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/cpu.h"
66 #include "libavutil/avutil.h"
67 #include "libavutil/mathematics.h"
68 #include "libavutil/bswap.h"
69 #include "libavutil/pixdesc.h"
75 //#define HAVE_AMD3DNOW
80 #define isPacked(x) ( \
82 || (x)==PIX_FMT_YUYV422 \
83 || (x)==PIX_FMT_UYVY422 \
84 || (x)==PIX_FMT_Y400A \
88 #define RGB2YUV_SHIFT 15
89 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
90 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
91 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
92 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
93 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
94 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
95 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
96 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
97 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
99 static const double rgb2yuv_table[8][9]={
100 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
101 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
102 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
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 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
105 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
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 {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
112 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
115 more intelligent misalignment avoidance for the horizontal scaler
116 write special vertical cubic upscale version
117 optimize C code (YV12 / minmax)
118 add support for packed pixel YUV input & output
119 add support for Y8 output
120 optimize BGR24 & BGR32
121 add BGR4 output support
122 write special BGR->BGR scaler
126 DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
127 DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
128 DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
129 DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
130 DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
131 DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
132 DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
133 DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
135 const DECLARE_ALIGNED(8, uint64_t, ff_dither4)[2] = {
136 0x0103010301030103LL,
137 0x0200020002000200LL,};
139 const DECLARE_ALIGNED(8, uint64_t, ff_dither8)[2] = {
140 0x0602060206020602LL,
141 0x0004000400040004LL,};
143 DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
144 DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
145 DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
146 DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
147 DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
148 DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
150 DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
151 DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
152 DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
155 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
156 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
157 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
159 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
160 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
161 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
162 #endif /* FAST_BGR2YV12 */
163 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
164 DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
165 DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
167 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
168 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
169 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
170 DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
171 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
173 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV)[2][4] = {
174 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
175 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
178 DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
180 #endif /* ARCH_X86 */
182 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
183 { 1, 3, 1, 3, 1, 3, 1, 3, },
184 { 2, 0, 2, 0, 2, 0, 2, 0, },
187 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
188 { 6, 2, 6, 2, 6, 2, 6, 2, },
189 { 0, 4, 0, 4, 0, 4, 0, 4, },
192 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
193 { 8, 4, 11, 7, 8, 4, 11, 7, },
194 { 2, 14, 1, 13, 2, 14, 1, 13, },
195 { 10, 6, 9, 5, 10, 6, 9, 5, },
196 { 0, 12, 3, 15, 0, 12, 3, 15, },
199 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
200 { 17, 9, 23, 15, 16, 8, 22, 14, },
201 { 5, 29, 3, 27, 4, 28, 2, 26, },
202 { 21, 13, 19, 11, 20, 12, 18, 10, },
203 { 0, 24, 6, 30, 1, 25, 7, 31, },
204 { 16, 8, 22, 14, 17, 9, 23, 15, },
205 { 4, 28, 2, 26, 5, 29, 3, 27, },
206 { 20, 12, 18, 10, 21, 13, 19, 11, },
207 { 1, 25, 7, 31, 0, 24, 6, 30, },
210 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
211 { 0, 55, 14, 68, 3, 58, 17, 72, },
212 { 37, 18, 50, 32, 40, 22, 54, 35, },
213 { 9, 64, 5, 59, 13, 67, 8, 63, },
214 { 46, 27, 41, 23, 49, 31, 44, 26, },
215 { 2, 57, 16, 71, 1, 56, 15, 70, },
216 { 39, 21, 52, 34, 38, 19, 51, 33, },
217 { 11, 66, 7, 62, 10, 65, 6, 60, },
218 { 48, 30, 43, 25, 47, 29, 42, 24, },
222 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
223 {117, 62, 158, 103, 113, 58, 155, 100, },
224 { 34, 199, 21, 186, 31, 196, 17, 182, },
225 {144, 89, 131, 76, 141, 86, 127, 72, },
226 { 0, 165, 41, 206, 10, 175, 52, 217, },
227 {110, 55, 151, 96, 120, 65, 162, 107, },
228 { 28, 193, 14, 179, 38, 203, 24, 189, },
229 {138, 83, 124, 69, 148, 93, 134, 79, },
230 { 7, 172, 48, 213, 3, 168, 45, 210, },
233 // tries to correct a gamma of 1.5
234 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
235 { 0, 143, 18, 200, 2, 156, 25, 215, },
236 { 78, 28, 125, 64, 89, 36, 138, 74, },
237 { 10, 180, 3, 161, 16, 195, 8, 175, },
238 {109, 51, 93, 38, 121, 60, 105, 47, },
239 { 1, 152, 23, 210, 0, 147, 20, 205, },
240 { 85, 33, 134, 71, 81, 30, 130, 67, },
241 { 14, 190, 6, 171, 12, 185, 5, 166, },
242 {117, 57, 101, 44, 113, 54, 97, 41, },
245 // tries to correct a gamma of 2.0
246 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
247 { 0, 124, 8, 193, 0, 140, 12, 213, },
248 { 55, 14, 104, 42, 66, 19, 119, 52, },
249 { 3, 168, 1, 145, 6, 187, 3, 162, },
250 { 86, 31, 70, 21, 99, 39, 82, 28, },
251 { 0, 134, 11, 206, 0, 129, 9, 200, },
252 { 62, 17, 114, 48, 58, 16, 109, 45, },
253 { 5, 181, 2, 157, 4, 175, 1, 151, },
254 { 95, 36, 78, 26, 90, 34, 74, 24, },
257 // tries to correct a gamma of 2.5
258 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
259 { 0, 107, 3, 187, 0, 125, 6, 212, },
260 { 39, 7, 86, 28, 49, 11, 102, 36, },
261 { 1, 158, 0, 131, 3, 180, 1, 151, },
262 { 68, 19, 52, 12, 81, 25, 64, 17, },
263 { 0, 119, 5, 203, 0, 113, 4, 195, },
264 { 45, 9, 96, 33, 42, 8, 91, 30, },
265 { 2, 172, 1, 144, 2, 165, 0, 137, },
266 { 77, 23, 60, 15, 72, 21, 56, 14, },
270 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
271 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
272 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
273 int dstW, int chrDstW, int big_endian)
275 //FIXME Optimize (just quickly written not optimized..)
278 for (i = 0; i < dstW; i++) {
282 for (j = 0; j < lumFilterSize; j++)
283 val += lumSrc[j][i] * lumFilter[j];
286 AV_WB16(&dest[i], av_clip_uint16(val >> 11));
288 AV_WL16(&dest[i], av_clip_uint16(val >> 11));
293 for (i = 0; i < chrDstW; i++) {
298 for (j = 0; j < chrFilterSize; j++) {
299 u += chrSrc[j][i ] * chrFilter[j];
300 v += chrSrc[j][i + VOFW] * chrFilter[j];
304 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
305 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
307 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
308 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
313 if (CONFIG_SWSCALE_ALPHA && aDest) {
314 for (i = 0; i < dstW; i++) {
318 for (j = 0; j < lumFilterSize; j++)
319 val += alpSrc[j][i] * lumFilter[j];
322 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
324 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
330 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
331 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
332 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
333 enum PixelFormat dstFormat)
335 if (isBE(dstFormat)) {
336 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
337 chrFilter, chrSrc, chrFilterSize,
339 dest, uDest, vDest, aDest,
342 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
343 chrFilter, chrSrc, chrFilterSize,
345 dest, uDest, vDest, aDest,
350 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
351 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
352 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
354 //FIXME Optimize (just quickly written not optimized..)
356 for (i=0; i<dstW; i++) {
359 for (j=0; j<lumFilterSize; j++)
360 val += lumSrc[j][i] * lumFilter[j];
362 dest[i]= av_clip_uint8(val>>19);
366 for (i=0; i<chrDstW; i++) {
370 for (j=0; j<chrFilterSize; j++) {
371 u += chrSrc[j][i] * chrFilter[j];
372 v += chrSrc[j][i + VOFW] * chrFilter[j];
375 uDest[i]= av_clip_uint8(u>>19);
376 vDest[i]= av_clip_uint8(v>>19);
379 if (CONFIG_SWSCALE_ALPHA && aDest)
380 for (i=0; i<dstW; i++) {
383 for (j=0; j<lumFilterSize; j++)
384 val += alpSrc[j][i] * lumFilter[j];
386 aDest[i]= av_clip_uint8(val>>19);
391 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
392 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
393 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
395 //FIXME Optimize (just quickly written not optimized..)
397 for (i=0; i<dstW; i++) {
400 for (j=0; j<lumFilterSize; j++)
401 val += lumSrc[j][i] * lumFilter[j];
403 dest[i]= av_clip_uint8(val>>19);
409 if (dstFormat == PIX_FMT_NV12)
410 for (i=0; i<chrDstW; i++) {
414 for (j=0; j<chrFilterSize; j++) {
415 u += chrSrc[j][i] * chrFilter[j];
416 v += chrSrc[j][i + VOFW] * chrFilter[j];
419 uDest[2*i]= av_clip_uint8(u>>19);
420 uDest[2*i+1]= av_clip_uint8(v>>19);
423 for (i=0; i<chrDstW; i++) {
427 for (j=0; j<chrFilterSize; j++) {
428 u += chrSrc[j][i] * chrFilter[j];
429 v += chrSrc[j][i + VOFW] * chrFilter[j];
432 uDest[2*i]= av_clip_uint8(v>>19);
433 uDest[2*i+1]= av_clip_uint8(u>>19);
437 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
438 for (i=0; i<(dstW>>1); i++) {\
444 int av_unused A1, A2;\
445 type av_unused *r, *b, *g;\
448 for (j=0; j<lumFilterSize; j++) {\
449 Y1 += lumSrc[j][i2] * lumFilter[j];\
450 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
452 for (j=0; j<chrFilterSize; j++) {\
453 U += chrSrc[j][i] * chrFilter[j];\
454 V += chrSrc[j][i+VOFW] * chrFilter[j];\
463 for (j=0; j<lumFilterSize; j++) {\
464 A1 += alpSrc[j][i2 ] * lumFilter[j];\
465 A2 += alpSrc[j][i2+1] * lumFilter[j];\
471 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
472 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
473 if ((Y1|Y2|U|V)&256) {\
474 if (Y1>255) Y1=255; \
475 else if (Y1<0)Y1=0; \
476 if (Y2>255) Y2=255; \
477 else if (Y2<0)Y2=0; \
483 if (alpha && ((A1|A2)&256)) {\
484 A1=av_clip_uint8(A1);\
485 A2=av_clip_uint8(A2);\
488 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
489 for (i=0; i<dstW; i++) {\
497 for (j=0; j<lumFilterSize; j++) {\
498 Y += lumSrc[j][i ] * lumFilter[j];\
500 for (j=0; j<chrFilterSize; j++) {\
501 U += chrSrc[j][i ] * chrFilter[j];\
502 V += chrSrc[j][i+VOFW] * chrFilter[j];\
509 for (j=0; j<lumFilterSize; j++)\
510 A += alpSrc[j][i ] * lumFilter[j];\
513 A = av_clip_uint8(A);\
516 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
517 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
518 Y-= c->yuv2rgb_y_offset;\
519 Y*= c->yuv2rgb_y_coeff;\
521 R= Y + V*c->yuv2rgb_v2r_coeff;\
522 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
523 B= Y + U*c->yuv2rgb_u2b_coeff;\
524 if ((R|G|B)&(0xC0000000)) {\
525 if (R>=(256<<22)) R=(256<<22)-1; \
527 if (G>=(256<<22)) G=(256<<22)-1; \
529 if (B>=(256<<22)) B=(256<<22)-1; \
533 #define YSCALE_YUV_2_GRAY16_C \
534 for (i=0; i<(dstW>>1); i++) {\
543 for (j=0; j<lumFilterSize; j++) {\
544 Y1 += lumSrc[j][i2] * lumFilter[j];\
545 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
549 if ((Y1|Y2|U|V)&65536) {\
550 if (Y1>65535) Y1=65535; \
551 else if (Y1<0)Y1=0; \
552 if (Y2>65535) Y2=65535; \
553 else if (Y2<0)Y2=0; \
556 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
557 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
558 r = (type *)c->table_rV[V]; \
559 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
560 b = (type *)c->table_bU[U];
562 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
563 for (i=0; i<(dstW>>1); i++) { \
565 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
566 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
567 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
568 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
569 type av_unused *r, *b, *g; \
570 int av_unused A1, A2; \
572 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
573 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
576 #define YSCALE_YUV_2_GRAY16_2_C \
577 for (i=0; i<(dstW>>1); i++) { \
579 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
580 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
582 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
583 YSCALE_YUV_2_PACKED2_C(type,alpha)\
584 r = (type *)c->table_rV[V];\
585 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
586 b = (type *)c->table_bU[U];
588 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
589 for (i=0; i<(dstW>>1); i++) {\
591 int Y1= buf0[i2 ]>>7;\
592 int Y2= buf0[i2+1]>>7;\
593 int U= (uvbuf1[i ])>>7;\
594 int V= (uvbuf1[i+VOFW])>>7;\
595 type av_unused *r, *b, *g;\
596 int av_unused A1, A2;\
602 #define YSCALE_YUV_2_GRAY16_1_C \
603 for (i=0; i<(dstW>>1); i++) {\
605 int Y1= buf0[i2 ]<<1;\
606 int Y2= buf0[i2+1]<<1;
608 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
609 YSCALE_YUV_2_PACKED1_C(type,alpha)\
610 r = (type *)c->table_rV[V];\
611 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
612 b = (type *)c->table_bU[U];
614 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
615 for (i=0; i<(dstW>>1); i++) {\
617 int Y1= buf0[i2 ]>>7;\
618 int Y2= buf0[i2+1]>>7;\
619 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
620 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
621 type av_unused *r, *b, *g;\
622 int av_unused A1, A2;\
628 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
629 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
630 r = (type *)c->table_rV[V];\
631 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
632 b = (type *)c->table_bU[U];
634 #define YSCALE_YUV_2_MONO2_C \
635 const uint8_t * const d128=dither_8x8_220[y&7];\
636 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
637 for (i=0; i<dstW-7; i+=8) {\
639 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
640 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
641 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
642 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
643 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
644 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
645 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
646 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
647 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
651 #define YSCALE_YUV_2_MONOX_C \
652 const uint8_t * const d128=dither_8x8_220[y&7];\
653 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
655 for (i=0; i<dstW-1; i+=2) {\
660 for (j=0; j<lumFilterSize; j++) {\
661 Y1 += lumSrc[j][i] * lumFilter[j];\
662 Y2 += lumSrc[j][i+1] * lumFilter[j];\
672 acc+= acc + g[Y1+d128[(i+0)&7]];\
673 acc+= acc + g[Y2+d128[(i+1)&7]];\
675 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
680 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
681 switch(c->dstFormat) {\
682 case PIX_FMT_RGB48BE:\
683 case PIX_FMT_RGB48LE:\
685 ((uint8_t*)dest)[ 0]= r[Y1];\
686 ((uint8_t*)dest)[ 1]= r[Y1];\
687 ((uint8_t*)dest)[ 2]= g[Y1];\
688 ((uint8_t*)dest)[ 3]= g[Y1];\
689 ((uint8_t*)dest)[ 4]= b[Y1];\
690 ((uint8_t*)dest)[ 5]= b[Y1];\
691 ((uint8_t*)dest)[ 6]= r[Y2];\
692 ((uint8_t*)dest)[ 7]= r[Y2];\
693 ((uint8_t*)dest)[ 8]= g[Y2];\
694 ((uint8_t*)dest)[ 9]= g[Y2];\
695 ((uint8_t*)dest)[10]= b[Y2];\
696 ((uint8_t*)dest)[11]= b[Y2];\
700 case PIX_FMT_BGR48BE:\
701 case PIX_FMT_BGR48LE:\
703 ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
704 ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
705 ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
706 ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
707 ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
708 ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
715 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
716 func(uint32_t,needAlpha)\
717 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
718 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
721 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
723 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
724 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
728 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
729 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
737 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
738 func(uint32_t,needAlpha)\
739 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
740 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
743 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
745 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
746 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
750 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
751 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
758 ((uint8_t*)dest)[0]= r[Y1];\
759 ((uint8_t*)dest)[1]= g[Y1];\
760 ((uint8_t*)dest)[2]= b[Y1];\
761 ((uint8_t*)dest)[3]= r[Y2];\
762 ((uint8_t*)dest)[4]= g[Y2];\
763 ((uint8_t*)dest)[5]= b[Y2];\
769 ((uint8_t*)dest)[0]= b[Y1];\
770 ((uint8_t*)dest)[1]= g[Y1];\
771 ((uint8_t*)dest)[2]= r[Y1];\
772 ((uint8_t*)dest)[3]= b[Y2];\
773 ((uint8_t*)dest)[4]= g[Y2];\
774 ((uint8_t*)dest)[5]= r[Y2];\
778 case PIX_FMT_RGB565BE:\
779 case PIX_FMT_RGB565LE:\
780 case PIX_FMT_BGR565BE:\
781 case PIX_FMT_BGR565LE:\
783 const int dr1= dither_2x2_8[y&1 ][0];\
784 const int dg1= dither_2x2_4[y&1 ][0];\
785 const int db1= dither_2x2_8[(y&1)^1][0];\
786 const int dr2= dither_2x2_8[y&1 ][1];\
787 const int dg2= dither_2x2_4[y&1 ][1];\
788 const int db2= dither_2x2_8[(y&1)^1][1];\
790 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
791 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
795 case PIX_FMT_RGB555BE:\
796 case PIX_FMT_RGB555LE:\
797 case PIX_FMT_BGR555BE:\
798 case PIX_FMT_BGR555LE:\
800 const int dr1= dither_2x2_8[y&1 ][0];\
801 const int dg1= dither_2x2_8[y&1 ][1];\
802 const int db1= dither_2x2_8[(y&1)^1][0];\
803 const int dr2= dither_2x2_8[y&1 ][1];\
804 const int dg2= dither_2x2_8[y&1 ][0];\
805 const int db2= dither_2x2_8[(y&1)^1][1];\
807 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
808 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
812 case PIX_FMT_RGB444BE:\
813 case PIX_FMT_RGB444LE:\
814 case PIX_FMT_BGR444BE:\
815 case PIX_FMT_BGR444LE:\
817 const int dr1= dither_4x4_16[y&3 ][0];\
818 const int dg1= dither_4x4_16[y&3 ][1];\
819 const int db1= dither_4x4_16[(y&3)^3][0];\
820 const int dr2= dither_4x4_16[y&3 ][1];\
821 const int dg2= dither_4x4_16[y&3 ][0];\
822 const int db2= dither_4x4_16[(y&3)^3][1];\
824 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
825 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
832 const uint8_t * const d64= dither_8x8_73[y&7];\
833 const uint8_t * const d32= dither_8x8_32[y&7];\
835 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
836 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
843 const uint8_t * const d64= dither_8x8_73 [y&7];\
844 const uint8_t * const d128=dither_8x8_220[y&7];\
846 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
847 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
851 case PIX_FMT_RGB4_BYTE:\
852 case PIX_FMT_BGR4_BYTE:\
854 const uint8_t * const d64= dither_8x8_73 [y&7];\
855 const uint8_t * const d128=dither_8x8_220[y&7];\
857 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
858 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
862 case PIX_FMT_MONOBLACK:\
863 case PIX_FMT_MONOWHITE:\
868 case PIX_FMT_YUYV422:\
870 ((uint8_t*)dest)[2*i2+0]= Y1;\
871 ((uint8_t*)dest)[2*i2+1]= U;\
872 ((uint8_t*)dest)[2*i2+2]= Y2;\
873 ((uint8_t*)dest)[2*i2+3]= V;\
876 case PIX_FMT_UYVY422:\
878 ((uint8_t*)dest)[2*i2+0]= U;\
879 ((uint8_t*)dest)[2*i2+1]= Y1;\
880 ((uint8_t*)dest)[2*i2+2]= V;\
881 ((uint8_t*)dest)[2*i2+3]= Y2;\
884 case PIX_FMT_GRAY16BE:\
886 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
887 ((uint8_t*)dest)[2*i2+1]= Y1;\
888 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
889 ((uint8_t*)dest)[2*i2+3]= Y2;\
892 case PIX_FMT_GRAY16LE:\
894 ((uint8_t*)dest)[2*i2+0]= Y1;\
895 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
896 ((uint8_t*)dest)[2*i2+2]= Y2;\
897 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
902 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
903 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
904 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
907 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)
910 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
911 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
912 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
915 int step= c->dstFormatBpp/8;
918 switch(c->dstFormat) {
926 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
927 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
928 dest[aidx]= needAlpha ? A : 255;
935 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
936 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
944 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
961 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
962 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
963 dest[aidx]= needAlpha ? A : 255;
970 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
971 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
979 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
994 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
997 uint8_t *ptr = plane + stride*y;
998 for (i=0; i<height; i++) {
999 memset(ptr, val, width);
1004 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
1008 for (i = 0; i < width; i++) {
1013 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1017 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1018 const uint8_t *src1, const uint8_t *src2,
1019 long width, uint32_t *unused)
1023 for (i = 0; i < width; i++) {
1024 int r = src1[6*i + 0];
1025 int g = src1[6*i + 2];
1026 int b = src1[6*i + 4];
1028 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1029 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1033 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1034 const uint8_t *src1, const uint8_t *src2,
1035 long width, uint32_t *unused)
1039 for (i = 0; i < width; i++) {
1040 int r= src1[12*i + 0] + src1[12*i + 6];
1041 int g= src1[12*i + 2] + src1[12*i + 8];
1042 int b= src1[12*i + 4] + src1[12*i + 10];
1044 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1045 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1049 static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,
1053 for (i = 0; i < width; i++) {
1058 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1062 static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1063 const uint8_t *src1, const uint8_t *src2,
1064 long width, uint32_t *unused)
1067 for (i = 0; i < width; i++) {
1068 int b = src1[6*i + 0];
1069 int g = src1[6*i + 2];
1070 int r = src1[6*i + 4];
1072 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1073 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1077 static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1078 const uint8_t *src1, const uint8_t *src2,
1079 long width, uint32_t *unused)
1082 for (i = 0; i < width; i++) {
1083 int b= src1[12*i + 0] + src1[12*i + 6];
1084 int g= src1[12*i + 2] + src1[12*i + 8];
1085 int r= src1[12*i + 4] + src1[12*i + 10];
1087 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1088 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1092 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1093 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1096 for (i=0; i<width; i++) {\
1097 int b= (((const type*)src)[i]>>shb)&maskb;\
1098 int g= (((const type*)src)[i]>>shg)&maskg;\
1099 int r= (((const type*)src)[i]>>shr)&maskr;\
1101 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1105 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1106 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1107 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1108 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1109 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1110 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1111 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1112 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1114 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1117 for (i=0; i<width; i++) {
1122 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1123 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1126 for (i=0; i<width; i++) {\
1127 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1128 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1129 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1131 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1132 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1135 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1138 for (i=0; i<width; i++) {\
1139 int pix0= ((const type*)src)[2*i+0]>>shp;\
1140 int pix1= ((const type*)src)[2*i+1]>>shp;\
1141 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1142 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1143 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1144 g&= maskg|(2*maskg);\
1148 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1149 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1153 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1154 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1155 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1156 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1157 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1158 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1159 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1160 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1162 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1165 for (i=0; i<width; i++) {
1168 dst[i]= pal[d] & 0xFF;
1172 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1173 const uint8_t *src1, const uint8_t *src2,
1174 long width, uint32_t *pal)
1177 assert(src1 == src2);
1178 for (i=0; i<width; i++) {
1179 int p= pal[src1[i]];
1186 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1189 for (i=0; i<width/8; i++) {
1192 dst[8*i+j]= ((d>>(7-j))&1)*255;
1196 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1199 for (i=0; i<width/8; i++) {
1202 dst[8*i+j]= ((d>>(7-j))&1)*255;
1206 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1208 #if CONFIG_RUNTIME_CPUDETECT
1209 # define COMPILE_C 1
1211 # define COMPILE_MMX 1
1212 # define COMPILE_MMX2 1
1213 # define COMPILE_3DNOW 1
1215 # define COMPILE_ALTIVEC HAVE_ALTIVEC
1217 #else /* CONFIG_RUNTIME_CPUDETECT */
1220 # define COMPILE_MMX2 1
1221 # elif HAVE_AMD3DNOW
1222 # define COMPILE_3DNOW 1
1224 # define COMPILE_MMX 1
1226 # define COMPILE_C 1
1228 # elif ARCH_PPC && HAVE_ALTIVEC
1229 # define COMPILE_ALTIVEC 1
1231 # define COMPILE_C 1
1236 # define COMPILE_C 0
1239 # define COMPILE_MMX 0
1241 #ifndef COMPILE_MMX2
1242 # define COMPILE_MMX2 0
1244 #ifndef COMPILE_3DNOW
1245 # define COMPILE_3DNOW 0
1247 #ifndef COMPILE_ALTIVEC
1248 # define COMPILE_ALTIVEC 0
1251 #define COMPILE_TEMPLATE_MMX 0
1252 #define COMPILE_TEMPLATE_MMX2 0
1253 #define COMPILE_TEMPLATE_AMD3DNOW 0
1254 #define COMPILE_TEMPLATE_ALTIVEC 0
1257 #define RENAME(a) a ## _C
1258 #include "swscale_template.c"
1263 #undef COMPILE_TEMPLATE_ALTIVEC
1264 #define COMPILE_TEMPLATE_ALTIVEC 1
1265 #define RENAME(a) a ## _altivec
1266 #include "swscale_template.c"
1274 #undef COMPILE_TEMPLATE_MMX
1275 #undef COMPILE_TEMPLATE_MMX2
1276 #undef COMPILE_TEMPLATE_AMD3DNOW
1277 #define COMPILE_TEMPLATE_MMX 1
1278 #define COMPILE_TEMPLATE_MMX2 0
1279 #define COMPILE_TEMPLATE_AMD3DNOW 0
1280 #define RENAME(a) a ## _MMX
1281 #include "swscale_template.c"
1287 #undef COMPILE_TEMPLATE_MMX
1288 #undef COMPILE_TEMPLATE_MMX2
1289 #undef COMPILE_TEMPLATE_AMD3DNOW
1290 #define COMPILE_TEMPLATE_MMX 1
1291 #define COMPILE_TEMPLATE_MMX2 1
1292 #define COMPILE_TEMPLATE_AMD3DNOW 0
1293 #define RENAME(a) a ## _MMX2
1294 #include "swscale_template.c"
1300 #undef COMPILE_TEMPLATE_MMX
1301 #undef COMPILE_TEMPLATE_MMX2
1302 #undef COMPILE_TEMPLATE_AMD3DNOW
1303 #define COMPILE_TEMPLATE_MMX 1
1304 #define COMPILE_TEMPLATE_MMX2 0
1305 #define COMPILE_TEMPLATE_AMD3DNOW 1
1306 #define RENAME(a) a ## _3DNow
1307 #include "swscale_template.c"
1312 SwsFunc ff_getSwsFunc(SwsContext *c)
1314 #if CONFIG_RUNTIME_CPUDETECT
1315 int flags = c->flags;
1317 int cpuflags = av_get_cpu_flags();
1319 flags |= (cpuflags & AV_CPU_FLAG_MMX ? SWS_CPU_CAPS_MMX : 0);
1320 flags |= (cpuflags & AV_CPU_FLAG_MMX2 ? SWS_CPU_CAPS_MMX2 : 0);
1321 flags |= (cpuflags & AV_CPU_FLAG_3DNOW ? SWS_CPU_CAPS_3DNOW : 0);
1324 // ordered per speed fastest first
1325 if (flags & SWS_CPU_CAPS_MMX2) {
1326 sws_init_swScale_MMX2(c);
1327 return swScale_MMX2;
1328 } else if (flags & SWS_CPU_CAPS_3DNOW) {
1329 sws_init_swScale_3DNow(c);
1330 return swScale_3DNow;
1331 } else if (flags & SWS_CPU_CAPS_MMX) {
1332 sws_init_swScale_MMX(c);
1335 sws_init_swScale_C(c);
1341 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1342 sws_init_swScale_altivec(c);
1343 return swScale_altivec;
1345 sws_init_swScale_C(c);
1349 sws_init_swScale_C(c);
1351 #endif /* ARCH_X86 */
1352 #else //CONFIG_RUNTIME_CPUDETECT
1353 #if COMPILE_TEMPLATE_MMX2
1354 sws_init_swScale_MMX2(c);
1355 return swScale_MMX2;
1356 #elif COMPILE_TEMPLATE_AMD3DNOW
1357 sws_init_swScale_3DNow(c);
1358 return swScale_3DNow;
1359 #elif COMPILE_TEMPLATE_MMX
1360 sws_init_swScale_MMX(c);
1362 #elif COMPILE_TEMPLATE_ALTIVEC
1363 sws_init_swScale_altivec(c);
1364 return swScale_altivec;
1366 sws_init_swScale_C(c);
1369 #endif //!CONFIG_RUNTIME_CPUDETECT
1372 static void copyPlane(const uint8_t *src, int srcStride,
1373 int srcSliceY, int srcSliceH, int width,
1374 uint8_t *dst, int dstStride)
1376 dst += dstStride * srcSliceY;
1377 if (dstStride == srcStride && srcStride > 0) {
1378 memcpy(dst, src, srcSliceH * dstStride);
1381 for (i=0; i<srcSliceH; i++) {
1382 memcpy(dst, src, width);
1389 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1390 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1392 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1394 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1395 dstParam[0], dstStride[0]);
1397 if (c->dstFormat == PIX_FMT_NV12)
1398 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1400 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1405 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1406 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1408 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1410 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1415 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1416 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1418 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1420 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1425 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1426 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1428 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1430 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1435 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1436 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1438 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1440 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1445 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1446 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1448 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1449 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1450 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1452 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1455 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1460 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1461 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1463 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1464 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1465 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1467 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1472 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1473 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1475 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1476 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1477 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1479 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1482 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1487 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1488 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1490 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1491 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1492 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1494 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1499 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1502 for (i=0; i<num_pixels; i++)
1503 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1506 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1510 for (i=0; i<num_pixels; i++)
1511 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1514 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1518 for (i=0; i<num_pixels; i++) {
1520 dst[0]= palette[src[i<<1]*4+0];
1521 dst[1]= palette[src[i<<1]*4+1];
1522 dst[2]= palette[src[i<<1]*4+2];
1527 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1528 int srcSliceH, uint8_t* dst[], int dstStride[])
1530 const enum PixelFormat srcFormat= c->srcFormat;
1531 const enum PixelFormat dstFormat= c->dstFormat;
1532 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1533 const uint8_t *palette)=NULL;
1535 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1536 const uint8_t *srcPtr= src[0];
1538 if (srcFormat == PIX_FMT_Y400A) {
1539 switch (dstFormat) {
1540 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1541 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1542 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1543 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1544 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1545 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1547 } else if (usePal(srcFormat)) {
1548 switch (dstFormat) {
1549 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1550 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1551 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1552 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1553 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1554 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1559 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1560 sws_format_name(srcFormat), sws_format_name(dstFormat));
1562 for (i=0; i<srcSliceH; i++) {
1563 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1564 srcPtr+= srcStride[0];
1565 dstPtr+= dstStride[0];
1572 #define isRGBA32(x) ( \
1573 (x) == PIX_FMT_ARGB \
1574 || (x) == PIX_FMT_RGBA \
1575 || (x) == PIX_FMT_BGRA \
1576 || (x) == PIX_FMT_ABGR \
1579 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1580 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1581 int srcSliceH, uint8_t* dst[], int dstStride[])
1583 const enum PixelFormat srcFormat= c->srcFormat;
1584 const enum PixelFormat dstFormat= c->dstFormat;
1585 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1586 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1587 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1588 const int dstId= c->dstFormatBpp >> 2;
1589 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1591 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1593 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1594 if ( CONV_IS(ABGR, RGBA)
1595 || CONV_IS(ARGB, BGRA)
1596 || CONV_IS(BGRA, ARGB)
1597 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1598 else if (CONV_IS(ABGR, ARGB)
1599 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1600 else if (CONV_IS(ABGR, BGRA)
1601 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1602 else if (CONV_IS(BGRA, RGBA)
1603 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1604 else if (CONV_IS(BGRA, ABGR)
1605 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1608 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1609 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1610 switch(srcId | (dstId<<4)) {
1611 case 0x34: conv= rgb16to15; break;
1612 case 0x36: conv= rgb24to15; break;
1613 case 0x38: conv= rgb32to15; break;
1614 case 0x43: conv= rgb15to16; break;
1615 case 0x46: conv= rgb24to16; break;
1616 case 0x48: conv= rgb32to16; break;
1617 case 0x63: conv= rgb15to24; break;
1618 case 0x64: conv= rgb16to24; break;
1619 case 0x68: conv= rgb32to24; break;
1620 case 0x83: conv= rgb15to32; break;
1621 case 0x84: conv= rgb16to32; break;
1622 case 0x86: conv= rgb24to32; break;
1624 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1625 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1626 switch(srcId | (dstId<<4)) {
1627 case 0x33: conv= rgb15tobgr15; break;
1628 case 0x34: conv= rgb16tobgr15; break;
1629 case 0x36: conv= rgb24tobgr15; break;
1630 case 0x38: conv= rgb32tobgr15; break;
1631 case 0x43: conv= rgb15tobgr16; break;
1632 case 0x44: conv= rgb16tobgr16; break;
1633 case 0x46: conv= rgb24tobgr16; break;
1634 case 0x48: conv= rgb32tobgr16; break;
1635 case 0x63: conv= rgb15tobgr24; break;
1636 case 0x64: conv= rgb16tobgr24; break;
1637 case 0x66: conv= rgb24tobgr24; break;
1638 case 0x68: conv= rgb32tobgr24; break;
1639 case 0x83: conv= rgb15tobgr32; break;
1640 case 0x84: conv= rgb16tobgr32; break;
1641 case 0x86: conv= rgb24tobgr32; break;
1646 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1647 sws_format_name(srcFormat), sws_format_name(dstFormat));
1649 const uint8_t *srcPtr= src[0];
1650 uint8_t *dstPtr= dst[0];
1651 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1652 srcPtr += ALT32_CORR;
1654 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1655 dstPtr += ALT32_CORR;
1657 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1658 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1661 dstPtr += dstStride[0]*srcSliceY;
1663 for (i=0; i<srcSliceH; i++) {
1664 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1665 srcPtr+= srcStride[0];
1666 dstPtr+= dstStride[0];
1673 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1674 int srcSliceH, uint8_t* dst[], int dstStride[])
1678 dst[0]+ srcSliceY *dstStride[0],
1679 dst[1]+(srcSliceY>>1)*dstStride[1],
1680 dst[2]+(srcSliceY>>1)*dstStride[2],
1682 dstStride[0], dstStride[1], srcStride[0]);
1684 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1688 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1689 int srcSliceH, uint8_t* dst[], int dstStride[])
1691 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1692 dst[0], dstStride[0]);
1694 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1695 srcSliceH >> 2, srcStride[1], dstStride[1]);
1696 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1697 srcSliceH >> 2, srcStride[2], dstStride[2]);
1699 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1703 /* unscaled copy like stuff (assumes nearly identical formats) */
1704 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1705 int srcSliceH, uint8_t* dst[], int dstStride[])
1707 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1708 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1711 const uint8_t *srcPtr= src[0];
1712 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1715 /* universal length finder */
1716 while(length+c->srcW <= FFABS(dstStride[0])
1717 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1720 for (i=0; i<srcSliceH; i++) {
1721 memcpy(dstPtr, srcPtr, length);
1722 srcPtr+= srcStride[0];
1723 dstPtr+= dstStride[0];
1729 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1730 int srcSliceH, uint8_t* dst[], int dstStride[])
1733 for (plane=0; plane<4; plane++) {
1734 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1735 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1736 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1737 const uint8_t *srcPtr= src[plane];
1738 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1740 if (!dst[plane]) continue;
1741 // ignore palette for GRAY8
1742 if (plane == 1 && !dst[2]) continue;
1743 if (!src[plane] || (plane == 1 && !src[2])) {
1744 if(is16BPS(c->dstFormat))
1746 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1748 if(isNBPS(c->srcFormat)) {
1749 const int depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
1750 uint16_t *srcPtr2 = (uint16_t*)srcPtr;
1752 if (is16BPS(c->dstFormat)) {
1753 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1754 for (i = 0; i < height; i++) {
1755 for (j = 0; j < length; j++)
1756 dstPtr2[j] = (srcPtr2[j]<<(16-depth)) | (srcPtr2[j]>>(2*depth-16));
1757 dstPtr2 += dstStride[plane]/2;
1758 srcPtr2 += srcStride[plane]/2;
1761 // FIXME Maybe dither instead.
1762 for (i = 0; i < height; i++) {
1763 for (j = 0; j < length; j++)
1764 dstPtr[j] = srcPtr2[j]>>(depth-8);
1765 dstPtr += dstStride[plane];
1766 srcPtr2 += srcStride[plane]/2;
1769 } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1770 if (!isBE(c->srcFormat)) srcPtr++;
1771 for (i=0; i<height; i++) {
1772 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1773 srcPtr+= srcStride[plane];
1774 dstPtr+= dstStride[plane];
1776 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1777 for (i=0; i<height; i++) {
1778 for (j=0; j<length; j++) {
1779 dstPtr[ j<<1 ] = srcPtr[j];
1780 dstPtr[(j<<1)+1] = srcPtr[j];
1782 srcPtr+= srcStride[plane];
1783 dstPtr+= dstStride[plane];
1785 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1786 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1788 for (i=0; i<height; i++) {
1789 for (j=0; j<length; j++)
1790 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1791 srcPtr+= srcStride[plane];
1792 dstPtr+= dstStride[plane];
1794 } else if (dstStride[plane] == srcStride[plane] &&
1795 srcStride[plane] > 0 && srcStride[plane] == length) {
1796 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1797 height*dstStride[plane]);
1799 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1801 for (i=0; i<height; i++) {
1802 memcpy(dstPtr, srcPtr, length);
1803 srcPtr+= srcStride[plane];
1804 dstPtr+= dstStride[plane];
1812 int ff_hardcodedcpuflags(void)
1815 #if COMPILE_TEMPLATE_MMX2
1816 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1817 #elif COMPILE_TEMPLATE_AMD3DNOW
1818 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1819 #elif COMPILE_TEMPLATE_MMX
1820 flags |= SWS_CPU_CAPS_MMX;
1821 #elif COMPILE_TEMPLATE_ALTIVEC
1822 flags |= SWS_CPU_CAPS_ALTIVEC;
1824 flags |= SWS_CPU_CAPS_BFIN;
1829 void ff_get_unscaled_swscale(SwsContext *c)
1831 const enum PixelFormat srcFormat = c->srcFormat;
1832 const enum PixelFormat dstFormat = c->dstFormat;
1833 const int flags = c->flags;
1834 const int dstH = c->dstH;
1837 needsDither= isAnyRGB(dstFormat)
1838 && c->dstFormatBpp < 24
1839 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1842 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1843 c->swScale= planarToNv12Wrapper;
1846 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1847 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1848 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1851 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1852 c->swScale= yvu9ToYv12Wrapper;
1856 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1857 c->swScale= bgr24ToYv12Wrapper;
1859 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1860 if ( isAnyRGB(srcFormat)
1861 && isAnyRGB(dstFormat)
1862 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1863 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1864 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1865 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1866 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1867 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1868 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1869 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1870 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1871 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1872 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
1873 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
1874 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1875 c->swScale= rgbToRgbWrapper;
1877 if ((usePal(srcFormat) && (
1878 dstFormat == PIX_FMT_RGB32 ||
1879 dstFormat == PIX_FMT_RGB32_1 ||
1880 dstFormat == PIX_FMT_RGB24 ||
1881 dstFormat == PIX_FMT_BGR32 ||
1882 dstFormat == PIX_FMT_BGR32_1 ||
1883 dstFormat == PIX_FMT_BGR24)))
1884 c->swScale= palToRgbWrapper;
1886 if (srcFormat == PIX_FMT_YUV422P) {
1887 if (dstFormat == PIX_FMT_YUYV422)
1888 c->swScale= yuv422pToYuy2Wrapper;
1889 else if (dstFormat == PIX_FMT_UYVY422)
1890 c->swScale= yuv422pToUyvyWrapper;
1893 /* LQ converters if -sws 0 or -sws 4*/
1894 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1896 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1897 if (dstFormat == PIX_FMT_YUYV422)
1898 c->swScale= planarToYuy2Wrapper;
1899 else if (dstFormat == PIX_FMT_UYVY422)
1900 c->swScale= planarToUyvyWrapper;
1903 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1904 c->swScale= yuyvToYuv420Wrapper;
1905 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1906 c->swScale= uyvyToYuv420Wrapper;
1907 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1908 c->swScale= yuyvToYuv422Wrapper;
1909 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1910 c->swScale= uyvyToYuv422Wrapper;
1913 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1914 !(c->flags & SWS_BITEXACT) &&
1915 srcFormat == PIX_FMT_YUV420P) {
1916 // unscaled YV12 -> packed YUV, we want speed
1917 if (dstFormat == PIX_FMT_YUYV422)
1918 c->swScale= yv12toyuy2_unscaled_altivec;
1919 else if (dstFormat == PIX_FMT_UYVY422)
1920 c->swScale= yv12touyvy_unscaled_altivec;
1925 if ( srcFormat == dstFormat
1926 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1927 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1928 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1929 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1930 || (isGray(dstFormat) && isGray(srcFormat))
1931 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1932 && c->chrDstHSubSample == c->chrSrcHSubSample
1933 && c->chrDstVSubSample == c->chrSrcVSubSample
1934 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1935 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1937 if (isPacked(c->srcFormat))
1938 c->swScale= packedCopyWrapper;
1939 else /* Planar YUV or gray */
1940 c->swScale= planarCopyWrapper;
1943 if (flags & SWS_CPU_CAPS_BFIN)
1944 ff_bfin_get_unscaled_swscale (c);
1948 static void reset_ptr(const uint8_t* src[], int format)
1950 if(!isALPHA(format))
1952 if(!isPlanarYUV(format)) {
1955 if (!usePal(format))
1960 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1961 const int linesizes[4])
1963 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1966 for (i = 0; i < 4; i++) {
1967 int plane = desc->comp[i].plane;
1968 if (!data[plane] || !linesizes[plane])
1976 * swscale wrapper, so we don't need to export the SwsContext.
1977 * Assumes planar YUV to be in YUV order instead of YVU.
1979 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1980 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1983 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1984 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1986 // do not mess up sliceDir if we have a "trailing" 0-size slice
1990 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1991 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1994 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1995 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1999 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2000 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2003 if (c->sliceDir == 0) {
2004 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2007 if (usePal(c->srcFormat)) {
2008 for (i=0; i<256; i++) {
2009 int p, r, g, b,y,u,v;
2010 if(c->srcFormat == PIX_FMT_PAL8) {
2011 p=((const uint32_t*)(src[1]))[i];
2015 } else if(c->srcFormat == PIX_FMT_RGB8) {
2019 } else if(c->srcFormat == PIX_FMT_BGR8) {
2023 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
2027 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
2030 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2035 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2036 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2037 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2038 c->pal_yuv[i]= y + (u<<8) + (v<<16);
2040 switch(c->dstFormat) {
2045 c->pal_rgb[i]= r + (g<<8) + (b<<16);
2047 case PIX_FMT_BGR32_1:
2051 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2053 case PIX_FMT_RGB32_1:
2057 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2064 c->pal_rgb[i]= b + (g<<8) + (r<<16);
2069 // copy strides, so they can safely be modified
2070 if (c->sliceDir == 1) {
2071 // slices go from top to bottom
2072 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2073 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2075 reset_ptr(src2, c->srcFormat);
2076 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2078 /* reset slice direction at end of frame */
2079 if (srcSliceY + srcSliceH == c->srcH)
2082 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2084 // slices go from bottom to top => we flip the image internally
2085 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2086 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2088 src2[0] += (srcSliceH-1)*srcStride[0];
2089 if (!usePal(c->srcFormat))
2090 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2091 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2092 src2[3] += (srcSliceH-1)*srcStride[3];
2093 dst2[0] += ( c->dstH -1)*dstStride[0];
2094 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2095 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2096 dst2[3] += ( c->dstH -1)*dstStride[3];
2098 reset_ptr(src2, c->srcFormat);
2099 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2101 /* reset slice direction at end of frame */
2105 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2109 #if LIBSWSCALE_VERSION_MAJOR < 1
2110 int sws_scale_ordered(SwsContext *c, const uint8_t* const src[], int srcStride[], int srcSliceY,
2111 int srcSliceH, uint8_t* dst[], int dstStride[])
2113 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
2117 /* Convert the palette to the same packed 32-bit format as the palette */
2118 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2122 for (i=0; i<num_pixels; i++)
2123 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2126 /* Palette format: ABCD -> dst format: ABC */
2127 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2131 for (i=0; i<num_pixels; i++) {
2133 dst[0]= palette[src[i]*4+0];
2134 dst[1]= palette[src[i]*4+1];
2135 dst[2]= palette[src[i]*4+2];