #endif /* FAST_BGR2YV12 */
172 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
173 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
174 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
175
176 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
177 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
178 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
179 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
180 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
181
182 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
183 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
184 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
185 };
186
187 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
188
189 #endif /* ARCH_X86 && CONFIG_GPL */
190
191 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
192 { 1, 3, 1, 3, 1, 3, 1, 3, },
193 { 2, 0, 2, 0, 2, 0, 2, 0, },
194 };
195
196 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
197 { 6, 2, 6, 2, 6, 2, 6, 2, },
198 { 0, 4, 0, 4, 0, 4, 0, 4, },
199 };
200
201 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
202 { 17, 9, 23, 15, 16, 8, 22, 14, },
203 { 5, 29, 3, 27, 4, 28, 2, 26, },
204 { 21, 13, 19, 11, 20, 12, 18, 10, },
205 { 0, 24, 6, 30, 1, 25, 7, 31, },
206 { 16, 8, 22, 14, 17, 9, 23, 15, },
207 { 4, 28, 2, 26, 5, 29, 3, 27, },
208 { 20, 12, 18, 10, 21, 13, 19, 11, },
209 { 1, 25, 7, 31, 0, 24, 6, 30, },
210 };
211
212 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
213 { 0, 55, 14, 68, 3, 58, 17, 72, },
214 { 37, 18, 50, 32, 40, 22, 54, 35, },
215 { 9, 64, 5, 59, 13, 67, 8, 63, },
216 { 46, 27, 41, 23, 49, 31, 44, 26, },
217 { 2, 57, 16, 71, 1, 56, 15, 70, },
218 { 39, 21, 52, 34, 38, 19, 51, 33, },
219 { 11, 66, 7, 62, 10, 65, 6, 60, },
220 { 48, 30, 43, 25, 47, 29, 42, 24, },
221 };
222
223 #if 1
224 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
225 {117, 62, 158, 103, 113, 58, 155, 100, },
226 { 34, 199, 21, 186, 31, 196, 17, 182, },
227 {144, 89, 131, 76, 141, 86, 127, 72, },
228 { 0, 165, 41, 206, 10, 175, 52, 217, },
229 {110, 55, 151, 96, 120, 65, 162, 107, },
230 { 28, 193, 14, 179, 38, 203, 24, 189, },
231 {138, 83, 124, 69, 148, 93, 134, 79, },
232 { 7, 172, 48, 213, 3, 168, 45, 210, },
233 };
234 #elif 1
235 // tries to correct a gamma of 1.5
236 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
237 { 0, 143, 18, 200, 2, 156, 25, 215, },
238 { 78, 28, 125, 64, 89, 36, 138, 74, },
239 { 10, 180, 3, 161, 16, 195, 8, 175, },
240 {109, 51, 93, 38, 121, 60, 105, 47, },
241 { 1, 152, 23, 210, 0, 147, 20, 205, },
242 { 85, 33, 134, 71, 81, 30, 130, 67, },
243 { 14, 190, 6, 171, 12, 185, 5, 166, },
244 {117, 57, 101, 44, 113, 54, 97, 41, },
245 };
246 #elif 1
247 // tries to correct a gamma of 2.0
248 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
249 { 0, 124, 8, 193, 0, 140, 12, 213, },
250 { 55, 14, 104, 42, 66, 19, 119, 52, },
251 { 3, 168, 1, 145, 6, 187, 3, 162, },
252 { 86, 31, 70, 21, 99, 39, 82, 28, },
253 { 0, 134, 11, 206, 0, 129, 9, 200, },
254 { 62, 17, 114, 48, 58, 16, 109, 45, },
255 { 5, 181, 2, 157, 4, 175, 1, 151, },
256 { 95, 36, 78, 26, 90, 34, 74, 24, },
257 };
258 #else
259 // tries to correct a gamma of 2.5
260 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
261 { 0, 107, 3, 187, 0, 125, 6, 212, },
262 { 39, 7, 86, 28, 49, 11, 102, 36, },
263 { 1, 158, 0, 131, 3, 180, 1, 151, },
264 { 68, 19, 52, 12, 81, 25, 64, 17, },
265 { 0, 119, 5, 203, 0, 113, 4, 195, },
266 { 45, 9, 96, 33, 42, 8, 91, 30, },
267 { 2, 172, 1, 144, 2, 165, 0, 137, },
268 { 77, 23, 60, 15, 72, 21, 56, 14, },
269 };
270 #endif
271
272 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
273 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
274 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
275 int dstW, int chrDstW, int big_endian)
276 {
277 //FIXME Optimize (just quickly written not optimized..)
278 int i;
279
280 for (i = 0; i < dstW; i++) {
281 int val = 1 << 10;
282 int j;
283
284 for (j = 0; j < lumFilterSize; j++)
285 val += lumSrc[j][i] * lumFilter[j];
286
287 if (big_endian) {
288 AV_WB16(&dest[i], av_clip_uint16(val >> 11));
289 } else {
290 AV_WL16(&dest[i], av_clip_uint16(val >> 11));
291 }
292 }
293
294 if (uDest) {
295 for (i = 0; i < chrDstW; i++) {
296 int u = 1 << 10;
297 int v = 1 << 10;
298 int j;
299
300 for (j = 0; j < chrFilterSize; j++) {
301 u += chrSrc[j][i ] * chrFilter[j];
302 v += chrSrc[j][i + VOFW] * chrFilter[j];
303 }
304
305 if (big_endian) {
306 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
307 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
308 } else {
309 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
310 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
311 }
312 }
313 }
314
315 if (CONFIG_SWSCALE_ALPHA && aDest) {
316 for (i = 0; i < dstW; i++) {
317 int val = 1 << 10;
318 int j;
319
320 for (j = 0; j < lumFilterSize; j++)
321 val += alpSrc[j][i] * lumFilter[j];
322
323 if (big_endian) {
324 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
325 } else {
326 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
327 }
328 }
329 }
330 }
331
332 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
333 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
334 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
335 enum PixelFormat dstFormat)
336 {
337 if (isBE(dstFormat)) {
338 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
339 chrFilter, chrSrc, chrFilterSize,
340 alpSrc,
341 dest, uDest, vDest, aDest,
342 dstW, chrDstW, 1);
343 } else {
344 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
345 chrFilter, chrSrc, chrFilterSize,
346 alpSrc,
347 dest, uDest, vDest, aDest,
348 dstW, chrDstW, 0);
349 }
350 }
351
352 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
353 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
354 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
355 {
356 //FIXME Optimize (just quickly written not optimized..)
357 int i;
358 for (i=0; i<dstW; i++) {
359 int val=1<<18;
360 int j;
361 for (j=0; j<lumFilterSize; j++)
362 val += lumSrc[j][i] * lumFilter[j];
363
364 dest[i]= av_clip_uint8(val>>19);
365 }
366
367 if (uDest)
368 for (i=0; i<chrDstW; i++) {
369 int u=1<<18;
370 int v=1<<18;
371 int j;
372 for (j=0; j<chrFilterSize; j++) {
373 u += chrSrc[j][i] * chrFilter[j];
374 v += chrSrc[j][i + VOFW] * chrFilter[j];
375 }
376
377 uDest[i]= av_clip_uint8(u>>19);
378 vDest[i]= av_clip_uint8(v>>19);
379 }
380
381 if (CONFIG_SWSCALE_ALPHA && aDest)
382 for (i=0; i<dstW; i++) {
383 int val=1<<18;
384 int j;
385 for (j=0; j<lumFilterSize; j++)
386 val += alpSrc[j][i] * lumFilter[j];
387
388 aDest[i]= av_clip_uint8(val>>19);
389 }
390
391 }
392
393 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
394 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
395 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
396 {
397 //FIXME Optimize (just quickly written not optimized..)
398 int i;
399 for (i=0; i<dstW; i++) {
400 int val=1<<18;
401 int j;
402 for (j=0; j<lumFilterSize; j++)
403 val += lumSrc[j][i] * lumFilter[j];
404
405 dest[i]= av_clip_uint8(val>>19);
406 }
407
408 if (!uDest)
409 return;
410
411 if (dstFormat == PIX_FMT_NV12)
412 for (i=0; i<chrDstW; i++) {
413 int u=1<<18;
414 int v=1<<18;
415 int j;
416 for (j=0; j<chrFilterSize; j++) {
417 u += chrSrc[j][i] * chrFilter[j];
418 v += chrSrc[j][i + VOFW] * chrFilter[j];
419 }
420
421 uDest[2*i]= av_clip_uint8(u>>19);
422 uDest[2*i+1]= av_clip_uint8(v>>19);
423 }
424 else
425 for (i=0; i<chrDstW; i++) {
426 int u=1<<18;
427 int v=1<<18;
428 int j;
429 for (j=0; j<chrFilterSize; j++) {
430 u += chrSrc[j][i] * chrFilter[j];
431 v += chrSrc[j][i + VOFW] * chrFilter[j];
432 }
433
434 uDest[2*i]= av_clip_uint8(v>>19);
435 uDest[2*i+1]= av_clip_uint8(u>>19);
436 }
437 }
438
439 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
440 for (i=0; i<(dstW>>1); i++) {\
441 int j;\
442 int Y1 = 1<<18;\
443 int Y2 = 1<<18;\
444 int U = 1<<18;\
445 int V = 1<<18;\
446 int av_unused A1, A2;\
447 type av_unused *r, *b, *g;\
448 const int i2= 2*i;\
449 \
450 for (j=0; j<lumFilterSize; j++) {\
451 Y1 += lumSrc[j][i2] * lumFilter[j];\
452 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
453 }\
454 for (j=0; j<chrFilterSize; j++) {\
455 U += chrSrc[j][i] * chrFilter[j];\
456 V += chrSrc[j][i+VOFW] * chrFilter[j];\
457 }\
458 Y1>>=19;\
459 Y2>>=19;\
460 U >>=19;\
461 V >>=19;\
462 if (alpha) {\
463 A1 = 1<<18;\
464 A2 = 1<<18;\
465 for (j=0; j<lumFilterSize; j++) {\
466 A1 += alpSrc[j][i2 ] * lumFilter[j];\
467 A2 += alpSrc[j][i2+1] * lumFilter[j];\
468 }\
469 A1>>=19;\
470 A2>>=19;\
471 }
472
473 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
474 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
475 if ((Y1|Y2|U|V)&256) {\
476 if (Y1>255) Y1=255; \
477 else if (Y1<0)Y1=0; \
478 if (Y2>255) Y2=255; \
479 else if (Y2<0)Y2=0; \
480 if (U>255) U=255; \
481 else if (U<0) U=0; \
482 if (V>255) V=255; \
483 else if (V<0) V=0; \
484 }\
485 if (alpha && ((A1|A2)&256)) {\
486 A1=av_clip_uint8(A1);\
487 A2=av_clip_uint8(A2);\
488 }
489
490 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
491 for (i=0; i<dstW; i++) {\
492 int j;\
493 int Y = 0;\
494 int U = -128<<19;\
495 int V = -128<<19;\
496 int av_unused A;\
497 int R,G,B;\
498 \
499 for (j=0; j<lumFilterSize; j++) {\
500 Y += lumSrc[j][i ] * lumFilter[j];\
501 }\
502 for (j=0; j<chrFilterSize; j++) {\
503 U += chrSrc[j][i ] * chrFilter[j];\
504 V += chrSrc[j][i+VOFW] * chrFilter[j];\
505 }\
506 Y >>=10;\
507 U >>=10;\
508 V >>=10;\
509 if (alpha) {\
510 A = rnd;\
511 for (j=0; j<lumFilterSize; j++)\
512 A += alpSrc[j][i ] * lumFilter[j];\
513 A >>=19;\
514 if (A&256)\
515 A = av_clip_uint8(A);\
516 }
517
518 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
519 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
520 Y-= c->yuv2rgb_y_offset;\
521 Y*= c->yuv2rgb_y_coeff;\
522 Y+= rnd;\
523 R= Y + V*c->yuv2rgb_v2r_coeff;\
524 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
525 B= Y + U*c->yuv2rgb_u2b_coeff;\
526 if ((R|G|B)&(0xC0000000)) {\
527 if (R>=(256<<22)) R=(256<<22)-1; \
528 else if (R<0)R=0; \
529 if (G>=(256<<22)) G=(256<<22)-1; \
530 else if (G<0)G=0; \
531 if (B>=(256<<22)) B=(256<<22)-1; \
532 else if (B<0)B=0; \
533 }
534
535 #define YSCALE_YUV_2_GRAY16_C \
536 for (i=0; i<(dstW>>1); i++) {\
537 int j;\
538 int Y1 = 1<<18;\
539 int Y2 = 1<<18;\
540 int U = 1<<18;\
541 int V = 1<<18;\
542 \
543 const int i2= 2*i;\
544 \
545 for (j=0; j<lumFilterSize; j++) {\
546 Y1 += lumSrc[j][i2] * lumFilter[j];\
547 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
548 }\
549 Y1>>=11;\
550 Y2>>=11;\
551 if ((Y1|Y2|U|V)&65536) {\
552 if (Y1>65535) Y1=65535; \
553 else if (Y1<0)Y1=0; \
554 if (Y2>65535) Y2=65535; \
555 else if (Y2<0)Y2=0; \
556 }
557
558 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
559 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
560 r = (type *)c->table_rV[V]; \
561 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
562 b = (type *)c->table_bU[U];
563
564 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
565 for (i=0; i<(dstW>>1); i++) { \
566 const int i2= 2*i; \
567 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
568 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
569 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
570 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
571 type av_unused *r, *b, *g; \
572 int av_unused A1, A2; \
573 if (alpha) {\
574 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
575 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
576 }
577
578 #define YSCALE_YUV_2_GRAY16_2_C \
579 for (i=0; i<(dstW>>1); i++) { \
580 const int i2= 2*i; \
581 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
582 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
583
584 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
585 YSCALE_YUV_2_PACKED2_C(type,alpha)\
586 r = (type *)c->table_rV[V];\
587 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
588 b = (type *)c->table_bU[U];
589
590 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
591 for (i=0; i<(dstW>>1); i++) {\
592 const int i2= 2*i;\
593 int Y1= buf0[i2 ]>>7;\
594 int Y2= buf0[i2+1]>>7;\
595 int U= (uvbuf1[i ])>>7;\
596 int V= (uvbuf1[i+VOFW])>>7;\
597 type av_unused *r, *b, *g;\
598 int av_unused A1, A2;\
599 if (alpha) {\
600 A1= abuf0[i2 ]>>7;\
601 A2= abuf0[i2+1]>>7;\
602 }
603
604 #define YSCALE_YUV_2_GRAY16_1_C \
605 for (i=0; i<(dstW>>1); i++) {\
606 const int i2= 2*i;\
607 int Y1= buf0[i2 ]<<1;\
608 int Y2= buf0[i2+1]<<1;
609
610 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
611 YSCALE_YUV_2_PACKED1_C(type,alpha)\
612 r = (type *)c->table_rV[V];\
613 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
614 b = (type *)c->table_bU[U];
615
616 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
617 for (i=0; i<(dstW>>1); i++) {\
618 const int i2= 2*i;\
619 int Y1= buf0[i2 ]>>7;\
620 int Y2= buf0[i2+1]>>7;\
621 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
622 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
623 type av_unused *r, *b, *g;\
624 int av_unused A1, A2;\
625 if (alpha) {\
626 A1= abuf0[i2 ]>>7;\
627 A2= abuf0[i2+1]>>7;\
628 }
629
630 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
631 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
632 r = (type *)c->table_rV[V];\
633 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
634 b = (type *)c->table_bU[U];
635
636 #define YSCALE_YUV_2_MONO2_C \
637 const uint8_t * const d128=dither_8x8_220[y&7];\
638 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
639 for (i=0; i<dstW-7; i+=8) {\
640 int acc;\
641 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
642 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
643 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
644 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
645 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
646 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
647 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
648 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
649 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
650 dest++;\
651 }
652
653 #define YSCALE_YUV_2_MONOX_C \
654 const uint8_t * const d128=dither_8x8_220[y&7];\
655 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
656 int acc=0;\
657 for (i=0; i<dstW-1; i+=2) {\
658 int j;\
659 int Y1=1<<18;\
660 int Y2=1<<18;\
661 \
662 for (j=0; j<lumFilterSize; j++) {\
663 Y1 += lumSrc[j][i] * lumFilter[j];\
664 Y2 += lumSrc[j][i+1] * lumFilter[j];\
665 }\
666 Y1>>=19;\
667 Y2>>=19;\
668 if ((Y1|Y2)&256) {\
669 if (Y1>255) Y1=255;\
670 else if (Y1<0)Y1=0;\
671 if (Y2>255) Y2=255;\
672 else if (Y2<0)Y2=0;\
673 }\
674 acc+= acc + g[Y1+d128[(i+0)&7]];\
675 acc+= acc + g[Y2+d128[(i+1)&7]];\
676 if ((i&7)==6) {\
677 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
678 dest++;\
679 }\
680 }
681
682 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
683 switch(c->dstFormat) {\
684 case PIX_FMT_RGB48BE:\
685 case PIX_FMT_RGB48LE:\
686 func(uint8_t,0)\
687 ((uint8_t*)dest)[ 0]= r[Y1];\
688 ((uint8_t*)dest)[ 1]= r[Y1];\
689 ((uint8_t*)dest)[ 2]= g[Y1];\
690 ((uint8_t*)dest)[ 3]= g[Y1];\
691 ((uint8_t*)dest)[ 4]= b[Y1];\
692 ((uint8_t*)dest)[ 5]= b[Y1];\
693 ((uint8_t*)dest)[ 6]= r[Y2];\
694 ((uint8_t*)dest)[ 7]= r[Y2];\
695 ((uint8_t*)dest)[ 8]= g[Y2];\
696 ((uint8_t*)dest)[ 9]= g[Y2];\
697 ((uint8_t*)dest)[10]= b[Y2];\
698 ((uint8_t*)dest)[11]= b[Y2];\
699 dest+=12;\
700 }\
701 break;\
702 case PIX_FMT_RGBA:\
703 case PIX_FMT_BGRA:\
704 if (CONFIG_SMALL) {\
705 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
706 func(uint32_t,needAlpha)\
707 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
708 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
709 }\
710 } else {\
711 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
712 func(uint32_t,1)\
713 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
714 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
715 }\
716 } else {\
717 func(uint32_t,0)\
718 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
719 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
720 }\
721 }\
722 }\
723 break;\
724 case PIX_FMT_ARGB:\
725 case PIX_FMT_ABGR:\
726 if (CONFIG_SMALL) {\
727 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
728 func(uint32_t,needAlpha)\
729 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
730 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
731 }\
732 } else {\
733 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
734 func(uint32_t,1)\
735 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
736 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
737 }\
738 } else {\
739 func(uint32_t,0)\
740 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
741 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
742 }\
743 }\
744 } \
745 break;\
746 case PIX_FMT_RGB24:\
747 func(uint8_t,0)\
748 ((uint8_t*)dest)[0]= r[Y1];\
749 ((uint8_t*)dest)[1]= g[Y1];\
750 ((uint8_t*)dest)[2]= b[Y1];\
751 ((uint8_t*)dest)[3]= r[Y2];\
752 ((uint8_t*)dest)[4]= g[Y2];\
753 ((uint8_t*)dest)[5]= b[Y2];\
754 dest+=6;\
755 }\
756 break;\
757 case PIX_FMT_BGR24:\
758 func(uint8_t,0)\
759 ((uint8_t*)dest)[0]= b[Y1];\
760 ((uint8_t*)dest)[1]= g[Y1];\
761 ((uint8_t*)dest)[2]= r[Y1];\
762 ((uint8_t*)dest)[3]= b[Y2];\
763 ((uint8_t*)dest)[4]= g[Y2];\
764 ((uint8_t*)dest)[5]= r[Y2];\
765 dest+=6;\
766 }\
767 break;\
768 case PIX_FMT_RGB565:\
769 case PIX_FMT_BGR565:\
770 {\
771 const int dr1= dither_2x2_8[y&1 ][0];\
772 const int dg1= dither_2x2_4[y&1 ][0];\
773 const int db1= dither_2x2_8[(y&1)^1][0];\
774 const int dr2= dither_2x2_8[y&1 ][1];\
775 const int dg2= dither_2x2_4[y&1 ][1];\
776 const int db2= dither_2x2_8[(y&1)^1][1];\
777 func(uint16_t,0)\
778 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
779 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
780 }\
781 }\
782 break;\
783 case PIX_FMT_RGB555:\
784 case PIX_FMT_BGR555:\
785 {\
786 const int dr1= dither_2x2_8[y&1 ][0];\
787 const int dg1= dither_2x2_8[y&1 ][1];\
788 const int db1= dither_2x2_8[(y&1)^1][0];\
789 const int dr2= dither_2x2_8[y&1 ][1];\
790 const int dg2= dither_2x2_8[y&1 ][0];\
791 const int db2= dither_2x2_8[(y&1)^1][1];\
792 func(uint16_t,0)\
793 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
794 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
795 }\
796 }\
797 break;\
798 case PIX_FMT_RGB8:\
799 case PIX_FMT_BGR8:\
800 {\
801 const uint8_t * const d64= dither_8x8_73[y&7];\
802 const uint8_t * const d32= dither_8x8_32[y&7];\
803 func(uint8_t,0)\
804 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
805 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
806 }\
807 }\
808 break;\
809 case PIX_FMT_RGB4:\
810 case PIX_FMT_BGR4:\
811 {\
812 const uint8_t * const d64= dither_8x8_73 [y&7];\
813 const uint8_t * const d128=dither_8x8_220[y&7];\
814 func(uint8_t,0)\
815 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
816 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
817 }\
818 }\
819 break;\
820 case PIX_FMT_RGB4_BYTE:\
821 case PIX_FMT_BGR4_BYTE:\
822 {\
823 const uint8_t * const d64= dither_8x8_73 [y&7];\
824 const uint8_t * const d128=dither_8x8_220[y&7];\
825 func(uint8_t,0)\
826 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
827 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
828 }\
829 }\
830 break;\
831 case PIX_FMT_MONOBLACK:\
832 case PIX_FMT_MONOWHITE:\
833 {\
834 func_monoblack\
835 }\
836 break;\
837 case PIX_FMT_YUYV422:\
838 func2\
839 ((uint8_t*)dest)[2*i2+0]= Y1;\
840 ((uint8_t*)dest)[2*i2+1]= U;\
841 ((uint8_t*)dest)[2*i2+2]= Y2;\
842 ((uint8_t*)dest)[2*i2+3]= V;\
843 } \
844 break;\
845 case PIX_FMT_UYVY422:\
846 func2\
847 ((uint8_t*)dest)[2*i2+0]= U;\
848 ((uint8_t*)dest)[2*i2+1]= Y1;\
849 ((uint8_t*)dest)[2*i2+2]= V;\
850 ((uint8_t*)dest)[2*i2+3]= Y2;\
851 } \
852 break;\
853 case PIX_FMT_GRAY16BE:\
854 func_g16\
855 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
856 ((uint8_t*)dest)[2*i2+1]= Y1;\
857 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
858 ((uint8_t*)dest)[2*i2+3]= Y2;\
859 } \
860 break;\
861 case PIX_FMT_GRAY16LE:\
862 func_g16\
863 ((uint8_t*)dest)[2*i2+0]= Y1;\
864 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
865 ((uint8_t*)dest)[2*i2+2]= Y2;\
866 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
867 } \
868 break;\
869 }
870
871 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
872 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
873 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
874 {
875 int i;
876 YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
877 }
878
879 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
880 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
881 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
882 {
883 int i;
884 int step= c->dstFormatBpp/8;
885 int aidx= 3;
886
887 switch(c->dstFormat) {
888 case PIX_FMT_ARGB:
889 dest++;
890 aidx= 0;
891 case PIX_FMT_RGB24:
892 aidx--;
893 case PIX_FMT_RGBA:
894 if (CONFIG_SMALL) {
895 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
896 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
897 dest[aidx]= needAlpha ? A : 255;
898 dest[0]= R>>22;
899 dest[1]= G>>22;
900 dest[2]= B>>22;
901 dest+= step;
902 }
903 } else {
904 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
905 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
906 dest[aidx]= A;
907 dest[0]= R>>22;
908 dest[1]= G>>22;
909 dest[2]= B>>22;
910 dest+= step;
911 }
912 } else {
913 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
914 dest[aidx]= 255;
915 dest[0]= R>>22;
916 dest[1]= G>>22;
917 dest[2]= B>>22;
918 dest+= step;
919 }
920 }
921 }
922 break;
923 case PIX_FMT_ABGR:
924 dest++;
925 aidx= 0;
926 case PIX_FMT_BGR24:
927 aidx--;
928 case PIX_FMT_BGRA:
929 if (CONFIG_SMALL) {
930 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
931 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
932 dest[aidx]= needAlpha ? A : 255;
933 dest[0]= B>>22;
934 dest[1]= G>>22;
935 dest[2]= R>>22;
936 dest+= step;
937 }
938 } else {
939 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
940 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
941 dest[aidx]= A;
942 dest[0]= B>>22;
943 dest[1]= G>>22;
944 dest[2]= R>>22;
945 dest+= step;
946 }
947 } else {
948 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
949 dest[aidx]= 255;
950 dest[0]= B>>22;
951 dest[1]= G>>22;
952 dest[2]= R>>22;
953 dest+= step;
954 }
955 }
956 }
957 break;
958 default:
959 assert(0);
960 }
961 }
962
963 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
964 {
965 int i;
966 uint8_t *ptr = plane + stride*y;
967 for (i=0; i<height; i++) {
968 memset(ptr, val, width);
969 ptr += stride;
970 }
971 }
972
973 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
974 uint32_t *unused)
975 {
976 int i;
977 for (i = 0; i < width; i++) {
978 int r = src[i*6+0];
979 int g = src[i*6+2];
980 int b = src[i*6+4];
981
982 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
983 }
984 }
985
986 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
987 const uint8_t *src1, const uint8_t *src2,
988 int width, uint32_t *unused)
989 {
990 int i;
991 assert(src1==src2);
992 for (i = 0; i < width; i++) {
993 int r = src1[6*i + 0];
994 int g = src1[6*i + 2];
995 int b = src1[6*i + 4];
996
997 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
998 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
999 }
1000 }
1001
1002 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1003 const uint8_t *src1, const uint8_t *src2,
1004 int width, uint32_t *unused)
1005 {
1006 int i;
1007 assert(src1==src2);
1008 for (i = 0; i < width; i++) {
1009 int r= src1[12*i + 0] + src1[12*i + 6];
1010 int g= src1[12*i + 2] + src1[12*i + 8];
1011 int b= src1[12*i + 4] + src1[12*i + 10];
1012
1013 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1014 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1015 }
1016 }
1017
1018 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1019 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1020 {\
1021 int i;\
1022 for (i=0; i<width; i++) {\
1023 int b= (((const type*)src)[i]>>shb)&maskb;\
1024 int g= (((const type*)src)[i]>>shg)&maskg;\
1025 int r= (((const type*)src)[i]>>shr)&maskr;\
1026 \
1027 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1028 }\
1029 }
1030
1031 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1032 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1033 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1034 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1035 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1036 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1037
1038 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1039 {
1040 int i;
1041 for (i=0; i<width; i++) {
1042 dst[i]= src[4*i];
1043 }
1044 }
1045
1046 #define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1047 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1048 {\
1049 int i;\
1050 for (i=0; i<width; i++) {\
1051 int b= (((const type*)src)[i]&maskb)>>shb;\
1052 int g= (((const type*)src)[i]&maskg)>>shg;\
1053 int r= (((const type*)src)[i]&maskr)>>shr;\
1054 \
1055 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1056 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1057 }\
1058 }\
1059 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1060 {\
1061 int i;\
1062 for (i=0; i<width; i++) {\
1063 int pix0= ((const type*)src)[2*i+0];\
1064 int pix1= ((const type*)src)[2*i+1];\
1065 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1066 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1067 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1068 g&= maskg|(2*maskg);\
1069 \
1070 g>>=shg;\
1071 \
1072 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1073 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1074 }\
1075 }
1076
1077 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1078 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1079 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1080 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1081 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1082 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1083
1084 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1085 {
1086 int i;
1087 for (i=0; i<width; i++) {
1088 int d= src[i];
1089
1090 dst[i]= pal[d] & 0xFF;
1091 }
1092 }
1093
1094 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1095 const uint8_t *src1, const uint8_t *src2,
1096 long width, uint32_t *pal)
1097 {
1098 int i;
1099 assert(src1 == src2);
1100 for (i=0; i<width; i++) {
1101 int p= pal[src1[i]];
1102
1103 dstU[i]= p>>8;
1104 dstV[i]= p>>16;
1105 }
1106 }
1107
1108 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1109 {
1110 int i, j;
1111 for (i=0; i<width/8; i++) {
1112 int d= ~src[i];
1113 for(j=0; j<8; j++)
1114 dst[8*i+j]= ((d>>(7-j))&1)*255;
1115 }
1116 }
1117
1118 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1119 {
1120 int i, j;
1121 for (i=0; i<width/8; i++) {
1122 int d= src[i];
1123 for(j=0; j<8; j++)
1124 dst[8*i+j]= ((d>>(7-j))&1)*255;
1125 }
1126 }
1127
1128 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1129 //Plain C versions
1130 #if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
1131 #define COMPILE_C
1132 #endif
1133
1134 #if ARCH_PPC
1135 #if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
1136 #define COMPILE_ALTIVEC
1137 #endif
1138 #endif //ARCH_PPC
1139
1140 #if ARCH_X86
1141
1142 #if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1143 #define COMPILE_MMX
1144 #endif
1145
1146 #if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1147 #define COMPILE_MMX2
1148 #endif
1149
1150 #if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
1151 #define COMPILE_3DNOW
1152 #endif
1153 #endif //ARCH_X86
1154
1155 #define COMPILE_TEMPLATE_MMX 0
1156 #define COMPILE_TEMPLATE_MMX2 0
1157 #define COMPILE_TEMPLATE_AMD3DNOW 0
1158 #define COMPILE_TEMPLATE_ALTIVEC 0
1159
1160 #ifdef COMPILE_C
1161 #define RENAME(a) a ## _C
1162 #include "swscale_template.c"
1163 #endif
1164
1165 #ifdef COMPILE_ALTIVEC
1166 #undef RENAME
1167 #undef COMPILE_TEMPLATE_ALTIVEC
1168 #define COMPILE_TEMPLATE_ALTIVEC 1
1169 #define RENAME(a) a ## _altivec
1170 #include "swscale_template.c"
1171 #endif
1172
1173 #if ARCH_X86
1174
1175 //MMX versions
1176 #ifdef COMPILE_MMX
1177 #undef RENAME
1178 #undef COMPILE_TEMPLATE_MMX
1179 #undef COMPILE_TEMPLATE_MMX2
1180 #undef COMPILE_TEMPLATE_AMD3DNOW
1181 #define COMPILE_TEMPLATE_MMX 1
1182 #define COMPILE_TEMPLATE_MMX2 0
1183 #define COMPILE_TEMPLATE_AMD3DNOW 0
1184 #define RENAME(a) a ## _MMX
1185 #include "swscale_template.c"
1186 #endif
1187
1188 //MMX2 versions
1189 #ifdef COMPILE_MMX2
1190 #undef RENAME
1191 #undef COMPILE_TEMPLATE_MMX
1192 #undef COMPILE_TEMPLATE_MMX2
1193 #undef COMPILE_TEMPLATE_AMD3DNOW
1194 #define COMPILE_TEMPLATE_MMX 1
1195 #define COMPILE_TEMPLATE_MMX2 1
1196 #define COMPILE_TEMPLATE_AMD3DNOW 0
1197 #define RENAME(a) a ## _MMX2
1198 #include "swscale_template.c"
1199 #endif
1200
1201 //3DNOW versions
1202 #ifdef COMPILE_3DNOW
1203 #undef RENAME
1204 #undef COMPILE_TEMPLATE_MMX
1205 #undef COMPILE_TEMPLATE_MMX2
1206 #undef COMPILE_TEMPLATE_AMD3DNOW
1207 #define COMPILE_TEMPLATE_MMX 1
1208 #define COMPILE_TEMPLATE_MMX2 0
1209 #define COMPILE_TEMPLATE_AMD3DNOW 1
1210 #define RENAME(a) a ## _3DNow
1211 #include "swscale_template.c"
1212 #endif
1213
1214 #endif //ARCH_X86
1215
1216 SwsFunc ff_getSwsFunc(SwsContext *c)
1217 {
1218 #if CONFIG_RUNTIME_CPUDETECT
1219 int flags = c->flags;
1220
1221 #if ARCH_X86 && CONFIG_GPL
1222 // ordered per speed fastest first
1223 if (flags & SWS_CPU_CAPS_MMX2) {
1224 sws_init_swScale_MMX2(c);
1225 return swScale_MMX2;
1226 } else if (flags & SWS_CPU_CAPS_3DNOW) {
1227 sws_init_swScale_3DNow(c);
1228 return swScale_3DNow;
1229 } else if (flags & SWS_CPU_CAPS_MMX) {
1230 sws_init_swScale_MMX(c);
1231 return swScale_MMX;
1232 } else {
1233 sws_init_swScale_C(c);
1234 return swScale_C;
1235 }
1236
1237 #else
1238 #if ARCH_PPC
1239 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1240 sws_init_swScale_altivec(c);
1241 return swScale_altivec;
1242 } else {
1243 sws_init_swScale_C(c);
1244 return swScale_C;
1245 }
1246 #endif
1247 sws_init_swScale_C(c);
1248 return swScale_C;
1249 #endif /* ARCH_X86 && CONFIG_GPL */
1250 #else //CONFIG_RUNTIME_CPUDETECT
1251 #if COMPILE_TEMPLATE_MMX2
1252 sws_init_swScale_MMX2(c);
1253 return swScale_MMX2;
1254 #elif COMPILE_TEMPLATE_AMD3DNOW
1255 sws_init_swScale_3DNow(c);
1256 return swScale_3DNow;
1257 #elif COMPILE_TEMPLATE_MMX
1258 sws_init_swScale_MMX(c);
1259 return swScale_MMX;
1260 #elif COMPILE_TEMPLATE_ALTIVEC
1261 sws_init_swScale_altivec(c);
1262 return swScale_altivec;
1263 #else
1264 sws_init_swScale_C(c);
1265 return swScale_C;
1266 #endif
1267 #endif //!CONFIG_RUNTIME_CPUDETECT
1268 }
1269
1270 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1271 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1272 {
1273 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1274 /* Copy Y plane */
1275 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1276 memcpy(dst, src[0], srcSliceH*dstStride[0]);
1277 else {
1278 int i;
1279 const uint8_t *srcPtr= src[0];
1280 uint8_t *dstPtr= dst;
1281 for (i=0; i<srcSliceH; i++) {
1282 memcpy(dstPtr, srcPtr, c->srcW);
1283 srcPtr+= srcStride[0];
1284 dstPtr+= dstStride[0];
1285 }
1286 }
1287 dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1288 if (c->dstFormat == PIX_FMT_NV12)
1289 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1290 else
1291 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1292
1293 return srcSliceH;
1294 }
1295
1296 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1297 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1298 {
1299 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1300
1301 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1302
1303 return srcSliceH;
1304 }
1305
1306 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1307 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1308 {
1309 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1310
1311 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1312
1313 return srcSliceH;
1314 }
1315
1316 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1317 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1318 {
1319 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1320
1321 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1322
1323 return srcSliceH;
1324 }
1325
1326 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1327 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1328 {
1329 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1330
1331 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1332
1333 return srcSliceH;
1334 }
1335
1336 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1337 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1338 {
1339 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1340 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1341 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1342
1343 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1344
1345 if (dstParam[3])
1346 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1347
1348 return srcSliceH;
1349 }
1350
1351 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1352 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1353 {
1354 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1355 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1356 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1357
1358 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1359
1360 return srcSliceH;
1361 }
1362
1363 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1364 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1365 {
1366 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1367 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1368 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1369
1370 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1371
1372 if (dstParam[3])
1373 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1374
1375 return srcSliceH;
1376 }
1377
1378 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1379 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1380 {
1381 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1382 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1383 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1384
1385 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1386
1387 return srcSliceH;
1388 }
1389
1390 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1391 int srcSliceH, uint8_t* dst[], int dstStride[])
1392 {
1393 const enum PixelFormat srcFormat= c->srcFormat;
1394 const enum PixelFormat dstFormat= c->dstFormat;
1395 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1396 const uint8_t *palette)=NULL;
1397 int i;
1398 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1399 const uint8_t *srcPtr= src[0];
1400
1401 if (usePal(srcFormat)) {
1402 switch (dstFormat) {
1403 case PIX_FMT_RGB32 : conv = palette8topacked32; break;
1404 case PIX_FMT_BGR32 : conv = palette8topacked32; break;
1405 case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
1406 case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
1407 case PIX_FMT_RGB24 : conv = palette8topacked24; break;
1408 case PIX_FMT_BGR24 : conv = palette8topacked24; break;
1409 }
1410 }
1411
1412 if (!conv)
1413 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1414 sws_format_name(srcFormat), sws_format_name(dstFormat));
1415 else {
1416 for (i=0; i<srcSliceH; i++) {
1417 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1418 srcPtr+= srcStride[0];
1419 dstPtr+= dstStride[0];
1420 }
1421 }
1422
1423 return srcSliceH;
1424 }
1425
1426 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1427 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1428 int srcSliceH, uint8_t* dst[], int dstStride[])
1429 {
1430 const enum PixelFormat srcFormat= c->srcFormat;
1431 const enum PixelFormat dstFormat= c->dstFormat;
1432 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1433 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1434 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1435 const int dstId= c->dstFormatBpp >> 2;
1436 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1437
1438 /* BGR -> BGR */
1439 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1440 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1441 switch(srcId | (dstId<<4)) {
1442 case 0x34: conv= rgb16to15; break;
1443 case 0x36: conv= rgb24to15; break;
1444 case 0x38: conv= rgb32to15; break;
1445 case 0x43: conv= rgb15to16; break;
1446 case 0x46: conv= rgb24to16; break;
1447 case 0x48: conv= rgb32to16; break;
1448 case 0x63: conv= rgb15to24; break;
1449 case 0x64: conv= rgb16to24; break;
1450 case 0x68: conv= rgb32to24; break;
1451 case 0x83: conv= rgb15to32; break;
1452 case 0x84: conv= rgb16to32; break;
1453 case 0x86: conv= rgb24to32; break;
1454 }
1455 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1456 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1457 switch(srcId | (dstId<<4)) {
1458 case 0x33: conv= rgb15tobgr15; break;
1459 case 0x34: conv= rgb16tobgr15; break;
1460 case 0x36: conv= rgb24tobgr15; break;
1461 case 0x38: conv= rgb32tobgr15; break;
1462 case 0x43: conv= rgb15tobgr16; break;
1463 case 0x44: conv= rgb16tobgr16; break;
1464 case 0x46: conv= rgb24tobgr16; break;
1465 case 0x48: conv= rgb32tobgr16; break;
1466 case 0x63: conv= rgb15tobgr24; break;
1467 case 0x64: conv= rgb16tobgr24; break;
1468 case 0x66: conv= rgb24tobgr24; break;
1469 case 0x68: conv= rgb32tobgr24; break;
1470 case 0x83: conv= rgb15tobgr32; break;
1471 case 0x84: conv= rgb16tobgr32; break;
1472 case 0x86: conv= rgb24tobgr32; break;
1473 case 0x88: conv= rgb32tobgr32; break;
1474 }
1475 }
1476
1477 if (!conv) {
1478 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1479 sws_format_name(srcFormat), sws_format_name(dstFormat));
1480 } else {
1481 const uint8_t *srcPtr= src[0];
1482 if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
1483 srcPtr += ALT32_CORR;
1484
1485 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1486 conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1487 else {
1488 int i;
1489 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1490
1491 for (i=0; i<srcSliceH; i++) {
1492 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1493 srcPtr+= srcStride[0];
1494 dstPtr+= dstStride[0];
1495 }
1496 }
1497 }
1498 return srcSliceH;
1499 }
1500
1501 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1502 int srcSliceH, uint8_t* dst[], int dstStride[])
1503 {
1504 rgb24toyv12(
1505 src[0],
1506 dst[0]+ srcSliceY *dstStride[0],
1507 dst[1]+(srcSliceY>>1)*dstStride[1],
1508 dst[2]+(srcSliceY>>1)*dstStride[2],
1509 c->srcW, srcSliceH,
1510 dstStride[0], dstStride[1], srcStride[0]);
1511 if (dst[3])
1512 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1513 return srcSliceH;
1514 }
1515
1516 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1517 int srcSliceH, uint8_t* dst[], int dstStride[])
1518 {
1519 int i;
1520
1521 /* copy Y */
1522 if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
1523 memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
1524 else {
1525 const uint8_t *srcPtr= src[0];
1526 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1527
1528 for (i=0; i<srcSliceH; i++) {
1529 memcpy(dstPtr, srcPtr, c->srcW);
1530 srcPtr+= srcStride[0];
1531 dstPtr+= dstStride[0];
1532 }
1533 }
1534
1535 if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
1536 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1537 srcSliceH >> 2, srcStride[1], dstStride[1]);
1538 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1539 srcSliceH >> 2, srcStride[2], dstStride[2]);
1540 } else {
1541 planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1542 srcSliceH >> 2, srcStride[1], dstStride[2]);
1543 planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1544 srcSliceH >> 2, srcStride[2], dstStride[1]);
1545 }
1546 if (dst[3])
1547 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1548 return srcSliceH;
1549 }
1550
1551 /* unscaled copy like stuff (assumes nearly identical formats) */
1552 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1553 int srcSliceH, uint8_t* dst[], int dstStride[])
1554 {
1555 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1556 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1557 else {
1558 int i;
1559 const uint8_t *srcPtr= src[0];
1560 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1561 int length=0;
1562
1563 /* universal length finder */
1564 while(length+c->srcW <= FFABS(dstStride[0])
1565 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1566 assert(length!=0);
1567
1568 for (i=0; i<srcSliceH; i++) {
1569 memcpy(dstPtr, srcPtr, length);
1570 srcPtr+= srcStride[0];
1571 dstPtr+= dstStride[0];
1572 }
1573 }
1574 return srcSliceH;
1575 }
1576
1577 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1578 int srcSliceH, uint8_t* dst[], int dstStride[])
1579 {
1580 int plane, i, j;
1581 for (plane=0; plane<4; plane++) {
1582 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1583 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1584 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1585 const uint8_t *srcPtr= src[plane];
1586 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1587
1588 if (!dst[plane]) continue;
1589 // ignore palette for GRAY8
1590 if (plane == 1 && !dst[2]) continue;
1591 if (!src[plane] || (plane == 1 && !src[2])) {
1592 if(is16BPS(c->dstFormat))
1593 length*=2;
1594 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1595 } else {
1596 if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1597 if (!isBE(c->srcFormat)) srcPtr++;
1598 for (i=0; i<height; i++) {
1599 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1600 srcPtr+= srcStride[plane];
1601 dstPtr+= dstStride[plane];
1602 }
1603 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1604 for (i=0; i<height; i++) {
1605 for (j=0; j<length; j++) {
1606 dstPtr[ j<<1 ] = srcPtr[j];
1607 dstPtr[(j<<1)+1] = srcPtr[j];
1608 }
1609 srcPtr+= srcStride[plane];
1610 dstPtr+= dstStride[plane];
1611 }
1612 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1613 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1614
1615 for (i=0; i<height; i++) {
1616 for (j=0; j<length; j++)
1617 ((uint16_t*)dstPtr)[j] = bswap_16(((const uint16_t*)srcPtr)[j]);
1618 srcPtr+= srcStride[plane];
1619 dstPtr+= dstStride[plane];
1620 }
1621 } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
1622 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
1623 else {
1624 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1625 length*=2;
1626 for (i=0; i<height; i++) {
1627 memcpy(dstPtr, srcPtr, length);
1628 srcPtr+= srcStride[plane];
1629 dstPtr+= dstStride[plane];
1630 }
1631 }
1632 }
1633 }
1634 return srcSliceH;
1635 }
1636
1637 int ff_hardcodedcpuflags(void)
1638 {
1639 int flags = 0;
1640 #if COMPILE_TEMPLATE_MMX2
1641 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1642 #elif COMPILE_TEMPLATE_AMD3DNOW
1643 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1644 #elif COMPILE_TEMPLATE_MMX
1645 flags |= SWS_CPU_CAPS_MMX;
1646 #elif COMPILE_TEMPLATE_ALTIVEC
1647 flags |= SWS_CPU_CAPS_ALTIVEC;
1648 #elif ARCH_BFIN
1649 flags |= SWS_CPU_CAPS_BFIN;
1650 #endif
1651 return flags;
1652 }
1653
1654 void ff_get_unscaled_swscale(SwsContext *c)
1655 {
1656 const enum PixelFormat srcFormat = c->srcFormat;
1657 const enum PixelFormat dstFormat = c->dstFormat;
1658 const int flags = c->flags;
1659 const int dstH = c->dstH;
1660 int needsDither;
1661
1662 needsDither= isAnyRGB(dstFormat)
1663 && c->dstFormatBpp < 24
1664 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1665
1666 /* yv12_to_nv12 */
1667 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1668 c->swScale= planarToNv12Wrapper;
1669 }
1670 /* yuv2bgr */
1671 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1672 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1673 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1674 }
1675
1676 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1677 c->swScale= yvu9ToYv12Wrapper;
1678 }
1679
1680 /* bgr24toYV12 */
1681 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1682 c->swScale= bgr24ToYv12Wrapper;
1683
1684 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1685 if ( isAnyRGB(srcFormat)
1686 && isAnyRGB(dstFormat)
1687 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1688 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1689 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1690 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1691 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1692 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1693 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1694 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1695 && dstFormat != PIX_FMT_RGB32_1
1696 && dstFormat != PIX_FMT_BGR32_1
1697 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1698 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1699 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1700 c->swScale= rgbToRgbWrapper;
1701
1702 if ((usePal(srcFormat) && (
1703 dstFormat == PIX_FMT_RGB32 ||
1704 dstFormat == PIX_FMT_RGB32_1 ||
1705 dstFormat == PIX_FMT_RGB24 ||
1706 dstFormat == PIX_FMT_BGR32 ||
1707 dstFormat == PIX_FMT_BGR32_1 ||
1708 dstFormat == PIX_FMT_BGR24)))
1709 c->swScale= palToRgbWrapper;
1710
1711 if (srcFormat == PIX_FMT_YUV422P) {
1712 if (dstFormat == PIX_FMT_YUYV422)
1713 c->swScale= yuv422pToYuy2Wrapper;
1714 else if (dstFormat == PIX_FMT_UYVY422)
1715 c->swScale= yuv422pToUyvyWrapper;
1716 }
1717
1718 /* LQ converters if -sws 0 or -sws 4*/
1719 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1720 /* yv12_to_yuy2 */
1721 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1722 if (dstFormat == PIX_FMT_YUYV422)
1723 c->swScale= planarToYuy2Wrapper;
1724 else if (dstFormat == PIX_FMT_UYVY422)
1725 c->swScale= planarToUyvyWrapper;
1726 }
1727 }
1728 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1729 c->swScale= yuyvToYuv420Wrapper;
1730 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1731 c->swScale= uyvyToYuv420Wrapper;
1732 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1733 c->swScale= yuyvToYuv422Wrapper;
1734 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1735 c->swScale= uyvyToYuv422Wrapper;
1736
1737 #ifdef COMPILE_ALTIVEC
1738 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1739 !(c->flags & SWS_BITEXACT) &&
1740 srcFormat == PIX_FMT_YUV420P) {
1741 // unscaled YV12 -> packed YUV, we want speed
1742 if (dstFormat == PIX_FMT_YUYV422)
1743 c->swScale= yv12toyuy2_unscaled_altivec;
1744 else if (dstFormat == PIX_FMT_UYVY422)
1745 c->swScale= yv12touyvy_unscaled_altivec;
1746 }
1747 #endif
1748
1749 /* simple copy */
1750 if ( srcFormat == dstFormat
1751 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1752 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1753 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1754 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1755 || (isGray(dstFormat) && isGray(srcFormat))
1756 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1757 && c->chrDstHSubSample == c->chrSrcHSubSample
1758 && c->chrDstVSubSample == c->chrSrcVSubSample
1759 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1760 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1761 {
1762 if (isPacked(c->srcFormat))
1763 c->swScale= packedCopyWrapper;
1764 else /* Planar YUV or gray */
1765 c->swScale= planarCopyWrapper;
1766 }
1767 #if ARCH_BFIN
1768 if (flags & SWS_CPU_CAPS_BFIN)
1769 ff_bfin_get_unscaled_swscale (c);
1770 #endif
1771 }
1772
1773 static void reset_ptr(const uint8_t* src[], int format)
1774 {
1775 if(!isALPHA(format))
1776 src[3]=NULL;
1777 if(!isPlanarYUV(format)) {
1778 src[3]=src[2]=NULL;
1779
1780 if (!usePal(format))
1781 src[1]= NULL;
1782 }
1783 }
1784
1785 /**
1786 * swscale wrapper, so we don't need to export the SwsContext.
1787 * Assumes planar YUV to be in YUV order instead of YVU.
1788 */
1789 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1790 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1791 {
1792 int i;
1793 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1794 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1795
1796 // do not mess up sliceDir if we have a "trailing" 0-size slice
1797 if (srcSliceH == 0)
1798 return 0;
1799
1800 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1801 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1802 return 0;
1803 }
1804 if (c->sliceDir == 0) {
1805 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1806 }
1807
1808 if (usePal(c->srcFormat)) {
1809 for (i=0; i<256; i++) {
1810 int p, r, g, b,y,u,v;
1811 if(c->srcFormat == PIX_FMT_PAL8) {
1812 p=((const uint32_t*)(src[1]))[i];
1813 r= (p>>16)&0xFF;
1814 g= (p>> 8)&0xFF;
1815 b= p &0xFF;
1816 } else if(c->srcFormat == PIX_FMT_RGB8) {
1817 r= (i>>5 )*36;
1818 g= ((i>>2)&7)*36;
1819 b= (i&3 )*85;
1820 } else if(c->srcFormat == PIX_FMT_BGR8) {
1821 b= (i>>6 )*85;
1822 g= ((i>>3)&7)*36;
1823 r= (i&7 )*36;
1824 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1825 r= (i>>3 )*255;
1826 g= ((i>>1)&3)*85;
1827 b= (i&1 )*255;
1828 } else {
1829 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1830 b= (i>>3 )*255;
1831 g= ((i>>1)&3)*85;
1832 r= (i&1 )*255;
1833 }
1834 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1835 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1836 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1837 c->pal_yuv[i]= y + (u<<8) + (v<<16);
1838
1839 switch(c->dstFormat) {
1840 case PIX_FMT_BGR32:
1841 #if !HAVE_BIGENDIAN
1842 case PIX_FMT_RGB24:
1843 #endif
1844 c->pal_rgb[i]= r + (g<<8) + (b<<16);
1845 break;
1846 case PIX_FMT_BGR32_1:
1847 #if HAVE_BIGENDIAN
1848 case PIX_FMT_BGR24:
1849 #endif
1850 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1851 break;
1852 case PIX_FMT_RGB32_1:
1853 #if HAVE_BIGENDIAN
1854 case PIX_FMT_RGB24:
1855 #endif
1856 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1857 break;
1858 case PIX_FMT_RGB32:
1859 #if !HAVE_BIGENDIAN
1860 case PIX_FMT_BGR24:
1861 #endif
1862 default:
1863 c->pal_rgb[i]= b + (g<<8) + (r<<16);
1864 }
1865 }
1866 }
1867
1868 // copy strides, so they can safely be modified
1869 if (c->sliceDir == 1) {
1870 // slices go from top to bottom
1871 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1872 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1873
1874 reset_ptr(src2, c->srcFormat);
1875 reset_ptr((const uint8_t**)dst2, c->dstFormat);
1876
1877 /* reset slice direction at end of frame */
1878 if (srcSliceY + srcSliceH == c->srcH)
1879 c->sliceDir = 0;
1880
1881 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
1882 } else {
1883 // slices go from bottom to top => we flip the image internally
1884 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
1885 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
1886
1887 src2[0] += (srcSliceH-1)*srcStride[0];
1888 if (!usePal(c->srcFormat))
1889 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
1890 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
1891 src2[3] += (srcSliceH-1)*srcStride[3];
1892 dst2[0] += ( c->dstH -1)*dstStride[0];
1893 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
1894 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
1895 dst2[3] += ( c->dstH -1)*dstStride[3];
1896
1897 reset_ptr(src2, c->srcFormat);
1898 reset_ptr((const uint8_t**)dst2, c->dstFormat);
1899
1900 /* reset slice direction at end of frame */
1901 if (!srcSliceY)
1902 c->sliceDir = 0;
1903
1904 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
1905 }
1906 }
1907
1908 #if LIBSWSCALE_VERSION_MAJOR < 1
1909 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
1910 int srcSliceH, uint8_t* dst[], int dstStride[])
1911 {
1912 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
1913 }
1914 #endif