?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
colorspace-test: fix build
[libswscale.git] / swscale.c
blob375171f8957686a4642ab59b585900cc5f9e05a5
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
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.
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 GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /*
22 supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
23 supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
24 {BGR,RGB}{1,4,8,15,16} support dithering
25
26 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
27 YV12 -> {BGR,RGB}{1,4,8,12,15,16,24,32}
28 x -> x
29 YUV9 -> YV12
30 YUV9/YV12 -> Y800
31 Y800 -> YUV9/YV12
32 BGR24 -> BGR32 & RGB24 -> RGB32
33 BGR32 -> BGR24 & RGB32 -> RGB24
34 BGR15 -> BGR16
35 */
36
37 /*
38 tested special converters (most are tested actually, but I did not write it down ...)
39 YV12 -> BGR12/BGR16
40 YV12 -> YV12
41 BGR15 -> BGR16
42 BGR16 -> BGR16
43 YVU9 -> YV12
44
45 untested special converters
46 YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
47 YV12/I420 -> YV12/I420
48 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
49 BGR24 -> BGR32 & RGB24 -> RGB32
50 BGR32 -> BGR24 & RGB32 -> RGB24
51 BGR24 -> YV12
52 */
53
54 #include <inttypes.h>
55 #include <string.h>
56 #include <math.h>
57 #include <stdio.h>
58 #include "config.h"
59 #include <assert.h>
60 #include "swscale.h"
61 #include "swscale_internal.h"
62 #include "rgb2rgb.h"
63 #include "libavutil/intreadwrite.h"
64 #include "libavutil/x86_cpu.h"
65 #include "libavutil/avutil.h"
66 #include "libavutil/mathematics.h"
67 #include "libavutil/bswap.h"
68 #include "libavutil/pixdesc.h"
69
70 #undef MOVNTQ
71 #undef PAVGB
72
73 //#undef HAVE_MMX2
74 //#define HAVE_AMD3DNOW
75 //#undef HAVE_MMX
76 //#undef ARCH_X86
77 #define DITHER1XBPP
78
79 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
80
81 #define isPacked(x) ( \
82 (x)==PIX_FMT_PAL8 \
83 || (x)==PIX_FMT_YUYV422 \
84 || (x)==PIX_FMT_UYVY422 \
85 || (x)==PIX_FMT_Y400A \
86 || isAnyRGB(x) \
87 )
88
89 #define RGB2YUV_SHIFT 15
90 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
91 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
92 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
93 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
94 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
95 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
96 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
97 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
98 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
99
100 static const double rgb2yuv_table[8][9]={
101 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
102 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
103 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
104 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
105 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
106 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
107 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
108 {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
109 };
110
111 /*
112 NOTES
113 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
114
115 TODO
116 more intelligent misalignment avoidance for the horizontal scaler
117 write special vertical cubic upscale version
118 optimize C code (YV12 / minmax)
119 add support for packed pixel YUV input & output
120 add support for Y8 output
121 optimize BGR24 & BGR32
122 add BGR4 output support
123 write special BGR->BGR scaler
124 */
125
126 #if ARCH_X86
127 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
128 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
129 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
130 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
131 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
132 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
133 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
134 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
135
136 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
137 0x0103010301030103LL,
138 0x0200020002000200LL,};
139
140 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
141 0x0602060206020602LL,
142 0x0004000400040004LL,};
143
144 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
145 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
146 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
147 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
148 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
149 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
150
151 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
152 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
153 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
154
155 #ifdef FAST_BGR2YV12
156 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
157 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
158 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
159 #else
160 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
161 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
162 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
163 #endif /* FAST_BGR2YV12 */
164 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
165 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
166 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
167
168 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
169 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
170 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
171 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
172 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
173
174 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
175 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
176 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
177 };
178
179 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
180
181 #endif /* ARCH_X86 */
182
183 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
184 { 1, 3, 1, 3, 1, 3, 1, 3, },
185 { 2, 0, 2, 0, 2, 0, 2, 0, },
186 };
187
188 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
189 { 6, 2, 6, 2, 6, 2, 6, 2, },
190 { 0, 4, 0, 4, 0, 4, 0, 4, },
191 };
192
193 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
194 { 8, 4, 11, 7, 8, 4, 11, 7, },
195 { 2, 14, 1, 13, 2, 14, 1, 13, },
196 { 10, 6, 9, 5, 10, 6, 9, 5, },
197 { 0, 12, 3, 15, 0, 12, 3, 15, },
198 };
199
200 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
201 { 17, 9, 23, 15, 16, 8, 22, 14, },
202 { 5, 29, 3, 27, 4, 28, 2, 26, },
203 { 21, 13, 19, 11, 20, 12, 18, 10, },
204 { 0, 24, 6, 30, 1, 25, 7, 31, },
205 { 16, 8, 22, 14, 17, 9, 23, 15, },
206 { 4, 28, 2, 26, 5, 29, 3, 27, },
207 { 20, 12, 18, 10, 21, 13, 19, 11, },
208 { 1, 25, 7, 31, 0, 24, 6, 30, },
209 };
210
211 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
212 { 0, 55, 14, 68, 3, 58, 17, 72, },
213 { 37, 18, 50, 32, 40, 22, 54, 35, },
214 { 9, 64, 5, 59, 13, 67, 8, 63, },
215 { 46, 27, 41, 23, 49, 31, 44, 26, },
216 { 2, 57, 16, 71, 1, 56, 15, 70, },
217 { 39, 21, 52, 34, 38, 19, 51, 33, },
218 { 11, 66, 7, 62, 10, 65, 6, 60, },
219 { 48, 30, 43, 25, 47, 29, 42, 24, },
220 };
221
222 #if 1
223 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
224 {117, 62, 158, 103, 113, 58, 155, 100, },
225 { 34, 199, 21, 186, 31, 196, 17, 182, },
226 {144, 89, 131, 76, 141, 86, 127, 72, },
227 { 0, 165, 41, 206, 10, 175, 52, 217, },
228 {110, 55, 151, 96, 120, 65, 162, 107, },
229 { 28, 193, 14, 179, 38, 203, 24, 189, },
230 {138, 83, 124, 69, 148, 93, 134, 79, },
231 { 7, 172, 48, 213, 3, 168, 45, 210, },
232 };
233 #elif 1
234 // tries to correct a gamma of 1.5
235 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
236 { 0, 143, 18, 200, 2, 156, 25, 215, },
237 { 78, 28, 125, 64, 89, 36, 138, 74, },
238 { 10, 180, 3, 161, 16, 195, 8, 175, },
239 {109, 51, 93, 38, 121, 60, 105, 47, },
240 { 1, 152, 23, 210, 0, 147, 20, 205, },
241 { 85, 33, 134, 71, 81, 30, 130, 67, },
242 { 14, 190, 6, 171, 12, 185, 5, 166, },
243 {117, 57, 101, 44, 113, 54, 97, 41, },
244 };
245 #elif 1
246 // tries to correct a gamma of 2.0
247 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
248 { 0, 124, 8, 193, 0, 140, 12, 213, },
249 { 55, 14, 104, 42, 66, 19, 119, 52, },
250 { 3, 168, 1, 145, 6, 187, 3, 162, },
251 { 86, 31, 70, 21, 99, 39, 82, 28, },
252 { 0, 134, 11, 206, 0, 129, 9, 200, },
253 { 62, 17, 114, 48, 58, 16, 109, 45, },
254 { 5, 181, 2, 157, 4, 175, 1, 151, },
255 { 95, 36, 78, 26, 90, 34, 74, 24, },
256 };
257 #else
258 // tries to correct a gamma of 2.5
259 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
260 { 0, 107, 3, 187, 0, 125, 6, 212, },
261 { 39, 7, 86, 28, 49, 11, 102, 36, },
262 { 1, 158, 0, 131, 3, 180, 1, 151, },
263 { 68, 19, 52, 12, 81, 25, 64, 17, },
264 { 0, 119, 5, 203, 0, 113, 4, 195, },
265 { 45, 9, 96, 33, 42, 8, 91, 30, },
266 { 2, 172, 1, 144, 2, 165, 0, 137, },
267 { 77, 23, 60, 15, 72, 21, 56, 14, },
268 };
269 #endif
270
271 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
272 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
273 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
274 int dstW, int chrDstW, int big_endian)
275 {
276 //FIXME Optimize (just quickly written not optimized..)
277 int i;
278
279 for (i = 0; i < dstW; i++) {
280 int val = 1 << 10;
281 int j;
282
283 for (j = 0; j < lumFilterSize; j++)
284 val += lumSrc[j][i] * lumFilter[j];
285
286 if (big_endian) {
287 AV_WB16(&dest[i], av_clip_uint16(val >> 11));
288 } else {
289 AV_WL16(&dest[i], av_clip_uint16(val >> 11));
290 }
291 }
292
293 if (uDest) {
294 for (i = 0; i < chrDstW; i++) {
295 int u = 1 << 10;
296 int v = 1 << 10;
297 int j;
298
299 for (j = 0; j < chrFilterSize; j++) {
300 u += chrSrc[j][i ] * chrFilter[j];
301 v += chrSrc[j][i + VOFW] * chrFilter[j];
302 }
303
304 if (big_endian) {
305 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
306 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
307 } else {
308 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
309 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
310 }
311 }
312 }
313
314 if (CONFIG_SWSCALE_ALPHA && aDest) {
315 for (i = 0; i < dstW; i++) {
316 int val = 1 << 10;
317 int j;
318
319 for (j = 0; j < lumFilterSize; j++)
320 val += alpSrc[j][i] * lumFilter[j];
321
322 if (big_endian) {
323 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
324 } else {
325 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
326 }
327 }
328 }
329 }
330
331 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
332 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
333 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
334 enum PixelFormat dstFormat)
335 {
336 if (isBE(dstFormat)) {
337 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
338 chrFilter, chrSrc, chrFilterSize,
339 alpSrc,
340 dest, uDest, vDest, aDest,
341 dstW, chrDstW, 1);
342 } else {
343 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
344 chrFilter, chrSrc, chrFilterSize,
345 alpSrc,
346 dest, uDest, vDest, aDest,
347 dstW, chrDstW, 0);
348 }
349 }
350
351 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
352 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
353 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
354 {
355 //FIXME Optimize (just quickly written not optimized..)
356 int i;
357 for (i=0; i<dstW; i++) {
358 int val=1<<18;
359 int j;
360 for (j=0; j<lumFilterSize; j++)
361 val += lumSrc[j][i] * lumFilter[j];
362
363 dest[i]= av_clip_uint8(val>>19);
364 }
365
366 if (uDest)
367 for (i=0; i<chrDstW; i++) {
368 int u=1<<18;
369 int v=1<<18;
370 int j;
371 for (j=0; j<chrFilterSize; j++) {
372 u += chrSrc[j][i] * chrFilter[j];
373 v += chrSrc[j][i + VOFW] * chrFilter[j];
374 }
375
376 uDest[i]= av_clip_uint8(u>>19);
377 vDest[i]= av_clip_uint8(v>>19);
378 }
379
380 if (CONFIG_SWSCALE_ALPHA && aDest)
381 for (i=0; i<dstW; i++) {
382 int val=1<<18;
383 int j;
384 for (j=0; j<lumFilterSize; j++)
385 val += alpSrc[j][i] * lumFilter[j];
386
387 aDest[i]= av_clip_uint8(val>>19);
388 }
389
390 }
391
392 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
393 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
394 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
395 {
396 //FIXME Optimize (just quickly written not optimized..)
397 int i;
398 for (i=0; i<dstW; i++) {
399 int val=1<<18;
400 int j;
401 for (j=0; j<lumFilterSize; j++)
402 val += lumSrc[j][i] * lumFilter[j];
403
404 dest[i]= av_clip_uint8(val>>19);
405 }
406
407 if (!uDest)
408 return;
409
410 if (dstFormat == PIX_FMT_NV12)
411 for (i=0; i<chrDstW; i++) {
412 int u=1<<18;
413 int v=1<<18;
414 int j;
415 for (j=0; j<chrFilterSize; j++) {
416 u += chrSrc[j][i] * chrFilter[j];
417 v += chrSrc[j][i + VOFW] * chrFilter[j];
418 }
419
420 uDest[2*i]= av_clip_uint8(u>>19);
421 uDest[2*i+1]= av_clip_uint8(v>>19);
422 }
423 else
424 for (i=0; i<chrDstW; i++) {
425 int u=1<<18;
426 int v=1<<18;
427 int j;
428 for (j=0; j<chrFilterSize; j++) {
429 u += chrSrc[j][i] * chrFilter[j];
430 v += chrSrc[j][i + VOFW] * chrFilter[j];
431 }
432
433 uDest[2*i]= av_clip_uint8(v>>19);
434 uDest[2*i+1]= av_clip_uint8(u>>19);
435 }
436 }
437
438 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
439 for (i=0; i<(dstW>>1); i++) {\
440 int j;\
441 int Y1 = 1<<18;\
442 int Y2 = 1<<18;\
443 int U = 1<<18;\
444 int V = 1<<18;\
445 int av_unused A1, A2;\
446 type av_unused *r, *b, *g;\
447 const int i2= 2*i;\
448 \
449 for (j=0; j<lumFilterSize; j++) {\
450 Y1 += lumSrc[j][i2] * lumFilter[j];\
451 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
452 }\
453 for (j=0; j<chrFilterSize; j++) {\
454 U += chrSrc[j][i] * chrFilter[j];\
455 V += chrSrc[j][i+VOFW] * chrFilter[j];\
456 }\
457 Y1>>=19;\
458 Y2>>=19;\
459 U >>=19;\
460 V >>=19;\
461 if (alpha) {\
462 A1 = 1<<18;\
463 A2 = 1<<18;\
464 for (j=0; j<lumFilterSize; j++) {\
465 A1 += alpSrc[j][i2 ] * lumFilter[j];\
466 A2 += alpSrc[j][i2+1] * lumFilter[j];\
467 }\
468 A1>>=19;\
469 A2>>=19;\
470 }
471
472 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
473 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
474 if ((Y1|Y2|U|V)&256) {\
475 if (Y1>255) Y1=255; \
476 else if (Y1<0)Y1=0; \
477 if (Y2>255) Y2=255; \
478 else if (Y2<0)Y2=0; \
479 if (U>255) U=255; \
480 else if (U<0) U=0; \
481 if (V>255) V=255; \
482 else if (V<0) V=0; \
483 }\
484 if (alpha && ((A1|A2)&256)) {\
485 A1=av_clip_uint8(A1);\
486 A2=av_clip_uint8(A2);\
487 }
488
489 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
490 for (i=0; i<dstW; i++) {\
491 int j;\
492 int Y = 0;\
493 int U = -128<<19;\
494 int V = -128<<19;\
495 int av_unused A;\
496 int R,G,B;\
497 \
498 for (j=0; j<lumFilterSize; j++) {\
499 Y += lumSrc[j][i ] * lumFilter[j];\
500 }\
501 for (j=0; j<chrFilterSize; j++) {\
502 U += chrSrc[j][i ] * chrFilter[j];\
503 V += chrSrc[j][i+VOFW] * chrFilter[j];\
504 }\
505 Y >>=10;\
506 U >>=10;\
507 V >>=10;\
508 if (alpha) {\
509 A = rnd;\
510 for (j=0; j<lumFilterSize; j++)\
511 A += alpSrc[j][i ] * lumFilter[j];\
512 A >>=19;\
513 if (A&256)\
514 A = av_clip_uint8(A);\
515 }
516
517 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
518 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
519 Y-= c->yuv2rgb_y_offset;\
520 Y*= c->yuv2rgb_y_coeff;\
521 Y+= rnd;\
522 R= Y + V*c->yuv2rgb_v2r_coeff;\
523 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
524 B= Y + U*c->yuv2rgb_u2b_coeff;\
525 if ((R|G|B)&(0xC0000000)) {\
526 if (R>=(256<<22)) R=(256<<22)-1; \
527 else if (R<0)R=0; \
528 if (G>=(256<<22)) G=(256<<22)-1; \
529 else if (G<0)G=0; \
530 if (B>=(256<<22)) B=(256<<22)-1; \
531 else if (B<0)B=0; \
532 }
533
534 #define YSCALE_YUV_2_GRAY16_C \
535 for (i=0; i<(dstW>>1); i++) {\
536 int j;\
537 int Y1 = 1<<18;\
538 int Y2 = 1<<18;\
539 int U = 1<<18;\
540 int V = 1<<18;\
541 \
542 const int i2= 2*i;\
543 \
544 for (j=0; j<lumFilterSize; j++) {\
545 Y1 += lumSrc[j][i2] * lumFilter[j];\
546 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
547 }\
548 Y1>>=11;\
549 Y2>>=11;\
550 if ((Y1|Y2|U|V)&65536) {\
551 if (Y1>65535) Y1=65535; \
552 else if (Y1<0)Y1=0; \
553 if (Y2>65535) Y2=65535; \
554 else if (Y2<0)Y2=0; \
555 }
556
557 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
558 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
559 r = (type *)c->table_rV[V]; \
560 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
561 b = (type *)c->table_bU[U];
562
563 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
564 for (i=0; i<(dstW>>1); i++) { \
565 const int i2= 2*i; \
566 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
567 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
568 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
569 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
570 type av_unused *r, *b, *g; \
571 int av_unused A1, A2; \
572 if (alpha) {\
573 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
574 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
575 }
576
577 #define YSCALE_YUV_2_GRAY16_2_C \
578 for (i=0; i<(dstW>>1); i++) { \
579 const int i2= 2*i; \
580 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
581 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
582
583 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
584 YSCALE_YUV_2_PACKED2_C(type,alpha)\
585 r = (type *)c->table_rV[V];\
586 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
587 b = (type *)c->table_bU[U];
588
589 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
590 for (i=0; i<(dstW>>1); i++) {\
591 const int i2= 2*i;\
592 int Y1= buf0[i2 ]>>7;\
593 int Y2= buf0[i2+1]>>7;\
594 int U= (uvbuf1[i ])>>7;\
595 int V= (uvbuf1[i+VOFW])>>7;\
596 type av_unused *r, *b, *g;\
597 int av_unused A1, A2;\
598 if (alpha) {\
599 A1= abuf0[i2 ]>>7;\
600 A2= abuf0[i2+1]>>7;\
601 }
602
603 #define YSCALE_YUV_2_GRAY16_1_C \
604 for (i=0; i<(dstW>>1); i++) {\
605 const int i2= 2*i;\
606 int Y1= buf0[i2 ]<<1;\
607 int Y2= buf0[i2+1]<<1;
608
609 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
610 YSCALE_YUV_2_PACKED1_C(type,alpha)\
611 r = (type *)c->table_rV[V];\
612 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
613 b = (type *)c->table_bU[U];
614
615 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
616 for (i=0; i<(dstW>>1); i++) {\
617 const int i2= 2*i;\
618 int Y1= buf0[i2 ]>>7;\
619 int Y2= buf0[i2+1]>>7;\
620 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
621 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
622 type av_unused *r, *b, *g;\
623 int av_unused A1, A2;\
624 if (alpha) {\
625 A1= abuf0[i2 ]>>7;\
626 A2= abuf0[i2+1]>>7;\
627 }
628
629 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
630 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
631 r = (type *)c->table_rV[V];\
632 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
633 b = (type *)c->table_bU[U];
634
635 #define YSCALE_YUV_2_MONO2_C \
636 const uint8_t * const d128=dither_8x8_220[y&7];\
637 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
638 for (i=0; i<dstW-7; i+=8) {\
639 int acc;\
640 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
641 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
642 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
643 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
644 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
645 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
646 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
647 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
648 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
649 dest++;\
650 }
651
652 #define YSCALE_YUV_2_MONOX_C \
653 const uint8_t * const d128=dither_8x8_220[y&7];\
654 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
655 int acc=0;\
656 for (i=0; i<dstW-1; i+=2) {\
657 int j;\
658 int Y1=1<<18;\
659 int Y2=1<<18;\
660 \
661 for (j=0; j<lumFilterSize; j++) {\
662 Y1 += lumSrc[j][i] * lumFilter[j];\
663 Y2 += lumSrc[j][i+1] * lumFilter[j];\
664 }\
665 Y1>>=19;\
666 Y2>>=19;\
667 if ((Y1|Y2)&256) {\
668 if (Y1>255) Y1=255;\
669 else if (Y1<0)Y1=0;\
670 if (Y2>255) Y2=255;\
671 else if (Y2<0)Y2=0;\
672 }\
673 acc+= acc + g[Y1+d128[(i+0)&7]];\
674 acc+= acc + g[Y2+d128[(i+1)&7]];\
675 if ((i&7)==6) {\
676 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
677 dest++;\
678 }\
679 }
680
681 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
682 switch(c->dstFormat) {\
683 case PIX_FMT_RGB48BE:\
684 case PIX_FMT_RGB48LE:\
685 func(uint8_t,0)\
686 ((uint8_t*)dest)[ 0]= r[Y1];\
687 ((uint8_t*)dest)[ 1]= r[Y1];\
688 ((uint8_t*)dest)[ 2]= g[Y1];\
689 ((uint8_t*)dest)[ 3]= g[Y1];\
690 ((uint8_t*)dest)[ 4]= b[Y1];\
691 ((uint8_t*)dest)[ 5]= b[Y1];\
692 ((uint8_t*)dest)[ 6]= r[Y2];\
693 ((uint8_t*)dest)[ 7]= r[Y2];\
694 ((uint8_t*)dest)[ 8]= g[Y2];\
695 ((uint8_t*)dest)[ 9]= g[Y2];\
696 ((uint8_t*)dest)[10]= b[Y2];\
697 ((uint8_t*)dest)[11]= b[Y2];\
698 dest+=12;\
699 }\
700 break;\
701 case PIX_FMT_RGBA:\
702 case PIX_FMT_BGRA:\
703 if (CONFIG_SMALL) {\
704 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
705 func(uint32_t,needAlpha)\
706 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
707 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
708 }\
709 } else {\
710 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
711 func(uint32_t,1)\
712 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
713 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
714 }\
715 } else {\
716 func(uint32_t,0)\
717 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
718 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
719 }\
720 }\
721 }\
722 break;\
723 case PIX_FMT_ARGB:\
724 case PIX_FMT_ABGR:\
725 if (CONFIG_SMALL) {\
726 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
727 func(uint32_t,needAlpha)\
728 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
729 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
730 }\
731 } else {\
732 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
733 func(uint32_t,1)\
734 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
735 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
736 }\
737 } else {\
738 func(uint32_t,0)\
739 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
740 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
741 }\
742 }\
743 } \
744 break;\
745 case PIX_FMT_RGB24:\
746 func(uint8_t,0)\
747 ((uint8_t*)dest)[0]= r[Y1];\
748 ((uint8_t*)dest)[1]= g[Y1];\
749 ((uint8_t*)dest)[2]= b[Y1];\
750 ((uint8_t*)dest)[3]= r[Y2];\
751 ((uint8_t*)dest)[4]= g[Y2];\
752 ((uint8_t*)dest)[5]= b[Y2];\
753 dest+=6;\
754 }\
755 break;\
756 case PIX_FMT_BGR24:\
757 func(uint8_t,0)\
758 ((uint8_t*)dest)[0]= b[Y1];\
759 ((uint8_t*)dest)[1]= g[Y1];\
760 ((uint8_t*)dest)[2]= r[Y1];\
761 ((uint8_t*)dest)[3]= b[Y2];\
762 ((uint8_t*)dest)[4]= g[Y2];\
763 ((uint8_t*)dest)[5]= r[Y2];\
764 dest+=6;\
765 }\
766 break;\
767 case PIX_FMT_RGB565BE:\
768 case PIX_FMT_RGB565LE:\
769 case PIX_FMT_BGR565BE:\
770 case PIX_FMT_BGR565LE:\
771 {\
772 const int dr1= dither_2x2_8[y&1 ][0];\
773 const int dg1= dither_2x2_4[y&1 ][0];\
774 const int db1= dither_2x2_8[(y&1)^1][0];\
775 const int dr2= dither_2x2_8[y&1 ][1];\
776 const int dg2= dither_2x2_4[y&1 ][1];\
777 const int db2= dither_2x2_8[(y&1)^1][1];\
778 func(uint16_t,0)\
779 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
780 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
781 }\
782 }\
783 break;\
784 case PIX_FMT_RGB555BE:\
785 case PIX_FMT_RGB555LE:\
786 case PIX_FMT_BGR555BE:\
787 case PIX_FMT_BGR555LE:\
788 {\
789 const int dr1= dither_2x2_8[y&1 ][0];\
790 const int dg1= dither_2x2_8[y&1 ][1];\
791 const int db1= dither_2x2_8[(y&1)^1][0];\
792 const int dr2= dither_2x2_8[y&1 ][1];\
793 const int dg2= dither_2x2_8[y&1 ][0];\
794 const int db2= dither_2x2_8[(y&1)^1][1];\
795 func(uint16_t,0)\
796 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
797 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
798 }\
799 }\
800 break;\
801 case PIX_FMT_RGB444BE:\
802 case PIX_FMT_RGB444LE:\
803 case PIX_FMT_BGR444BE:\
804 case PIX_FMT_BGR444LE:\
805 {\
806 const int dr1= dither_4x4_16[y&3 ][0];\
807 const int dg1= dither_4x4_16[y&3 ][1];\
808 const int db1= dither_4x4_16[(y&3)^3][0];\
809 const int dr2= dither_4x4_16[y&3 ][1];\
810 const int dg2= dither_4x4_16[y&3 ][0];\
811 const int db2= dither_4x4_16[(y&3)^3][1];\
812 func(uint16_t,0)\
813 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
814 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
815 }\
816 }\
817 break;\
818 case PIX_FMT_RGB8:\
819 case PIX_FMT_BGR8:\
820 {\
821 const uint8_t * const d64= dither_8x8_73[y&7];\
822 const uint8_t * const d32= dither_8x8_32[y&7];\
823 func(uint8_t,0)\
824 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
825 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
826 }\
827 }\
828 break;\
829 case PIX_FMT_RGB4:\
830 case PIX_FMT_BGR4:\
831 {\
832 const uint8_t * const d64= dither_8x8_73 [y&7];\
833 const uint8_t * const d128=dither_8x8_220[y&7];\
834 func(uint8_t,0)\
835 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
836 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
837 }\
838 }\
839 break;\
840 case PIX_FMT_RGB4_BYTE:\
841 case PIX_FMT_BGR4_BYTE:\
842 {\
843 const uint8_t * const d64= dither_8x8_73 [y&7];\
844 const uint8_t * const d128=dither_8x8_220[y&7];\
845 func(uint8_t,0)\
846 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
847 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
848 }\
849 }\
850 break;\
851 case PIX_FMT_MONOBLACK:\
852 case PIX_FMT_MONOWHITE:\
853 {\
854 func_monoblack\
855 }\
856 break;\
857 case PIX_FMT_YUYV422:\
858 func2\
859 ((uint8_t*)dest)[2*i2+0]= Y1;\
860 ((uint8_t*)dest)[2*i2+1]= U;\
861 ((uint8_t*)dest)[2*i2+2]= Y2;\
862 ((uint8_t*)dest)[2*i2+3]= V;\
863 } \
864 break;\
865 case PIX_FMT_UYVY422:\
866 func2\
867 ((uint8_t*)dest)[2*i2+0]= U;\
868 ((uint8_t*)dest)[2*i2+1]= Y1;\
869 ((uint8_t*)dest)[2*i2+2]= V;\
870 ((uint8_t*)dest)[2*i2+3]= Y2;\
871 } \
872 break;\
873 case PIX_FMT_GRAY16BE:\
874 func_g16\
875 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
876 ((uint8_t*)dest)[2*i2+1]= Y1;\
877 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
878 ((uint8_t*)dest)[2*i2+3]= Y2;\
879 } \
880 break;\
881 case PIX_FMT_GRAY16LE:\
882 func_g16\
883 ((uint8_t*)dest)[2*i2+0]= Y1;\
884 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
885 ((uint8_t*)dest)[2*i2+2]= Y2;\
886 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
887 } \
888 break;\
889 }
890
891 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
892 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
893 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
894 {
895 int i;
896 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)
897 }
898
899 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
900 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
901 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
902 {
903 int i;
904 int step= c->dstFormatBpp/8;
905 int aidx= 3;
906
907 switch(c->dstFormat) {
908 case PIX_FMT_ARGB:
909 dest++;
910 aidx= 0;
911 case PIX_FMT_RGB24:
912 aidx--;
913 case PIX_FMT_RGBA:
914 if (CONFIG_SMALL) {
915 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
916 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
917 dest[aidx]= needAlpha ? A : 255;
918 dest[0]= R>>22;
919 dest[1]= G>>22;
920 dest[2]= B>>22;
921 dest+= step;
922 }
923 } else {
924 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
925 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
926 dest[aidx]= A;
927 dest[0]= R>>22;
928 dest[1]= G>>22;
929 dest[2]= B>>22;
930 dest+= step;
931 }
932 } else {
933 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
934 dest[aidx]= 255;
935 dest[0]= R>>22;
936 dest[1]= G>>22;
937 dest[2]= B>>22;
938 dest+= step;
939 }
940 }
941 }
942 break;
943 case PIX_FMT_ABGR:
944 dest++;
945 aidx= 0;
946 case PIX_FMT_BGR24:
947 aidx--;
948 case PIX_FMT_BGRA:
949 if (CONFIG_SMALL) {
950 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
951 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
952 dest[aidx]= needAlpha ? A : 255;
953 dest[0]= B>>22;
954 dest[1]= G>>22;
955 dest[2]= R>>22;
956 dest+= step;
957 }
958 } else {
959 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
960 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
961 dest[aidx]= A;
962 dest[0]= B>>22;
963 dest[1]= G>>22;
964 dest[2]= R>>22;
965 dest+= step;
966 }
967 } else {
968 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
969 dest[aidx]= 255;
970 dest[0]= B>>22;
971 dest[1]= G>>22;
972 dest[2]= R>>22;
973 dest+= step;
974 }
975 }
976 }
977 break;
978 default:
979 assert(0);
980 }
981 }
982
983 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
984 {
985 int i;
986 uint8_t *ptr = plane + stride*y;
987 for (i=0; i<height; i++) {
988 memset(ptr, val, width);
989 ptr += stride;
990 }
991 }
992
993 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
994 uint32_t *unused)
995 {
996 int i;
997 for (i = 0; i < width; i++) {
998 int r = src[i*6+0];
999 int g = src[i*6+2];
1000 int b = src[i*6+4];
1001
1002 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1003 }
1004 }
1005
1006 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1007 const uint8_t *src1, const uint8_t *src2,
1008 long width, uint32_t *unused)
1009 {
1010 int i;
1011 assert(src1==src2);
1012 for (i = 0; i < width; i++) {
1013 int r = src1[6*i + 0];
1014 int g = src1[6*i + 2];
1015 int b = src1[6*i + 4];
1016
1017 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1018 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1019 }
1020 }
1021
1022 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1023 const uint8_t *src1, const uint8_t *src2,
1024 long width, uint32_t *unused)
1025 {
1026 int i;
1027 assert(src1==src2);
1028 for (i = 0; i < width; i++) {
1029 int r= src1[12*i + 0] + src1[12*i + 6];
1030 int g= src1[12*i + 2] + src1[12*i + 8];
1031 int b= src1[12*i + 4] + src1[12*i + 10];
1032
1033 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1034 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1035 }
1036 }
1037
1038 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1039 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1040 {\
1041 int i;\
1042 for (i=0; i<width; i++) {\
1043 int b= (((const type*)src)[i]>>shb)&maskb;\
1044 int g= (((const type*)src)[i]>>shg)&maskg;\
1045 int r= (((const type*)src)[i]>>shr)&maskr;\
1046 \
1047 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1048 }\
1049 }
1050
1051 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1052 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1053 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1054 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1055 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1056 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1057 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1058 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1059
1060 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1061 {
1062 int i;
1063 for (i=0; i<width; i++) {
1064 dst[i]= src[4*i];
1065 }
1066 }
1067
1068 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1069 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1070 {\
1071 int i;\
1072 for (i=0; i<width; i++) {\
1073 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1074 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1075 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1076 \
1077 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1078 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1079 }\
1080 }\
1081 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1082 {\
1083 int i;\
1084 for (i=0; i<width; i++) {\
1085 int pix0= ((const type*)src)[2*i+0]>>shp;\
1086 int pix1= ((const type*)src)[2*i+1]>>shp;\
1087 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1088 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1089 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1090 g&= maskg|(2*maskg);\
1091 \
1092 g>>=shg;\
1093 \
1094 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1095 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1096 }\
1097 }
1098
1099 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1100 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1101 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1102 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1103 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1104 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1105 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1106 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1107
1108 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1109 {
1110 int i;
1111 for (i=0; i<width; i++) {
1112 int d= src[i];
1113
1114 dst[i]= pal[d] & 0xFF;
1115 }
1116 }
1117
1118 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1119 const uint8_t *src1, const uint8_t *src2,
1120 long width, uint32_t *pal)
1121 {
1122 int i;
1123 assert(src1 == src2);
1124 for (i=0; i<width; i++) {
1125 int p= pal[src1[i]];
1126
1127 dstU[i]= p>>8;
1128 dstV[i]= p>>16;
1129 }
1130 }
1131
1132 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1133 {
1134 int i, j;
1135 for (i=0; i<width/8; i++) {
1136 int d= ~src[i];
1137 for(j=0; j<8; j++)
1138 dst[8*i+j]= ((d>>(7-j))&1)*255;
1139 }
1140 }
1141
1142 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1143 {
1144 int i, j;
1145 for (i=0; i<width/8; i++) {
1146 int d= src[i];
1147 for(j=0; j<8; j++)
1148 dst[8*i+j]= ((d>>(7-j))&1)*255;
1149 }
1150 }
1151
1152 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1153 //Plain C versions
1154 #if CONFIG_RUNTIME_CPUDETECT
1155 # define COMPILE_C 1
1156 # if ARCH_X86
1157 # define COMPILE_MMX HAVE_MMX
1158 # define COMPILE_MMX2 HAVE_MMX2
1159 # define COMPILE_3DNOW HAVE_AMD3DNOW
1160 # elif ARCH_PPC
1161 # define COMPILE_ALTIVEC HAVE_ALTIVEC
1162 # endif
1163 #else /* CONFIG_RUNTIME_CPUDETECT */
1164 # if ARCH_X86
1165 # if HAVE_MMX2
1166 # define COMPILE_MMX2 1
1167 # elif HAVE_AMD3DNOW
1168 # define COMPILE_3DNOW 1
1169 # elif HAVE_MMX
1170 # define COMPILE_MMX 1
1171 # else
1172 # define COMPILE_C 1
1173 # endif
1174 # elif ARCH_PPC && HAVE_ALTIVEC
1175 # define COMPILE_ALTIVEC 1
1176 # else
1177 # define COMPILE_C 1
1178 # endif
1179 #endif
1180
1181 #ifndef COMPILE_C
1182 # define COMPILE_C 0
1183 #endif
1184 #ifndef COMPILE_MMX
1185 # define COMPILE_MMX 0
1186 #endif
1187 #ifndef COMPILE_MMX2
1188 # define COMPILE_MMX2 0
1189 #endif
1190 #ifndef COMPILE_3DNOW
1191 # define COMPILE_3DNOW 0
1192 #endif
1193 #ifndef COMPILE_ALTIVEC
1194 # define COMPILE_ALTIVEC 0
1195 #endif
1196
1197 #define COMPILE_TEMPLATE_MMX 0
1198 #define COMPILE_TEMPLATE_MMX2 0
1199 #define COMPILE_TEMPLATE_AMD3DNOW 0
1200 #define COMPILE_TEMPLATE_ALTIVEC 0
1201
1202 #if COMPILE_C
1203 #define RENAME(a) a ## _C
1204 #include "swscale_template.c"
1205 #endif
1206
1207 #if COMPILE_ALTIVEC
1208 #undef RENAME
1209 #undef COMPILE_TEMPLATE_ALTIVEC
1210 #define COMPILE_TEMPLATE_ALTIVEC 1
1211 #define RENAME(a) a ## _altivec
1212 #include "swscale_template.c"
1213 #endif
1214
1215 #if ARCH_X86
1216
1217 //MMX versions
1218 #if COMPILE_MMX
1219 #undef RENAME
1220 #undef COMPILE_TEMPLATE_MMX
1221 #undef COMPILE_TEMPLATE_MMX2
1222 #undef COMPILE_TEMPLATE_AMD3DNOW
1223 #define COMPILE_TEMPLATE_MMX 1
1224 #define COMPILE_TEMPLATE_MMX2 0
1225 #define COMPILE_TEMPLATE_AMD3DNOW 0
1226 #define RENAME(a) a ## _MMX
1227 #include "swscale_template.c"
1228 #endif
1229
1230 //MMX2 versions
1231 #if COMPILE_MMX2
1232 #undef RENAME
1233 #undef COMPILE_TEMPLATE_MMX
1234 #undef COMPILE_TEMPLATE_MMX2
1235 #undef COMPILE_TEMPLATE_AMD3DNOW
1236 #define COMPILE_TEMPLATE_MMX 1
1237 #define COMPILE_TEMPLATE_MMX2 1
1238 #define COMPILE_TEMPLATE_AMD3DNOW 0
1239 #define RENAME(a) a ## _MMX2
1240 #include "swscale_template.c"
1241 #endif
1242
1243 //3DNOW versions
1244 #if COMPILE_3DNOW
1245 #undef RENAME
1246 #undef COMPILE_TEMPLATE_MMX
1247 #undef COMPILE_TEMPLATE_MMX2
1248 #undef COMPILE_TEMPLATE_AMD3DNOW
1249 #define COMPILE_TEMPLATE_MMX 1
1250 #define COMPILE_TEMPLATE_MMX2 0
1251 #define COMPILE_TEMPLATE_AMD3DNOW 1
1252 #define RENAME(a) a ## _3DNow
1253 #include "swscale_template.c"
1254 #endif
1255
1256 #endif //ARCH_X86
1257
1258 SwsFunc ff_getSwsFunc(SwsContext *c)
1259 {
1260 #if CONFIG_RUNTIME_CPUDETECT
1261 int flags = c->flags;
1262
1263 #if ARCH_X86
1264 // ordered per speed fastest first
1265 if (flags & SWS_CPU_CAPS_MMX2) {
1266 sws_init_swScale_MMX2(c);
1267 return swScale_MMX2;
1268 } else if (flags & SWS_CPU_CAPS_3DNOW) {
1269 sws_init_swScale_3DNow(c);
1270 return swScale_3DNow;
1271 } else if (flags & SWS_CPU_CAPS_MMX) {
1272 sws_init_swScale_MMX(c);
1273 return swScale_MMX;
1274 } else {
1275 sws_init_swScale_C(c);
1276 return swScale_C;
1277 }
1278
1279 #else
1280 #if COMPILE_ALTIVEC
1281 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1282 sws_init_swScale_altivec(c);
1283 return swScale_altivec;
1284 } else {
1285 sws_init_swScale_C(c);
1286 return swScale_C;
1287 }
1288 #endif
1289 sws_init_swScale_C(c);
1290 return swScale_C;
1291 #endif /* ARCH_X86 */
1292 #else //CONFIG_RUNTIME_CPUDETECT
1293 #if COMPILE_TEMPLATE_MMX2
1294 sws_init_swScale_MMX2(c);
1295 return swScale_MMX2;
1296 #elif COMPILE_TEMPLATE_AMD3DNOW
1297 sws_init_swScale_3DNow(c);
1298 return swScale_3DNow;
1299 #elif COMPILE_TEMPLATE_MMX
1300 sws_init_swScale_MMX(c);
1301 return swScale_MMX;
1302 #elif COMPILE_TEMPLATE_ALTIVEC
1303 sws_init_swScale_altivec(c);
1304 return swScale_altivec;
1305 #else
1306 sws_init_swScale_C(c);
1307 return swScale_C;
1308 #endif
1309 #endif //!CONFIG_RUNTIME_CPUDETECT
1310 }
1311
1312 static void copyPlane(const uint8_t *src, int srcStride,
1313 int srcSliceY, int srcSliceH, int width,
1314 uint8_t *dst, int dstStride)
1315 {
1316 dst += dstStride * srcSliceY;
1317 if (dstStride == srcStride && srcStride > 0) {
1318 memcpy(dst, src, srcSliceH * dstStride);
1319 } else {
1320 int i;
1321 for (i=0; i<srcSliceH; i++) {
1322 memcpy(dst, src, width);
1323 src += srcStride;
1324 dst += dstStride;
1325 }
1326 }
1327 }
1328
1329 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1330 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1331 {
1332 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1333
1334 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1335 dstParam[0], dstStride[0]);
1336
1337 if (c->dstFormat == PIX_FMT_NV12)
1338 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1339 else
1340 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1341
1342 return srcSliceH;
1343 }
1344
1345 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1346 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1347 {
1348 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1349
1350 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1351
1352 return srcSliceH;
1353 }
1354
1355 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1356 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1357 {
1358 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1359
1360 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1361
1362 return srcSliceH;
1363 }
1364
1365 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1366 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1367 {
1368 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1369
1370 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1371
1372 return srcSliceH;
1373 }
1374
1375 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1376 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1377 {
1378 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1379
1380 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1381
1382 return srcSliceH;
1383 }
1384
1385 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1386 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1387 {
1388 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1389 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1390 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1391
1392 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1393
1394 if (dstParam[3])
1395 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1396
1397 return srcSliceH;
1398 }
1399
1400 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1401 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1402 {
1403 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1404 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1405 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1406
1407 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1408
1409 return srcSliceH;
1410 }
1411
1412 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1413 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1414 {
1415 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1416 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1417 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1418
1419 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1420
1421 if (dstParam[3])
1422 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1423
1424 return srcSliceH;
1425 }
1426
1427 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1428 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1429 {
1430 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1431 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1432 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1433
1434 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1435
1436 return srcSliceH;
1437 }
1438
1439 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1440 {
1441 long i;
1442 for (i=0; i<num_pixels; i++)
1443 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1444 }
1445
1446 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1447 {
1448 long i;
1449
1450 for (i=0; i<num_pixels; i++)
1451 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1452 }
1453
1454 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1455 {
1456 long i;
1457
1458 for (i=0; i<num_pixels; i++) {
1459 //FIXME slow?
1460 dst[0]= palette[src[i<<1]*4+0];
1461 dst[1]= palette[src[i<<1]*4+1];
1462 dst[2]= palette[src[i<<1]*4+2];
1463 dst+= 3;
1464 }
1465 }
1466
1467 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1468 int srcSliceH, uint8_t* dst[], int dstStride[])
1469 {
1470 const enum PixelFormat srcFormat= c->srcFormat;
1471 const enum PixelFormat dstFormat= c->dstFormat;
1472 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1473 const uint8_t *palette)=NULL;
1474 int i;
1475 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1476 const uint8_t *srcPtr= src[0];
1477
1478 if (srcFormat == PIX_FMT_Y400A) {
1479 switch (dstFormat) {
1480 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1481 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1482 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1483 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1484 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1485 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1486 }
1487 } else if (usePal(srcFormat)) {
1488 switch (dstFormat) {
1489 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1490 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1491 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1492 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1493 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1494 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1495 }
1496 }
1497
1498 if (!conv)
1499 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1500 sws_format_name(srcFormat), sws_format_name(dstFormat));
1501 else {
1502 for (i=0; i<srcSliceH; i++) {
1503 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1504 srcPtr+= srcStride[0];
1505 dstPtr+= dstStride[0];
1506 }
1507 }
1508
1509 return srcSliceH;
1510 }
1511
1512 #define isRGBA32(x) ( \
1513 (x) == PIX_FMT_ARGB \
1514 || (x) == PIX_FMT_RGBA \
1515 || (x) == PIX_FMT_BGRA \
1516 || (x) == PIX_FMT_ABGR \
1517 )
1518
1519 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1520 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1521 int srcSliceH, uint8_t* dst[], int dstStride[])
1522 {
1523 const enum PixelFormat srcFormat= c->srcFormat;
1524 const enum PixelFormat dstFormat= c->dstFormat;
1525 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1526 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1527 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1528 const int dstId= c->dstFormatBpp >> 2;
1529 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1530
1531 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1532
1533 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1534 if ( CONV_IS(ABGR, RGBA)
1535 || CONV_IS(ARGB, BGRA)
1536 || CONV_IS(BGRA, ARGB)
1537 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1538 else if (CONV_IS(ABGR, ARGB)
1539 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1540 else if (CONV_IS(ABGR, BGRA)
1541 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1542 else if (CONV_IS(BGRA, RGBA)
1543 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1544 else if (CONV_IS(BGRA, ABGR)
1545 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1546 } else
1547 /* BGR -> BGR */
1548 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1549 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1550 switch(srcId | (dstId<<4)) {
1551 case 0x34: conv= rgb16to15; break;
1552 case 0x36: conv= rgb24to15; break;
1553 case 0x38: conv= rgb32to15; break;
1554 case 0x43: conv= rgb15to16; break;
1555 case 0x46: conv= rgb24to16; break;
1556 case 0x48: conv= rgb32to16; break;
1557 case 0x63: conv= rgb15to24; break;
1558 case 0x64: conv= rgb16to24; break;
1559 case 0x68: conv= rgb32to24; break;
1560 case 0x83: conv= rgb15to32; break;
1561 case 0x84: conv= rgb16to32; break;
1562 case 0x86: conv= rgb24to32; break;
1563 }
1564 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1565 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1566 switch(srcId | (dstId<<4)) {
1567 case 0x33: conv= rgb15tobgr15; break;
1568 case 0x34: conv= rgb16tobgr15; break;
1569 case 0x36: conv= rgb24tobgr15; break;
1570 case 0x38: conv= rgb32tobgr15; break;
1571 case 0x43: conv= rgb15tobgr16; break;
1572 case 0x44: conv= rgb16tobgr16; break;
1573 case 0x46: conv= rgb24tobgr16; break;
1574 case 0x48: conv= rgb32tobgr16; break;
1575 case 0x63: conv= rgb15tobgr24; break;
1576 case 0x64: conv= rgb16tobgr24; break;
1577 case 0x66: conv= rgb24tobgr24; break;
1578 case 0x68: conv= rgb32tobgr24; break;
1579 case 0x83: conv= rgb15tobgr32; break;
1580 case 0x84: conv= rgb16tobgr32; break;
1581 case 0x86: conv= rgb24tobgr32; break;
1582 }
1583 }
1584
1585 if (!conv) {
1586 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1587 sws_format_name(srcFormat), sws_format_name(dstFormat));
1588 } else {
1589 const uint8_t *srcPtr= src[0];
1590 uint8_t *dstPtr= dst[0];
1591 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1592 srcPtr += ALT32_CORR;
1593
1594 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1595 dstPtr += ALT32_CORR;
1596
1597 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1598 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1599 else {
1600 int i;
1601 dstPtr += dstStride[0]*srcSliceY;
1602
1603 for (i=0; i<srcSliceH; i++) {
1604 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1605 srcPtr+= srcStride[0];
1606 dstPtr+= dstStride[0];
1607 }
1608 }
1609 }
1610 return srcSliceH;
1611 }
1612
1613 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1614 int srcSliceH, uint8_t* dst[], int dstStride[])
1615 {
1616 rgb24toyv12(
1617 src[0],
1618 dst[0]+ srcSliceY *dstStride[0],
1619 dst[1]+(srcSliceY>>1)*dstStride[1],
1620 dst[2]+(srcSliceY>>1)*dstStride[2],
1621 c->srcW, srcSliceH,
1622 dstStride[0], dstStride[1], srcStride[0]);
1623 if (dst[3])
1624 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1625 return srcSliceH;
1626 }
1627
1628 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1629 int srcSliceH, uint8_t* dst[], int dstStride[])
1630 {
1631 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1632 dst[0], dstStride[0]);
1633
1634 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1635 srcSliceH >> 2, srcStride[1], dstStride[1]);
1636 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1637 srcSliceH >> 2, srcStride[2], dstStride[2]);
1638 if (dst[3])
1639 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1640 return srcSliceH;
1641 }
1642
1643 /* unscaled copy like stuff (assumes nearly identical formats) */
1644 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1645 int srcSliceH, uint8_t* dst[], int dstStride[])
1646 {
1647 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1648 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1649 else {
1650 int i;
1651 const uint8_t *srcPtr= src[0];
1652 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1653 int length=0;
1654
1655 /* universal length finder */
1656 while(length+c->srcW <= FFABS(dstStride[0])
1657 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1658 assert(length!=0);
1659
1660 for (i=0; i<srcSliceH; i++) {
1661 memcpy(dstPtr, srcPtr, length);
1662 srcPtr+= srcStride[0];
1663 dstPtr+= dstStride[0];
1664 }
1665 }
1666 return srcSliceH;
1667 }
1668
1669 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1670 int srcSliceH, uint8_t* dst[], int dstStride[])
1671 {
1672 int plane, i, j;
1673 for (plane=0; plane<4; plane++) {
1674 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1675 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1676 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1677 const uint8_t *srcPtr= src[plane];
1678 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1679
1680 if (!dst[plane]) continue;
1681 // ignore palette for GRAY8
1682 if (plane == 1 && !dst[2]) continue;
1683 if (!src[plane] || (plane == 1 && !src[2])) {
1684 if(is16BPS(c->dstFormat))
1685 length*=2;
1686 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1687 } else {
1688 if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1689 if (!isBE(c->srcFormat)) srcPtr++;
1690 for (i=0; i<height; i++) {
1691 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1692 srcPtr+= srcStride[plane];
1693 dstPtr+= dstStride[plane];
1694 }
1695 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1696 for (i=0; i<height; i++) {
1697 for (j=0; j<length; j++) {
1698 dstPtr[ j<<1 ] = srcPtr[j];
1699 dstPtr[(j<<1)+1] = srcPtr[j];
1700 }
1701 srcPtr+= srcStride[plane];
1702 dstPtr+= dstStride[plane];
1703 }
1704 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1705 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1706
1707 for (i=0; i<height; i++) {
1708 for (j=0; j<length; j++)
1709 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1710 srcPtr+= srcStride[plane];
1711 dstPtr+= dstStride[plane];
1712 }
1713 } else if (dstStride[plane] == srcStride[plane] &&
1714 srcStride[plane] > 0 && srcStride[plane] == length) {
1715 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1716 height*dstStride[plane]);
1717 } else {
1718 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1719 length*=2;
1720 for (i=0; i<height; i++) {
1721 memcpy(dstPtr, srcPtr, length);
1722 srcPtr+= srcStride[plane];
1723 dstPtr+= dstStride[plane];
1724 }
1725 }
1726 }
1727 }
1728 return srcSliceH;
1729 }
1730
1731 int ff_hardcodedcpuflags(void)
1732 {
1733 int flags = 0;
1734 #if COMPILE_TEMPLATE_MMX2
1735 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1736 #elif COMPILE_TEMPLATE_AMD3DNOW
1737 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1738 #elif COMPILE_TEMPLATE_MMX
1739 flags |= SWS_CPU_CAPS_MMX;
1740 #elif COMPILE_TEMPLATE_ALTIVEC
1741 flags |= SWS_CPU_CAPS_ALTIVEC;
1742 #elif ARCH_BFIN
1743 flags |= SWS_CPU_CAPS_BFIN;
1744 #endif
1745 return flags;
1746 }
1747
1748 void ff_get_unscaled_swscale(SwsContext *c)
1749 {
1750 const enum PixelFormat srcFormat = c->srcFormat;
1751 const enum PixelFormat dstFormat = c->dstFormat;
1752 const int flags = c->flags;
1753 const int dstH = c->dstH;
1754 int needsDither;
1755
1756 needsDither= isAnyRGB(dstFormat)
1757 && c->dstFormatBpp < 24
1758 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1759
1760 /* yv12_to_nv12 */
1761 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1762 c->swScale= planarToNv12Wrapper;
1763 }
1764 /* yuv2bgr */
1765 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1766 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1767 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1768 }
1769
1770 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1771 c->swScale= yvu9ToYv12Wrapper;
1772 }
1773
1774 /* bgr24toYV12 */
1775 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1776 c->swScale= bgr24ToYv12Wrapper;
1777
1778 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1779 if ( isAnyRGB(srcFormat)
1780 && isAnyRGB(dstFormat)
1781 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1782 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1783 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1784 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1785 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1786 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1787 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1788 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1789 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1790 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1791 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1792 c->swScale= rgbToRgbWrapper;
1793
1794 if ((usePal(srcFormat) && (
1795 dstFormat == PIX_FMT_RGB32 ||
1796 dstFormat == PIX_FMT_RGB32_1 ||
1797 dstFormat == PIX_FMT_RGB24 ||
1798 dstFormat == PIX_FMT_BGR32 ||
1799 dstFormat == PIX_FMT_BGR32_1 ||
1800 dstFormat == PIX_FMT_BGR24)))
1801 c->swScale= palToRgbWrapper;
1802
1803 if (srcFormat == PIX_FMT_YUV422P) {
1804 if (dstFormat == PIX_FMT_YUYV422)
1805 c->swScale= yuv422pToYuy2Wrapper;
1806 else if (dstFormat == PIX_FMT_UYVY422)
1807 c->swScale= yuv422pToUyvyWrapper;
1808 }
1809
1810 /* LQ converters if -sws 0 or -sws 4*/
1811 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1812 /* yv12_to_yuy2 */
1813 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1814 if (dstFormat == PIX_FMT_YUYV422)
1815 c->swScale= planarToYuy2Wrapper;
1816 else if (dstFormat == PIX_FMT_UYVY422)
1817 c->swScale= planarToUyvyWrapper;
1818 }
1819 }
1820 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1821 c->swScale= yuyvToYuv420Wrapper;
1822 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1823 c->swScale= uyvyToYuv420Wrapper;
1824 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1825 c->swScale= yuyvToYuv422Wrapper;
1826 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1827 c->swScale= uyvyToYuv422Wrapper;
1828
1829 #if COMPILE_ALTIVEC
1830 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1831 !(c->flags & SWS_BITEXACT) &&
1832 srcFormat == PIX_FMT_YUV420P) {
1833 // unscaled YV12 -> packed YUV, we want speed
1834 if (dstFormat == PIX_FMT_YUYV422)
1835 c->swScale= yv12toyuy2_unscaled_altivec;
1836 else if (dstFormat == PIX_FMT_UYVY422)
1837 c->swScale= yv12touyvy_unscaled_altivec;
1838 }
1839 #endif
1840
1841 /* simple copy */
1842 if ( srcFormat == dstFormat
1843 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1844 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1845 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1846 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1847 || (isGray(dstFormat) && isGray(srcFormat))
1848 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1849 && c->chrDstHSubSample == c->chrSrcHSubSample
1850 && c->chrDstVSubSample == c->chrSrcVSubSample
1851 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1852 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1853 {
1854 if (isPacked(c->srcFormat))
1855 c->swScale= packedCopyWrapper;
1856 else /* Planar YUV or gray */
1857 c->swScale= planarCopyWrapper;
1858 }
1859 #if ARCH_BFIN
1860 if (flags & SWS_CPU_CAPS_BFIN)
1861 ff_bfin_get_unscaled_swscale (c);
1862 #endif
1863 }
1864
1865 static void reset_ptr(const uint8_t* src[], int format)
1866 {
1867 if(!isALPHA(format))
1868 src[3]=NULL;
1869 if(!isPlanarYUV(format)) {
1870 src[3]=src[2]=NULL;
1871
1872 if (!usePal(format))
1873 src[1]= NULL;
1874 }
1875 }
1876
1877 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1878 const int linesizes[4])
1879 {
1880 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1881 int i;
1882
1883 for (i = 0; i < 4; i++) {
1884 int plane = desc->comp[i].plane;
1885 if (!data[plane] || !linesizes[plane])
1886 return 0;
1887 }
1888
1889 return 1;
1890 }
1891
1892 /**
1893 * swscale wrapper, so we don't need to export the SwsContext.
1894 * Assumes planar YUV to be in YUV order instead of YVU.
1895 */
1896 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1897 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1898 {
1899 int i;
1900 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1901 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1902
1903 // do not mess up sliceDir if we have a "trailing" 0-size slice
1904 if (srcSliceH == 0)
1905 return 0;
1906
1907 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1908 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1909 return 0;
1910 }
1911 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1912 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1913 return 0;
1914 }
1915
1916 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1917 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1918 return 0;
1919 }
1920 if (c->sliceDir == 0) {
1921 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1922 }
1923
1924 if (usePal(c->srcFormat)) {
1925 for (i=0; i<256; i++) {
1926 int p, r, g, b,y,u,v;
1927 if(c->srcFormat == PIX_FMT_PAL8) {
1928 p=((const uint32_t*)(src[1]))[i];
1929 r= (p>>16)&0xFF;
1930 g= (p>> 8)&0xFF;
1931 b= p &0xFF;
1932 } else if(c->srcFormat == PIX_FMT_RGB8) {
1933 r= (i>>5 )*36;
1934 g= ((i>>2)&7)*36;
1935 b= (i&3 )*85;
1936 } else if(c->srcFormat == PIX_FMT_BGR8) {
1937 b= (i>>6 )*85;
1938 g= ((i>>3)&7)*36;
1939 r= (i&7 )*36;
1940 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1941 r= (i>>3 )*255;
1942 g= ((i>>1)&3)*85;
1943 b= (i&1 )*255;
1944 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
1945 r = g = b = i;
1946 } else {
1947 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1948 b= (i>>3 )*255;
1949 g= ((i>>1)&3)*85;
1950 r= (i&1 )*255;
1951 }
1952 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1953 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1954 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1955 c->pal_yuv[i]= y + (u<<8) + (v<<16);
1956
1957 switch(c->dstFormat) {
1958 case PIX_FMT_BGR32:
1959 #if !HAVE_BIGENDIAN
1960 case PIX_FMT_RGB24:
1961 #endif
1962 c->pal_rgb[i]= r + (g<<8) + (b<<16);
1963 break;
1964 case PIX_FMT_BGR32_1:
1965 #if HAVE_BIGENDIAN
1966 case PIX_FMT_BGR24:
1967 #endif
1968 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1969 break;
1970 case PIX_FMT_RGB32_1:
1971 #if HAVE_BIGENDIAN
1972 case PIX_FMT_RGB24:
1973 #endif
1974 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
1975 break;
1976 case PIX_FMT_RGB32:
1977 #if !HAVE_BIGENDIAN
1978 case PIX_FMT_BGR24:
1979 #endif
1980 default:
1981 c->pal_rgb[i]= b + (g<<8) + (r<<16);
1982 }
1983 }
1984 }
1985
1986 // copy strides, so they can safely be modified
1987 if (c->sliceDir == 1) {
1988 // slices go from top to bottom
1989 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
1990 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
1991
1992 reset_ptr(src2, c->srcFormat);
1993 reset_ptr((const uint8_t**)dst2, c->dstFormat);
1994
1995 /* reset slice direction at end of frame */
1996 if (srcSliceY + srcSliceH == c->srcH)
1997 c->sliceDir = 0;
1998
1999 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2000 } else {
2001 // slices go from bottom to top => we flip the image internally
2002 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2003 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2004
2005 src2[0] += (srcSliceH-1)*srcStride[0];
2006 if (!usePal(c->srcFormat))
2007 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2008 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2009 src2[3] += (srcSliceH-1)*srcStride[3];
2010 dst2[0] += ( c->dstH -1)*dstStride[0];
2011 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2012 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2013 dst2[3] += ( c->dstH -1)*dstStride[3];
2014
2015 reset_ptr(src2, c->srcFormat);
2016 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2017
2018 /* reset slice direction at end of frame */
2019 if (!srcSliceY)
2020 c->sliceDir = 0;
2021
2022 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2023 }
2024 }
2025
2026 #if LIBSWSCALE_VERSION_MAJOR < 1
2027 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
2028 int srcSliceH, uint8_t* dst[], int dstStride[])
2029 {
2030 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2031 }
2032 #endif
2033
2034 /* Convert the palette to the same packed 32-bit format as the palette */
2035 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2036 {
2037 long i;
2038
2039 for (i=0; i<num_pixels; i++)
2040 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2041 }
2042
2043 /* Palette format: ABCD -> dst format: ABC */
2044 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2045 {
2046 long i;
2047
2048 for (i=0; i<num_pixels; i++) {
2049 //FIXME slow?
2050 dst[0]= palette[src[i]*4+0];
2051 dst[1]= palette[src[i]*4+1];
2052 dst[2]= palette[src[i]*4+2];
2053 dst+= 3;
2054 }
2055 }