2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
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
26 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
27 YV12 -> {BGR,RGB}{1,4,8,12,15,16,24,32}
32 BGR24 -> BGR32 & RGB24 -> RGB32
33 BGR32 -> BGR24 & RGB32 -> RGB24
38 tested special converters (most are tested actually, but I did not write it down ...)
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
61 #include "swscale_internal.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"
74 //#define HAVE_AMD3DNOW
79 #define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
81 #define isPacked(x) ( \
83 || (x)==PIX_FMT_YUYV422 \
84 || (x)==PIX_FMT_UYVY422 \
85 || (x)==PIX_FMT_Y400A \
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))
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
113 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
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
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;
136 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
137 0x0103010301030103LL,
138 0x0200020002000200LL,};
140 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
141 0x0602060206020602LL,
142 0x0004000400040004LL,};
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;
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;
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;
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;
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;
174 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
175 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
176 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
179 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
181 #endif /* ARCH_X86 */
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, },
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, },
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, },
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, },
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, },
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, },
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, },
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, },
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, },
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)
276 //FIXME Optimize (just quickly written not optimized..)
279 for (i = 0; i < dstW; i++) {
283 for (j = 0; j < lumFilterSize; j++)
284 val += lumSrc[j][i] * lumFilter[j];
287 AV_WB16(&dest[i], av_clip_uint16(val >> 11));
289 AV_WL16(&dest[i], av_clip_uint16(val >> 11));
294 for (i = 0; i < chrDstW; i++) {
299 for (j = 0; j < chrFilterSize; j++) {
300 u += chrSrc[j][i ] * chrFilter[j];
301 v += chrSrc[j][i + VOFW] * chrFilter[j];
305 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
306 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
308 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
309 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
314 if (CONFIG_SWSCALE_ALPHA && aDest) {
315 for (i = 0; i < dstW; i++) {
319 for (j = 0; j < lumFilterSize; j++)
320 val += alpSrc[j][i] * lumFilter[j];
323 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
325 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
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)
336 if (isBE(dstFormat)) {
337 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
338 chrFilter, chrSrc, chrFilterSize,
340 dest, uDest, vDest, aDest,
343 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
344 chrFilter, chrSrc, chrFilterSize,
346 dest, uDest, vDest, aDest,
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)
355 //FIXME Optimize (just quickly written not optimized..)
357 for (i=0; i<dstW; i++) {
360 for (j=0; j<lumFilterSize; j++)
361 val += lumSrc[j][i] * lumFilter[j];
363 dest[i]= av_clip_uint8(val>>19);
367 for (i=0; i<chrDstW; i++) {
371 for (j=0; j<chrFilterSize; j++) {
372 u += chrSrc[j][i] * chrFilter[j];
373 v += chrSrc[j][i + VOFW] * chrFilter[j];
376 uDest[i]= av_clip_uint8(u>>19);
377 vDest[i]= av_clip_uint8(v>>19);
380 if (CONFIG_SWSCALE_ALPHA && aDest)
381 for (i=0; i<dstW; i++) {
384 for (j=0; j<lumFilterSize; j++)
385 val += alpSrc[j][i] * lumFilter[j];
387 aDest[i]= av_clip_uint8(val>>19);
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)
396 //FIXME Optimize (just quickly written not optimized..)
398 for (i=0; i<dstW; i++) {
401 for (j=0; j<lumFilterSize; j++)
402 val += lumSrc[j][i] * lumFilter[j];
404 dest[i]= av_clip_uint8(val>>19);
410 if (dstFormat == PIX_FMT_NV12)
411 for (i=0; i<chrDstW; i++) {
415 for (j=0; j<chrFilterSize; j++) {
416 u += chrSrc[j][i] * chrFilter[j];
417 v += chrSrc[j][i + VOFW] * chrFilter[j];
420 uDest[2*i]= av_clip_uint8(u>>19);
421 uDest[2*i+1]= av_clip_uint8(v>>19);
424 for (i=0; i<chrDstW; i++) {
428 for (j=0; j<chrFilterSize; j++) {
429 u += chrSrc[j][i] * chrFilter[j];
430 v += chrSrc[j][i + VOFW] * chrFilter[j];
433 uDest[2*i]= av_clip_uint8(v>>19);
434 uDest[2*i+1]= av_clip_uint8(u>>19);
438 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
439 for (i=0; i<(dstW>>1); i++) {\
445 int av_unused A1, A2;\
446 type av_unused *r, *b, *g;\
449 for (j=0; j<lumFilterSize; j++) {\
450 Y1 += lumSrc[j][i2] * lumFilter[j];\
451 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
453 for (j=0; j<chrFilterSize; j++) {\
454 U += chrSrc[j][i] * chrFilter[j];\
455 V += chrSrc[j][i+VOFW] * chrFilter[j];\
464 for (j=0; j<lumFilterSize; j++) {\
465 A1 += alpSrc[j][i2 ] * lumFilter[j];\
466 A2 += alpSrc[j][i2+1] * lumFilter[j];\
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; \
484 if (alpha && ((A1|A2)&256)) {\
485 A1=av_clip_uint8(A1);\
486 A2=av_clip_uint8(A2);\
489 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
490 for (i=0; i<dstW; i++) {\
498 for (j=0; j<lumFilterSize; j++) {\
499 Y += lumSrc[j][i ] * lumFilter[j];\
501 for (j=0; j<chrFilterSize; j++) {\
502 U += chrSrc[j][i ] * chrFilter[j];\
503 V += chrSrc[j][i+VOFW] * chrFilter[j];\
510 for (j=0; j<lumFilterSize; j++)\
511 A += alpSrc[j][i ] * lumFilter[j];\
514 A = av_clip_uint8(A);\
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;\
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; \
528 if (G>=(256<<22)) G=(256<<22)-1; \
530 if (B>=(256<<22)) B=(256<<22)-1; \
534 #define YSCALE_YUV_2_GRAY16_C \
535 for (i=0; i<(dstW>>1); i++) {\
544 for (j=0; j<lumFilterSize; j++) {\
545 Y1 += lumSrc[j][i2] * lumFilter[j];\
546 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
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; \
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];
563 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
564 for (i=0; i<(dstW>>1); 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; \
573 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
574 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
577 #define YSCALE_YUV_2_GRAY16_2_C \
578 for (i=0; i<(dstW>>1); i++) { \
580 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
581 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
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];
589 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
590 for (i=0; i<(dstW>>1); 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;\
603 #define YSCALE_YUV_2_GRAY16_1_C \
604 for (i=0; i<(dstW>>1); i++) {\
606 int Y1= buf0[i2 ]<<1;\
607 int Y2= buf0[i2+1]<<1;
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];
615 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
616 for (i=0; i<(dstW>>1); 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;\
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];
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) {\
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;\
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];\
656 for (i=0; i<dstW-1; i+=2) {\
661 for (j=0; j<lumFilterSize; j++) {\
662 Y1 += lumSrc[j][i] * lumFilter[j];\
663 Y2 += lumSrc[j][i+1] * lumFilter[j];\
673 acc+= acc + g[Y1+d128[(i+0)&7]];\
674 acc+= acc + g[Y2+d128[(i+1)&7]];\
676 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
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:\
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];\
701 case PIX_FMT_BGR48BE:\
702 case PIX_FMT_BGR48LE:\
704 ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
705 ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
706 ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
707 ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
708 ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
709 ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
716 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
717 func(uint32_t,needAlpha)\
718 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
719 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
722 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
724 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
725 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
729 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
730 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
738 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
739 func(uint32_t,needAlpha)\
740 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
741 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
744 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
746 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
747 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
751 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
752 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
759 ((uint8_t*)dest)[0]= r[Y1];\
760 ((uint8_t*)dest)[1]= g[Y1];\
761 ((uint8_t*)dest)[2]= b[Y1];\
762 ((uint8_t*)dest)[3]= r[Y2];\
763 ((uint8_t*)dest)[4]= g[Y2];\
764 ((uint8_t*)dest)[5]= b[Y2];\
770 ((uint8_t*)dest)[0]= b[Y1];\
771 ((uint8_t*)dest)[1]= g[Y1];\
772 ((uint8_t*)dest)[2]= r[Y1];\
773 ((uint8_t*)dest)[3]= b[Y2];\
774 ((uint8_t*)dest)[4]= g[Y2];\
775 ((uint8_t*)dest)[5]= r[Y2];\
779 case PIX_FMT_RGB565BE:\
780 case PIX_FMT_RGB565LE:\
781 case PIX_FMT_BGR565BE:\
782 case PIX_FMT_BGR565LE:\
784 const int dr1= dither_2x2_8[y&1 ][0];\
785 const int dg1= dither_2x2_4[y&1 ][0];\
786 const int db1= dither_2x2_8[(y&1)^1][0];\
787 const int dr2= dither_2x2_8[y&1 ][1];\
788 const int dg2= dither_2x2_4[y&1 ][1];\
789 const int db2= dither_2x2_8[(y&1)^1][1];\
791 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
792 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
796 case PIX_FMT_RGB555BE:\
797 case PIX_FMT_RGB555LE:\
798 case PIX_FMT_BGR555BE:\
799 case PIX_FMT_BGR555LE:\
801 const int dr1= dither_2x2_8[y&1 ][0];\
802 const int dg1= dither_2x2_8[y&1 ][1];\
803 const int db1= dither_2x2_8[(y&1)^1][0];\
804 const int dr2= dither_2x2_8[y&1 ][1];\
805 const int dg2= dither_2x2_8[y&1 ][0];\
806 const int db2= dither_2x2_8[(y&1)^1][1];\
808 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
809 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
813 case PIX_FMT_RGB444BE:\
814 case PIX_FMT_RGB444LE:\
815 case PIX_FMT_BGR444BE:\
816 case PIX_FMT_BGR444LE:\
818 const int dr1= dither_4x4_16[y&3 ][0];\
819 const int dg1= dither_4x4_16[y&3 ][1];\
820 const int db1= dither_4x4_16[(y&3)^3][0];\
821 const int dr2= dither_4x4_16[y&3 ][1];\
822 const int dg2= dither_4x4_16[y&3 ][0];\
823 const int db2= dither_4x4_16[(y&3)^3][1];\
825 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
826 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
833 const uint8_t * const d64= dither_8x8_73[y&7];\
834 const uint8_t * const d32= dither_8x8_32[y&7];\
836 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
837 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
844 const uint8_t * const d64= dither_8x8_73 [y&7];\
845 const uint8_t * const d128=dither_8x8_220[y&7];\
847 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
848 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
852 case PIX_FMT_RGB4_BYTE:\
853 case PIX_FMT_BGR4_BYTE:\
855 const uint8_t * const d64= dither_8x8_73 [y&7];\
856 const uint8_t * const d128=dither_8x8_220[y&7];\
858 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
859 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
863 case PIX_FMT_MONOBLACK:\
864 case PIX_FMT_MONOWHITE:\
869 case PIX_FMT_YUYV422:\
871 ((uint8_t*)dest)[2*i2+0]= Y1;\
872 ((uint8_t*)dest)[2*i2+1]= U;\
873 ((uint8_t*)dest)[2*i2+2]= Y2;\
874 ((uint8_t*)dest)[2*i2+3]= V;\
877 case PIX_FMT_UYVY422:\
879 ((uint8_t*)dest)[2*i2+0]= U;\
880 ((uint8_t*)dest)[2*i2+1]= Y1;\
881 ((uint8_t*)dest)[2*i2+2]= V;\
882 ((uint8_t*)dest)[2*i2+3]= Y2;\
885 case PIX_FMT_GRAY16BE:\
887 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
888 ((uint8_t*)dest)[2*i2+1]= Y1;\
889 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
890 ((uint8_t*)dest)[2*i2+3]= Y2;\
893 case PIX_FMT_GRAY16LE:\
895 ((uint8_t*)dest)[2*i2+0]= Y1;\
896 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
897 ((uint8_t*)dest)[2*i2+2]= Y2;\
898 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
903 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
904 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
905 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
908 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)
911 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
912 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
913 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
916 int step= c->dstFormatBpp/8;
919 switch(c->dstFormat) {
927 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
928 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
929 dest[aidx]= needAlpha ? A : 255;
936 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
937 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
945 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
962 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
963 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
964 dest[aidx]= needAlpha ? A : 255;
971 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
972 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
980 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
995 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
998 uint8_t *ptr = plane + stride*y;
999 for (i=0; i<height; i++) {
1000 memset(ptr, val, width);
1005 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
1009 for (i = 0; i < width; i++) {
1014 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1018 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1019 const uint8_t *src1, const uint8_t *src2,
1020 long width, uint32_t *unused)
1024 for (i = 0; i < width; i++) {
1025 int r = src1[6*i + 0];
1026 int g = src1[6*i + 2];
1027 int b = src1[6*i + 4];
1029 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1030 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1034 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1035 const uint8_t *src1, const uint8_t *src2,
1036 long width, uint32_t *unused)
1040 for (i = 0; i < width; i++) {
1041 int r= src1[12*i + 0] + src1[12*i + 6];
1042 int g= src1[12*i + 2] + src1[12*i + 8];
1043 int b= src1[12*i + 4] + src1[12*i + 10];
1045 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1046 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1050 static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,
1054 for (i = 0; i < width; i++) {
1059 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1063 static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1064 const uint8_t *src1, const uint8_t *src2,
1065 long width, uint32_t *unused)
1068 for (i = 0; i < width; i++) {
1069 int b = src1[6*i + 0];
1070 int g = src1[6*i + 2];
1071 int r = src1[6*i + 4];
1073 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1074 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1078 static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1079 const uint8_t *src1, const uint8_t *src2,
1080 long width, uint32_t *unused)
1083 for (i = 0; i < width; i++) {
1084 int b= src1[12*i + 0] + src1[12*i + 6];
1085 int g= src1[12*i + 2] + src1[12*i + 8];
1086 int r= src1[12*i + 4] + src1[12*i + 10];
1088 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1089 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1093 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1094 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1097 for (i=0; i<width; i++) {\
1098 int b= (((const type*)src)[i]>>shb)&maskb;\
1099 int g= (((const type*)src)[i]>>shg)&maskg;\
1100 int r= (((const type*)src)[i]>>shr)&maskr;\
1102 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1106 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1107 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1108 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1109 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1110 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1111 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1112 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1113 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1115 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1118 for (i=0; i<width; i++) {
1123 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1124 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1127 for (i=0; i<width; i++) {\
1128 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1129 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1130 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1132 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1133 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1136 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1139 for (i=0; i<width; i++) {\
1140 int pix0= ((const type*)src)[2*i+0]>>shp;\
1141 int pix1= ((const type*)src)[2*i+1]>>shp;\
1142 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1143 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1144 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1145 g&= maskg|(2*maskg);\
1149 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1150 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1154 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1155 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1156 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1157 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1158 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1159 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1160 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1161 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1163 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1166 for (i=0; i<width; i++) {
1169 dst[i]= pal[d] & 0xFF;
1173 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1174 const uint8_t *src1, const uint8_t *src2,
1175 long width, uint32_t *pal)
1178 assert(src1 == src2);
1179 for (i=0; i<width; i++) {
1180 int p= pal[src1[i]];
1187 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1190 for (i=0; i<width/8; i++) {
1193 dst[8*i+j]= ((d>>(7-j))&1)*255;
1197 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1200 for (i=0; i<width/8; i++) {
1203 dst[8*i+j]= ((d>>(7-j))&1)*255;
1207 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1209 #if CONFIG_RUNTIME_CPUDETECT
1210 # define COMPILE_C 1
1212 # define COMPILE_MMX 1
1213 # define COMPILE_MMX2 1
1214 # define COMPILE_3DNOW 1
1216 # define COMPILE_ALTIVEC HAVE_ALTIVEC
1218 #else /* CONFIG_RUNTIME_CPUDETECT */
1221 # define COMPILE_MMX2 1
1222 # elif HAVE_AMD3DNOW
1223 # define COMPILE_3DNOW 1
1225 # define COMPILE_MMX 1
1227 # define COMPILE_C 1
1229 # elif ARCH_PPC && HAVE_ALTIVEC
1230 # define COMPILE_ALTIVEC 1
1232 # define COMPILE_C 1
1237 # define COMPILE_C 0
1240 # define COMPILE_MMX 0
1242 #ifndef COMPILE_MMX2
1243 # define COMPILE_MMX2 0
1245 #ifndef COMPILE_3DNOW
1246 # define COMPILE_3DNOW 0
1248 #ifndef COMPILE_ALTIVEC
1249 # define COMPILE_ALTIVEC 0
1252 #define COMPILE_TEMPLATE_MMX 0
1253 #define COMPILE_TEMPLATE_MMX2 0
1254 #define COMPILE_TEMPLATE_AMD3DNOW 0
1255 #define COMPILE_TEMPLATE_ALTIVEC 0
1258 #define RENAME(a) a ## _C
1259 #include "swscale_template.c"
1264 #undef COMPILE_TEMPLATE_ALTIVEC
1265 #define COMPILE_TEMPLATE_ALTIVEC 1
1266 #define RENAME(a) a ## _altivec
1267 #include "swscale_template.c"
1275 #undef COMPILE_TEMPLATE_MMX
1276 #undef COMPILE_TEMPLATE_MMX2
1277 #undef COMPILE_TEMPLATE_AMD3DNOW
1278 #define COMPILE_TEMPLATE_MMX 1
1279 #define COMPILE_TEMPLATE_MMX2 0
1280 #define COMPILE_TEMPLATE_AMD3DNOW 0
1281 #define RENAME(a) a ## _MMX
1282 #include "swscale_template.c"
1288 #undef COMPILE_TEMPLATE_MMX
1289 #undef COMPILE_TEMPLATE_MMX2
1290 #undef COMPILE_TEMPLATE_AMD3DNOW
1291 #define COMPILE_TEMPLATE_MMX 1
1292 #define COMPILE_TEMPLATE_MMX2 1
1293 #define COMPILE_TEMPLATE_AMD3DNOW 0
1294 #define RENAME(a) a ## _MMX2
1295 #include "swscale_template.c"
1301 #undef COMPILE_TEMPLATE_MMX
1302 #undef COMPILE_TEMPLATE_MMX2
1303 #undef COMPILE_TEMPLATE_AMD3DNOW
1304 #define COMPILE_TEMPLATE_MMX 1
1305 #define COMPILE_TEMPLATE_MMX2 0
1306 #define COMPILE_TEMPLATE_AMD3DNOW 1
1307 #define RENAME(a) a ## _3DNow
1308 #include "swscale_template.c"
1313 SwsFunc ff_getSwsFunc(SwsContext *c)
1315 #if CONFIG_RUNTIME_CPUDETECT
1316 int flags = c->flags;
1319 // ordered per speed fastest first
1320 if (flags & SWS_CPU_CAPS_MMX2) {
1321 sws_init_swScale_MMX2(c);
1322 return swScale_MMX2;
1323 } else if (flags & SWS_CPU_CAPS_3DNOW) {
1324 sws_init_swScale_3DNow(c);
1325 return swScale_3DNow;
1326 } else if (flags & SWS_CPU_CAPS_MMX) {
1327 sws_init_swScale_MMX(c);
1330 sws_init_swScale_C(c);
1336 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1337 sws_init_swScale_altivec(c);
1338 return swScale_altivec;
1340 sws_init_swScale_C(c);
1344 sws_init_swScale_C(c);
1346 #endif /* ARCH_X86 */
1347 #else //CONFIG_RUNTIME_CPUDETECT
1348 #if COMPILE_TEMPLATE_MMX2
1349 sws_init_swScale_MMX2(c);
1350 return swScale_MMX2;
1351 #elif COMPILE_TEMPLATE_AMD3DNOW
1352 sws_init_swScale_3DNow(c);
1353 return swScale_3DNow;
1354 #elif COMPILE_TEMPLATE_MMX
1355 sws_init_swScale_MMX(c);
1357 #elif COMPILE_TEMPLATE_ALTIVEC
1358 sws_init_swScale_altivec(c);
1359 return swScale_altivec;
1361 sws_init_swScale_C(c);
1364 #endif //!CONFIG_RUNTIME_CPUDETECT
1367 static void copyPlane(const uint8_t *src, int srcStride,
1368 int srcSliceY, int srcSliceH, int width,
1369 uint8_t *dst, int dstStride)
1371 dst += dstStride * srcSliceY;
1372 if (dstStride == srcStride && srcStride > 0) {
1373 memcpy(dst, src, srcSliceH * dstStride);
1376 for (i=0; i<srcSliceH; i++) {
1377 memcpy(dst, src, width);
1384 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1385 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1387 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1389 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1390 dstParam[0], dstStride[0]);
1392 if (c->dstFormat == PIX_FMT_NV12)
1393 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1395 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1400 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1401 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1403 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1405 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1410 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1411 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1413 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1415 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1420 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1421 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1423 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1425 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1430 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1431 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1433 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1435 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1440 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1441 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1443 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1444 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1445 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1447 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1450 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1455 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1456 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1458 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1459 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1460 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1462 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1467 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1468 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1470 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1471 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1472 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1474 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1477 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1482 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1483 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1485 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1486 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1487 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1489 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1494 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1497 for (i=0; i<num_pixels; i++)
1498 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1501 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1505 for (i=0; i<num_pixels; i++)
1506 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1509 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1513 for (i=0; i<num_pixels; i++) {
1515 dst[0]= palette[src[i<<1]*4+0];
1516 dst[1]= palette[src[i<<1]*4+1];
1517 dst[2]= palette[src[i<<1]*4+2];
1522 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1523 int srcSliceH, uint8_t* dst[], int dstStride[])
1525 const enum PixelFormat srcFormat= c->srcFormat;
1526 const enum PixelFormat dstFormat= c->dstFormat;
1527 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1528 const uint8_t *palette)=NULL;
1530 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1531 const uint8_t *srcPtr= src[0];
1533 if (srcFormat == PIX_FMT_Y400A) {
1534 switch (dstFormat) {
1535 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1536 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1537 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1538 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1539 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1540 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1542 } else if (usePal(srcFormat)) {
1543 switch (dstFormat) {
1544 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1545 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1546 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1547 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1548 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1549 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1554 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1555 sws_format_name(srcFormat), sws_format_name(dstFormat));
1557 for (i=0; i<srcSliceH; i++) {
1558 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1559 srcPtr+= srcStride[0];
1560 dstPtr+= dstStride[0];
1567 #define isRGBA32(x) ( \
1568 (x) == PIX_FMT_ARGB \
1569 || (x) == PIX_FMT_RGBA \
1570 || (x) == PIX_FMT_BGRA \
1571 || (x) == PIX_FMT_ABGR \
1574 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1575 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1576 int srcSliceH, uint8_t* dst[], int dstStride[])
1578 const enum PixelFormat srcFormat= c->srcFormat;
1579 const enum PixelFormat dstFormat= c->dstFormat;
1580 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1581 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1582 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1583 const int dstId= c->dstFormatBpp >> 2;
1584 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1586 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1588 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1589 if ( CONV_IS(ABGR, RGBA)
1590 || CONV_IS(ARGB, BGRA)
1591 || CONV_IS(BGRA, ARGB)
1592 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1593 else if (CONV_IS(ABGR, ARGB)
1594 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1595 else if (CONV_IS(ABGR, BGRA)
1596 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1597 else if (CONV_IS(BGRA, RGBA)
1598 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1599 else if (CONV_IS(BGRA, ABGR)
1600 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1603 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1604 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1605 switch(srcId | (dstId<<4)) {
1606 case 0x34: conv= rgb16to15; break;
1607 case 0x36: conv= rgb24to15; break;
1608 case 0x38: conv= rgb32to15; break;
1609 case 0x43: conv= rgb15to16; break;
1610 case 0x46: conv= rgb24to16; break;
1611 case 0x48: conv= rgb32to16; break;
1612 case 0x63: conv= rgb15to24; break;
1613 case 0x64: conv= rgb16to24; break;
1614 case 0x68: conv= rgb32to24; break;
1615 case 0x83: conv= rgb15to32; break;
1616 case 0x84: conv= rgb16to32; break;
1617 case 0x86: conv= rgb24to32; break;
1619 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1620 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1621 switch(srcId | (dstId<<4)) {
1622 case 0x33: conv= rgb15tobgr15; break;
1623 case 0x34: conv= rgb16tobgr15; break;
1624 case 0x36: conv= rgb24tobgr15; break;
1625 case 0x38: conv= rgb32tobgr15; break;
1626 case 0x43: conv= rgb15tobgr16; break;
1627 case 0x44: conv= rgb16tobgr16; break;
1628 case 0x46: conv= rgb24tobgr16; break;
1629 case 0x48: conv= rgb32tobgr16; break;
1630 case 0x63: conv= rgb15tobgr24; break;
1631 case 0x64: conv= rgb16tobgr24; break;
1632 case 0x66: conv= rgb24tobgr24; break;
1633 case 0x68: conv= rgb32tobgr24; break;
1634 case 0x83: conv= rgb15tobgr32; break;
1635 case 0x84: conv= rgb16tobgr32; break;
1636 case 0x86: conv= rgb24tobgr32; break;
1641 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1642 sws_format_name(srcFormat), sws_format_name(dstFormat));
1644 const uint8_t *srcPtr= src[0];
1645 uint8_t *dstPtr= dst[0];
1646 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1647 srcPtr += ALT32_CORR;
1649 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1650 dstPtr += ALT32_CORR;
1652 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1653 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1656 dstPtr += dstStride[0]*srcSliceY;
1658 for (i=0; i<srcSliceH; i++) {
1659 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1660 srcPtr+= srcStride[0];
1661 dstPtr+= dstStride[0];
1668 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1669 int srcSliceH, uint8_t* dst[], int dstStride[])
1673 dst[0]+ srcSliceY *dstStride[0],
1674 dst[1]+(srcSliceY>>1)*dstStride[1],
1675 dst[2]+(srcSliceY>>1)*dstStride[2],
1677 dstStride[0], dstStride[1], srcStride[0]);
1679 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1683 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1684 int srcSliceH, uint8_t* dst[], int dstStride[])
1686 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1687 dst[0], dstStride[0]);
1689 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1690 srcSliceH >> 2, srcStride[1], dstStride[1]);
1691 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1692 srcSliceH >> 2, srcStride[2], dstStride[2]);
1694 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1698 /* unscaled copy like stuff (assumes nearly identical formats) */
1699 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1700 int srcSliceH, uint8_t* dst[], int dstStride[])
1702 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1703 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1706 const uint8_t *srcPtr= src[0];
1707 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1710 /* universal length finder */
1711 while(length+c->srcW <= FFABS(dstStride[0])
1712 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1715 for (i=0; i<srcSliceH; i++) {
1716 memcpy(dstPtr, srcPtr, length);
1717 srcPtr+= srcStride[0];
1718 dstPtr+= dstStride[0];
1724 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1725 int srcSliceH, uint8_t* dst[], int dstStride[])
1728 for (plane=0; plane<4; plane++) {
1729 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1730 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1731 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1732 const uint8_t *srcPtr= src[plane];
1733 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1735 if (!dst[plane]) continue;
1736 // ignore palette for GRAY8
1737 if (plane == 1 && !dst[2]) continue;
1738 if (!src[plane] || (plane == 1 && !src[2])) {
1739 if(is16BPS(c->dstFormat))
1741 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1743 if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1744 if (!isBE(c->srcFormat)) srcPtr++;
1745 for (i=0; i<height; i++) {
1746 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1747 srcPtr+= srcStride[plane];
1748 dstPtr+= dstStride[plane];
1750 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1751 for (i=0; i<height; i++) {
1752 for (j=0; j<length; j++) {
1753 dstPtr[ j<<1 ] = srcPtr[j];
1754 dstPtr[(j<<1)+1] = srcPtr[j];
1756 srcPtr+= srcStride[plane];
1757 dstPtr+= dstStride[plane];
1759 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1760 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1762 for (i=0; i<height; i++) {
1763 for (j=0; j<length; j++)
1764 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1765 srcPtr+= srcStride[plane];
1766 dstPtr+= dstStride[plane];
1768 } else if (dstStride[plane] == srcStride[plane] &&
1769 srcStride[plane] > 0 && srcStride[plane] == length) {
1770 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1771 height*dstStride[plane]);
1773 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1775 for (i=0; i<height; i++) {
1776 memcpy(dstPtr, srcPtr, length);
1777 srcPtr+= srcStride[plane];
1778 dstPtr+= dstStride[plane];
1786 int ff_hardcodedcpuflags(void)
1789 #if COMPILE_TEMPLATE_MMX2
1790 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1791 #elif COMPILE_TEMPLATE_AMD3DNOW
1792 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1793 #elif COMPILE_TEMPLATE_MMX
1794 flags |= SWS_CPU_CAPS_MMX;
1795 #elif COMPILE_TEMPLATE_ALTIVEC
1796 flags |= SWS_CPU_CAPS_ALTIVEC;
1798 flags |= SWS_CPU_CAPS_BFIN;
1803 void ff_get_unscaled_swscale(SwsContext *c)
1805 const enum PixelFormat srcFormat = c->srcFormat;
1806 const enum PixelFormat dstFormat = c->dstFormat;
1807 const int flags = c->flags;
1808 const int dstH = c->dstH;
1811 needsDither= isAnyRGB(dstFormat)
1812 && c->dstFormatBpp < 24
1813 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1816 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1817 c->swScale= planarToNv12Wrapper;
1820 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1821 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1822 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1825 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1826 c->swScale= yvu9ToYv12Wrapper;
1830 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1831 c->swScale= bgr24ToYv12Wrapper;
1833 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1834 if ( isAnyRGB(srcFormat)
1835 && isAnyRGB(dstFormat)
1836 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1837 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1838 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1839 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1840 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1841 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1842 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1843 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1844 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1845 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1846 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
1847 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
1848 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1849 c->swScale= rgbToRgbWrapper;
1851 if ((usePal(srcFormat) && (
1852 dstFormat == PIX_FMT_RGB32 ||
1853 dstFormat == PIX_FMT_RGB32_1 ||
1854 dstFormat == PIX_FMT_RGB24 ||
1855 dstFormat == PIX_FMT_BGR32 ||
1856 dstFormat == PIX_FMT_BGR32_1 ||
1857 dstFormat == PIX_FMT_BGR24)))
1858 c->swScale= palToRgbWrapper;
1860 if (srcFormat == PIX_FMT_YUV422P) {
1861 if (dstFormat == PIX_FMT_YUYV422)
1862 c->swScale= yuv422pToYuy2Wrapper;
1863 else if (dstFormat == PIX_FMT_UYVY422)
1864 c->swScale= yuv422pToUyvyWrapper;
1867 /* LQ converters if -sws 0 or -sws 4*/
1868 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1870 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1871 if (dstFormat == PIX_FMT_YUYV422)
1872 c->swScale= planarToYuy2Wrapper;
1873 else if (dstFormat == PIX_FMT_UYVY422)
1874 c->swScale= planarToUyvyWrapper;
1877 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1878 c->swScale= yuyvToYuv420Wrapper;
1879 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1880 c->swScale= uyvyToYuv420Wrapper;
1881 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1882 c->swScale= yuyvToYuv422Wrapper;
1883 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1884 c->swScale= uyvyToYuv422Wrapper;
1887 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1888 !(c->flags & SWS_BITEXACT) &&
1889 srcFormat == PIX_FMT_YUV420P) {
1890 // unscaled YV12 -> packed YUV, we want speed
1891 if (dstFormat == PIX_FMT_YUYV422)
1892 c->swScale= yv12toyuy2_unscaled_altivec;
1893 else if (dstFormat == PIX_FMT_UYVY422)
1894 c->swScale= yv12touyvy_unscaled_altivec;
1899 if ( srcFormat == dstFormat
1900 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1901 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1902 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1903 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1904 || (isGray(dstFormat) && isGray(srcFormat))
1905 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1906 && c->chrDstHSubSample == c->chrSrcHSubSample
1907 && c->chrDstVSubSample == c->chrSrcVSubSample
1908 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1909 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1911 if (isPacked(c->srcFormat))
1912 c->swScale= packedCopyWrapper;
1913 else /* Planar YUV or gray */
1914 c->swScale= planarCopyWrapper;
1917 if (flags & SWS_CPU_CAPS_BFIN)
1918 ff_bfin_get_unscaled_swscale (c);
1922 static void reset_ptr(const uint8_t* src[], int format)
1924 if(!isALPHA(format))
1926 if(!isPlanarYUV(format)) {
1929 if (!usePal(format))
1934 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1935 const int linesizes[4])
1937 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1940 for (i = 0; i < 4; i++) {
1941 int plane = desc->comp[i].plane;
1942 if (!data[plane] || !linesizes[plane])
1950 * swscale wrapper, so we don't need to export the SwsContext.
1951 * Assumes planar YUV to be in YUV order instead of YVU.
1953 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1954 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1957 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1958 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1960 // do not mess up sliceDir if we have a "trailing" 0-size slice
1964 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1965 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1968 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1969 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1973 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1974 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1977 if (c->sliceDir == 0) {
1978 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1981 if (usePal(c->srcFormat)) {
1982 for (i=0; i<256; i++) {
1983 int p, r, g, b,y,u,v;
1984 if(c->srcFormat == PIX_FMT_PAL8) {
1985 p=((const uint32_t*)(src[1]))[i];
1989 } else if(c->srcFormat == PIX_FMT_RGB8) {
1993 } else if(c->srcFormat == PIX_FMT_BGR8) {
1997 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
2001 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
2004 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2009 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2010 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2011 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2012 c->pal_yuv[i]= y + (u<<8) + (v<<16);
2014 switch(c->dstFormat) {
2019 c->pal_rgb[i]= r + (g<<8) + (b<<16);
2021 case PIX_FMT_BGR32_1:
2025 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2027 case PIX_FMT_RGB32_1:
2031 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2038 c->pal_rgb[i]= b + (g<<8) + (r<<16);
2043 // copy strides, so they can safely be modified
2044 if (c->sliceDir == 1) {
2045 // slices go from top to bottom
2046 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2047 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2049 reset_ptr(src2, c->srcFormat);
2050 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2052 /* reset slice direction at end of frame */
2053 if (srcSliceY + srcSliceH == c->srcH)
2056 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2058 // slices go from bottom to top => we flip the image internally
2059 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2060 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2062 src2[0] += (srcSliceH-1)*srcStride[0];
2063 if (!usePal(c->srcFormat))
2064 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2065 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2066 src2[3] += (srcSliceH-1)*srcStride[3];
2067 dst2[0] += ( c->dstH -1)*dstStride[0];
2068 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2069 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2070 dst2[3] += ( c->dstH -1)*dstStride[3];
2072 reset_ptr(src2, c->srcFormat);
2073 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2075 /* reset slice direction at end of frame */
2079 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2083 #if LIBSWSCALE_VERSION_MAJOR < 1
2084 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
2085 int srcSliceH, uint8_t* dst[], int dstStride[])
2087 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2091 /* Convert the palette to the same packed 32-bit format as the palette */
2092 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2096 for (i=0; i<num_pixels; i++)
2097 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2100 /* Palette format: ABCD -> dst format: ABC */
2101 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2105 for (i=0; i<num_pixels; i++) {
2107 dst[0]= palette[src[i]*4+0];
2108 dst[1]= palette[src[i]*4+1];
2109 dst[2]= palette[src[i]*4+2];