2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of Libav.
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
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"
76 #define isPacked(x) ( \
78 || (x)==PIX_FMT_YUYV422 \
79 || (x)==PIX_FMT_UYVY422 \
80 || (x)==PIX_FMT_Y400A \
84 #define RGB2YUV_SHIFT 15
85 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
86 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
87 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
88 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
89 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
90 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
91 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
92 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
93 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
95 static const double rgb2yuv_table[8][9]={
96 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
97 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
98 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
99 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
100 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
101 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
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 {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
108 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
111 more intelligent misalignment avoidance for the horizontal scaler
112 write special vertical cubic upscale version
113 optimize C code (YV12 / minmax)
114 add support for packed pixel YUV input & output
115 add support for Y8 output
116 optimize BGR24 & BGR32
117 add BGR4 output support
118 write special BGR->BGR scaler
121 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
122 { 1, 3, 1, 3, 1, 3, 1, 3, },
123 { 2, 0, 2, 0, 2, 0, 2, 0, },
126 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
127 { 6, 2, 6, 2, 6, 2, 6, 2, },
128 { 0, 4, 0, 4, 0, 4, 0, 4, },
131 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
132 { 8, 4, 11, 7, 8, 4, 11, 7, },
133 { 2, 14, 1, 13, 2, 14, 1, 13, },
134 { 10, 6, 9, 5, 10, 6, 9, 5, },
135 { 0, 12, 3, 15, 0, 12, 3, 15, },
138 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
139 { 17, 9, 23, 15, 16, 8, 22, 14, },
140 { 5, 29, 3, 27, 4, 28, 2, 26, },
141 { 21, 13, 19, 11, 20, 12, 18, 10, },
142 { 0, 24, 6, 30, 1, 25, 7, 31, },
143 { 16, 8, 22, 14, 17, 9, 23, 15, },
144 { 4, 28, 2, 26, 5, 29, 3, 27, },
145 { 20, 12, 18, 10, 21, 13, 19, 11, },
146 { 1, 25, 7, 31, 0, 24, 6, 30, },
149 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
150 { 0, 55, 14, 68, 3, 58, 17, 72, },
151 { 37, 18, 50, 32, 40, 22, 54, 35, },
152 { 9, 64, 5, 59, 13, 67, 8, 63, },
153 { 46, 27, 41, 23, 49, 31, 44, 26, },
154 { 2, 57, 16, 71, 1, 56, 15, 70, },
155 { 39, 21, 52, 34, 38, 19, 51, 33, },
156 { 11, 66, 7, 62, 10, 65, 6, 60, },
157 { 48, 30, 43, 25, 47, 29, 42, 24, },
161 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
162 {117, 62, 158, 103, 113, 58, 155, 100, },
163 { 34, 199, 21, 186, 31, 196, 17, 182, },
164 {144, 89, 131, 76, 141, 86, 127, 72, },
165 { 0, 165, 41, 206, 10, 175, 52, 217, },
166 {110, 55, 151, 96, 120, 65, 162, 107, },
167 { 28, 193, 14, 179, 38, 203, 24, 189, },
168 {138, 83, 124, 69, 148, 93, 134, 79, },
169 { 7, 172, 48, 213, 3, 168, 45, 210, },
172 // tries to correct a gamma of 1.5
173 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
174 { 0, 143, 18, 200, 2, 156, 25, 215, },
175 { 78, 28, 125, 64, 89, 36, 138, 74, },
176 { 10, 180, 3, 161, 16, 195, 8, 175, },
177 {109, 51, 93, 38, 121, 60, 105, 47, },
178 { 1, 152, 23, 210, 0, 147, 20, 205, },
179 { 85, 33, 134, 71, 81, 30, 130, 67, },
180 { 14, 190, 6, 171, 12, 185, 5, 166, },
181 {117, 57, 101, 44, 113, 54, 97, 41, },
184 // tries to correct a gamma of 2.0
185 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
186 { 0, 124, 8, 193, 0, 140, 12, 213, },
187 { 55, 14, 104, 42, 66, 19, 119, 52, },
188 { 3, 168, 1, 145, 6, 187, 3, 162, },
189 { 86, 31, 70, 21, 99, 39, 82, 28, },
190 { 0, 134, 11, 206, 0, 129, 9, 200, },
191 { 62, 17, 114, 48, 58, 16, 109, 45, },
192 { 5, 181, 2, 157, 4, 175, 1, 151, },
193 { 95, 36, 78, 26, 90, 34, 74, 24, },
196 // tries to correct a gamma of 2.5
197 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
198 { 0, 107, 3, 187, 0, 125, 6, 212, },
199 { 39, 7, 86, 28, 49, 11, 102, 36, },
200 { 1, 158, 0, 131, 3, 180, 1, 151, },
201 { 68, 19, 52, 12, 81, 25, 64, 17, },
202 { 0, 119, 5, 203, 0, 113, 4, 195, },
203 { 45, 9, 96, 33, 42, 8, 91, 30, },
204 { 2, 172, 1, 144, 2, 165, 0, 137, },
205 { 77, 23, 60, 15, 72, 21, 56, 14, },
209 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
210 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
211 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
212 int dstW, int chrDstW, int big_endian, int output_bits)
214 //FIXME Optimize (just quickly written not optimized..)
216 int shift = 11 + 16 - output_bits;
218 #define output_pixel(pos, val) \
220 if (output_bits == 16) { \
221 AV_WB16(pos, av_clip_uint16(val >> shift)); \
223 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
226 if (output_bits == 16) { \
227 AV_WL16(pos, av_clip_uint16(val >> shift)); \
229 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
232 for (i = 0; i < dstW; i++) {
233 int val = 1 << (26-output_bits);
236 for (j = 0; j < lumFilterSize; j++)
237 val += lumSrc[j][i] * lumFilter[j];
239 output_pixel(&dest[i], val);
243 for (i = 0; i < chrDstW; i++) {
244 int u = 1 << (26-output_bits);
245 int v = 1 << (26-output_bits);
248 for (j = 0; j < chrFilterSize; j++) {
249 u += chrSrc[j][i ] * chrFilter[j];
250 v += chrSrc[j][i + VOFW] * chrFilter[j];
253 output_pixel(&uDest[i], u);
254 output_pixel(&vDest[i], v);
258 if (CONFIG_SWSCALE_ALPHA && aDest) {
259 for (i = 0; i < dstW; i++) {
260 int val = 1 << (26-output_bits);
263 for (j = 0; j < lumFilterSize; j++)
264 val += alpSrc[j][i] * lumFilter[j];
266 output_pixel(&aDest[i], val);
271 #define yuv2NBPS(bits, BE_LE, is_be) \
272 static void yuv2yuvX ## bits ## BE_LE ## _c(const int16_t *lumFilter, \
273 const int16_t **lumSrc, int lumFilterSize, \
274 const int16_t *chrFilter, const int16_t **chrSrc, \
275 int chrFilterSize, const int16_t **alpSrc, \
276 uint16_t *dest, uint16_t *uDest, uint16_t *vDest, \
277 uint16_t *aDest, int dstW, int chrDstW) \
279 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \
280 chrFilter, chrSrc, chrFilterSize, \
282 dest, uDest, vDest, aDest, \
283 dstW, chrDstW, is_be, bits); \
292 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
293 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
294 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
295 enum PixelFormat dstFormat)
297 #define conv16(bits) \
298 if (isBE(dstFormat)) { \
299 yuv2yuvX ## bits ## BE_c(lumFilter, lumSrc, lumFilterSize, \
300 chrFilter, chrSrc, chrFilterSize, \
302 dest, uDest, vDest, aDest, \
305 yuv2yuvX ## bits ## LE_c(lumFilter, lumSrc, lumFilterSize, \
306 chrFilter, chrSrc, chrFilterSize, \
308 dest, uDest, vDest, aDest, \
311 if (is16BPS(dstFormat)) {
313 } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
321 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
322 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
323 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
325 //FIXME Optimize (just quickly written not optimized..)
327 for (i=0; i<dstW; i++) {
330 for (j=0; j<lumFilterSize; j++)
331 val += lumSrc[j][i] * lumFilter[j];
333 dest[i]= av_clip_uint8(val>>19);
337 for (i=0; i<chrDstW; i++) {
341 for (j=0; j<chrFilterSize; j++) {
342 u += chrSrc[j][i] * chrFilter[j];
343 v += chrSrc[j][i + VOFW] * chrFilter[j];
346 uDest[i]= av_clip_uint8(u>>19);
347 vDest[i]= av_clip_uint8(v>>19);
350 if (CONFIG_SWSCALE_ALPHA && aDest)
351 for (i=0; i<dstW; i++) {
354 for (j=0; j<lumFilterSize; j++)
355 val += alpSrc[j][i] * lumFilter[j];
357 aDest[i]= av_clip_uint8(val>>19);
362 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
363 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
364 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
366 //FIXME Optimize (just quickly written not optimized..)
368 for (i=0; i<dstW; i++) {
371 for (j=0; j<lumFilterSize; j++)
372 val += lumSrc[j][i] * lumFilter[j];
374 dest[i]= av_clip_uint8(val>>19);
380 if (dstFormat == PIX_FMT_NV12)
381 for (i=0; i<chrDstW; i++) {
385 for (j=0; j<chrFilterSize; j++) {
386 u += chrSrc[j][i] * chrFilter[j];
387 v += chrSrc[j][i + VOFW] * chrFilter[j];
390 uDest[2*i]= av_clip_uint8(u>>19);
391 uDest[2*i+1]= av_clip_uint8(v>>19);
394 for (i=0; i<chrDstW; i++) {
398 for (j=0; j<chrFilterSize; j++) {
399 u += chrSrc[j][i] * chrFilter[j];
400 v += chrSrc[j][i + VOFW] * chrFilter[j];
403 uDest[2*i]= av_clip_uint8(v>>19);
404 uDest[2*i+1]= av_clip_uint8(u>>19);
408 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
409 for (i=0; i<(dstW>>1); i++) {\
415 int av_unused A1, A2;\
416 type av_unused *r, *b, *g;\
419 for (j=0; j<lumFilterSize; j++) {\
420 Y1 += lumSrc[j][i2] * lumFilter[j];\
421 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
423 for (j=0; j<chrFilterSize; j++) {\
424 U += chrSrc[j][i] * chrFilter[j];\
425 V += chrSrc[j][i+VOFW] * chrFilter[j];\
434 for (j=0; j<lumFilterSize; j++) {\
435 A1 += alpSrc[j][i2 ] * lumFilter[j];\
436 A2 += alpSrc[j][i2+1] * lumFilter[j];\
442 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
443 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
444 if ((Y1|Y2|U|V)&256) {\
445 if (Y1>255) Y1=255; \
446 else if (Y1<0)Y1=0; \
447 if (Y2>255) Y2=255; \
448 else if (Y2<0)Y2=0; \
454 if (alpha && ((A1|A2)&256)) {\
455 A1=av_clip_uint8(A1);\
456 A2=av_clip_uint8(A2);\
459 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
460 for (i=0; i<dstW; i++) {\
468 for (j=0; j<lumFilterSize; j++) {\
469 Y += lumSrc[j][i ] * lumFilter[j];\
471 for (j=0; j<chrFilterSize; j++) {\
472 U += chrSrc[j][i ] * chrFilter[j];\
473 V += chrSrc[j][i+VOFW] * chrFilter[j];\
480 for (j=0; j<lumFilterSize; j++)\
481 A += alpSrc[j][i ] * lumFilter[j];\
484 A = av_clip_uint8(A);\
487 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
488 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
489 Y-= c->yuv2rgb_y_offset;\
490 Y*= c->yuv2rgb_y_coeff;\
492 R= Y + V*c->yuv2rgb_v2r_coeff;\
493 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
494 B= Y + U*c->yuv2rgb_u2b_coeff;\
495 if ((R|G|B)&(0xC0000000)) {\
496 if (R>=(256<<22)) R=(256<<22)-1; \
498 if (G>=(256<<22)) G=(256<<22)-1; \
500 if (B>=(256<<22)) B=(256<<22)-1; \
504 #define YSCALE_YUV_2_GRAY16_C \
505 for (i=0; i<(dstW>>1); i++) {\
514 for (j=0; j<lumFilterSize; j++) {\
515 Y1 += lumSrc[j][i2] * lumFilter[j];\
516 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
520 if ((Y1|Y2|U|V)&65536) {\
521 if (Y1>65535) Y1=65535; \
522 else if (Y1<0)Y1=0; \
523 if (Y2>65535) Y2=65535; \
524 else if (Y2<0)Y2=0; \
527 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
528 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
529 r = (type *)c->table_rV[V]; \
530 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
531 b = (type *)c->table_bU[U];
533 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
534 for (i=0; i<(dstW>>1); i++) { \
536 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
537 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
538 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
539 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
540 type av_unused *r, *b, *g; \
541 int av_unused A1, A2; \
543 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
544 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
547 #define YSCALE_YUV_2_GRAY16_2_C \
548 for (i=0; i<(dstW>>1); i++) { \
550 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
551 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
553 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
554 YSCALE_YUV_2_PACKED2_C(type,alpha)\
555 r = (type *)c->table_rV[V];\
556 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
557 b = (type *)c->table_bU[U];
559 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
560 for (i=0; i<(dstW>>1); i++) {\
562 int Y1= buf0[i2 ]>>7;\
563 int Y2= buf0[i2+1]>>7;\
564 int U= (uvbuf1[i ])>>7;\
565 int V= (uvbuf1[i+VOFW])>>7;\
566 type av_unused *r, *b, *g;\
567 int av_unused A1, A2;\
573 #define YSCALE_YUV_2_GRAY16_1_C \
574 for (i=0; i<(dstW>>1); i++) {\
576 int Y1= buf0[i2 ]<<1;\
577 int Y2= buf0[i2+1]<<1;
579 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
580 YSCALE_YUV_2_PACKED1_C(type,alpha)\
581 r = (type *)c->table_rV[V];\
582 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
583 b = (type *)c->table_bU[U];
585 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
586 for (i=0; i<(dstW>>1); i++) {\
588 int Y1= buf0[i2 ]>>7;\
589 int Y2= buf0[i2+1]>>7;\
590 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
591 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
592 type av_unused *r, *b, *g;\
593 int av_unused A1, A2;\
599 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
600 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
601 r = (type *)c->table_rV[V];\
602 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
603 b = (type *)c->table_bU[U];
605 #define YSCALE_YUV_2_MONO2_C \
606 const uint8_t * const d128=dither_8x8_220[y&7];\
607 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
608 for (i=0; i<dstW-7; i+=8) {\
610 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
611 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
612 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
613 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
614 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
615 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
616 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
617 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
618 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
622 #define YSCALE_YUV_2_MONOX_C \
623 const uint8_t * const d128=dither_8x8_220[y&7];\
624 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
626 for (i=0; i<dstW-1; i+=2) {\
631 for (j=0; j<lumFilterSize; j++) {\
632 Y1 += lumSrc[j][i] * lumFilter[j];\
633 Y2 += lumSrc[j][i+1] * lumFilter[j];\
643 acc+= acc + g[Y1+d128[(i+0)&7]];\
644 acc+= acc + g[Y2+d128[(i+1)&7]];\
646 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
651 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
652 switch(c->dstFormat) {\
653 case PIX_FMT_RGB48BE:\
654 case PIX_FMT_RGB48LE:\
656 ((uint8_t*)dest)[ 0]= r[Y1];\
657 ((uint8_t*)dest)[ 1]= r[Y1];\
658 ((uint8_t*)dest)[ 2]= g[Y1];\
659 ((uint8_t*)dest)[ 3]= g[Y1];\
660 ((uint8_t*)dest)[ 4]= b[Y1];\
661 ((uint8_t*)dest)[ 5]= b[Y1];\
662 ((uint8_t*)dest)[ 6]= r[Y2];\
663 ((uint8_t*)dest)[ 7]= r[Y2];\
664 ((uint8_t*)dest)[ 8]= g[Y2];\
665 ((uint8_t*)dest)[ 9]= g[Y2];\
666 ((uint8_t*)dest)[10]= b[Y2];\
667 ((uint8_t*)dest)[11]= b[Y2];\
671 case PIX_FMT_BGR48BE:\
672 case PIX_FMT_BGR48LE:\
674 ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
675 ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
676 ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
677 ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
678 ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
679 ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
686 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
687 func(uint32_t,needAlpha)\
688 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
689 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
692 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
694 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
695 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
699 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
700 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
708 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
709 func(uint32_t,needAlpha)\
710 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
711 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
714 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
716 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
717 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
721 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
722 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
729 ((uint8_t*)dest)[0]= r[Y1];\
730 ((uint8_t*)dest)[1]= g[Y1];\
731 ((uint8_t*)dest)[2]= b[Y1];\
732 ((uint8_t*)dest)[3]= r[Y2];\
733 ((uint8_t*)dest)[4]= g[Y2];\
734 ((uint8_t*)dest)[5]= b[Y2];\
740 ((uint8_t*)dest)[0]= b[Y1];\
741 ((uint8_t*)dest)[1]= g[Y1];\
742 ((uint8_t*)dest)[2]= r[Y1];\
743 ((uint8_t*)dest)[3]= b[Y2];\
744 ((uint8_t*)dest)[4]= g[Y2];\
745 ((uint8_t*)dest)[5]= r[Y2];\
749 case PIX_FMT_RGB565BE:\
750 case PIX_FMT_RGB565LE:\
751 case PIX_FMT_BGR565BE:\
752 case PIX_FMT_BGR565LE:\
754 const int dr1= dither_2x2_8[y&1 ][0];\
755 const int dg1= dither_2x2_4[y&1 ][0];\
756 const int db1= dither_2x2_8[(y&1)^1][0];\
757 const int dr2= dither_2x2_8[y&1 ][1];\
758 const int dg2= dither_2x2_4[y&1 ][1];\
759 const int db2= dither_2x2_8[(y&1)^1][1];\
761 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
762 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
766 case PIX_FMT_RGB555BE:\
767 case PIX_FMT_RGB555LE:\
768 case PIX_FMT_BGR555BE:\
769 case PIX_FMT_BGR555LE:\
771 const int dr1= dither_2x2_8[y&1 ][0];\
772 const int dg1= dither_2x2_8[y&1 ][1];\
773 const int db1= dither_2x2_8[(y&1)^1][0];\
774 const int dr2= dither_2x2_8[y&1 ][1];\
775 const int dg2= dither_2x2_8[y&1 ][0];\
776 const int db2= dither_2x2_8[(y&1)^1][1];\
778 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
779 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
783 case PIX_FMT_RGB444BE:\
784 case PIX_FMT_RGB444LE:\
785 case PIX_FMT_BGR444BE:\
786 case PIX_FMT_BGR444LE:\
788 const int dr1= dither_4x4_16[y&3 ][0];\
789 const int dg1= dither_4x4_16[y&3 ][1];\
790 const int db1= dither_4x4_16[(y&3)^3][0];\
791 const int dr2= dither_4x4_16[y&3 ][1];\
792 const int dg2= dither_4x4_16[y&3 ][0];\
793 const int db2= dither_4x4_16[(y&3)^3][1];\
795 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
796 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
803 const uint8_t * const d64= dither_8x8_73[y&7];\
804 const uint8_t * const d32= dither_8x8_32[y&7];\
806 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
807 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
814 const uint8_t * const d64= dither_8x8_73 [y&7];\
815 const uint8_t * const d128=dither_8x8_220[y&7];\
817 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
818 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
822 case PIX_FMT_RGB4_BYTE:\
823 case PIX_FMT_BGR4_BYTE:\
825 const uint8_t * const d64= dither_8x8_73 [y&7];\
826 const uint8_t * const d128=dither_8x8_220[y&7];\
828 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
829 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
833 case PIX_FMT_MONOBLACK:\
834 case PIX_FMT_MONOWHITE:\
839 case PIX_FMT_YUYV422:\
841 ((uint8_t*)dest)[2*i2+0]= Y1;\
842 ((uint8_t*)dest)[2*i2+1]= U;\
843 ((uint8_t*)dest)[2*i2+2]= Y2;\
844 ((uint8_t*)dest)[2*i2+3]= V;\
847 case PIX_FMT_UYVY422:\
849 ((uint8_t*)dest)[2*i2+0]= U;\
850 ((uint8_t*)dest)[2*i2+1]= Y1;\
851 ((uint8_t*)dest)[2*i2+2]= V;\
852 ((uint8_t*)dest)[2*i2+3]= Y2;\
855 case PIX_FMT_GRAY16BE:\
857 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
858 ((uint8_t*)dest)[2*i2+1]= Y1;\
859 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
860 ((uint8_t*)dest)[2*i2+3]= Y2;\
863 case PIX_FMT_GRAY16LE:\
865 ((uint8_t*)dest)[2*i2+0]= Y1;\
866 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
867 ((uint8_t*)dest)[2*i2+2]= Y2;\
868 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
873 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
874 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
875 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
878 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)
881 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
882 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
883 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
886 int step= c->dstFormatBpp/8;
889 switch(c->dstFormat) {
897 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
898 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
899 dest[aidx]= needAlpha ? A : 255;
906 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
907 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
915 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
932 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
933 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
934 dest[aidx]= needAlpha ? A : 255;
941 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
942 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
950 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
965 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
968 uint8_t *ptr = plane + stride*y;
969 for (i=0; i<height; i++) {
970 memset(ptr, val, width);
975 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
979 for (i = 0; i < width; i++) {
984 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
988 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
989 const uint8_t *src1, const uint8_t *src2,
990 long width, uint32_t *unused)
994 for (i = 0; i < width; i++) {
995 int r = src1[6*i + 0];
996 int g = src1[6*i + 2];
997 int b = src1[6*i + 4];
999 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1000 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1004 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1005 const uint8_t *src1, const uint8_t *src2,
1006 long width, uint32_t *unused)
1010 for (i = 0; i < width; i++) {
1011 int r= src1[12*i + 0] + src1[12*i + 6];
1012 int g= src1[12*i + 2] + src1[12*i + 8];
1013 int b= src1[12*i + 4] + src1[12*i + 10];
1015 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1016 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1020 static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,
1024 for (i = 0; i < width; i++) {
1029 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1033 static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1034 const uint8_t *src1, const uint8_t *src2,
1035 long width, uint32_t *unused)
1038 for (i = 0; i < width; i++) {
1039 int b = src1[6*i + 0];
1040 int g = src1[6*i + 2];
1041 int r = src1[6*i + 4];
1043 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1044 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1048 static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1049 const uint8_t *src1, const uint8_t *src2,
1050 long width, uint32_t *unused)
1053 for (i = 0; i < width; i++) {
1054 int b= src1[12*i + 0] + src1[12*i + 6];
1055 int g= src1[12*i + 2] + src1[12*i + 8];
1056 int r= src1[12*i + 4] + src1[12*i + 10];
1058 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1059 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1063 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1064 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1067 for (i=0; i<width; i++) {\
1068 int b= (((const type*)src)[i]>>shb)&maskb;\
1069 int g= (((const type*)src)[i]>>shg)&maskg;\
1070 int r= (((const type*)src)[i]>>shr)&maskr;\
1072 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1076 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1077 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1078 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1079 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1080 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1081 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1082 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1083 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1085 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1088 for (i=0; i<width; i++) {
1093 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1094 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1097 for (i=0; i<width; i++) {\
1098 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1099 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1100 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1102 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1103 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1106 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1109 for (i=0; i<width; i++) {\
1110 int pix0= ((const type*)src)[2*i+0]>>shp;\
1111 int pix1= ((const type*)src)[2*i+1]>>shp;\
1112 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1113 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1114 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1115 g&= maskg|(2*maskg);\
1119 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1120 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1124 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1125 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1126 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1127 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1128 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1129 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1130 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1131 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1133 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1136 for (i=0; i<width; i++) {
1139 dst[i]= pal[d] & 0xFF;
1143 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1144 const uint8_t *src1, const uint8_t *src2,
1145 long width, uint32_t *pal)
1148 assert(src1 == src2);
1149 for (i=0; i<width; i++) {
1150 int p= pal[src1[i]];
1157 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1160 for (i=0; i<width/8; i++) {
1163 dst[8*i+j]= ((d>>(7-j))&1)*255;
1167 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1170 for (i=0; i<width/8; i++) {
1173 dst[8*i+j]= ((d>>(7-j))&1)*255;
1177 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1180 #define COMPILE_TEMPLATE_MMX2 0
1181 #define COMPILE_TEMPLATE_ALTIVEC 0
1183 #include "swscale_template.c"
1187 #undef COMPILE_TEMPLATE_ALTIVEC
1188 #define COMPILE_TEMPLATE_ALTIVEC 1
1189 #define RENAME(a) a ## _altivec
1190 #include "ppc/swscale_template.c"
1196 #undef COMPILE_TEMPLATE_MMX2
1197 #define COMPILE_TEMPLATE_MMX2 0
1198 #define RENAME(a) a ## _MMX
1199 #include "x86/swscale_template.c"
1205 #undef COMPILE_TEMPLATE_MMX2
1206 #define COMPILE_TEMPLATE_MMX2 1
1207 #define RENAME(a) a ## _MMX2
1208 #include "x86/swscale_template.c"
1211 SwsFunc ff_getSwsFunc(SwsContext *c)
1213 int cpu_flags = av_get_cpu_flags();
1215 sws_init_swScale_c(c);
1218 if (cpu_flags & AV_CPU_FLAG_MMX)
1219 sws_init_swScale_MMX(c);
1222 if (cpu_flags & AV_CPU_FLAG_MMX2)
1223 sws_init_swScale_MMX2(c);
1226 if (cpu_flags & AV_CPU_FLAG_ALTIVEC)
1227 sws_init_swScale_altivec(c);
1233 static void copyPlane(const uint8_t *src, int srcStride,
1234 int srcSliceY, int srcSliceH, int width,
1235 uint8_t *dst, int dstStride)
1237 dst += dstStride * srcSliceY;
1238 if (dstStride == srcStride && srcStride > 0) {
1239 memcpy(dst, src, srcSliceH * dstStride);
1242 for (i=0; i<srcSliceH; i++) {
1243 memcpy(dst, src, width);
1250 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1251 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1253 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1255 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1256 dstParam[0], dstStride[0]);
1258 if (c->dstFormat == PIX_FMT_NV12)
1259 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1261 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1266 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1267 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1269 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1271 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1276 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1277 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1279 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1281 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1286 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1287 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1289 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1291 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1296 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1297 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1299 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1301 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1306 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1307 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1309 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1310 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1311 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1313 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1316 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1321 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1322 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1324 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1325 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1326 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1328 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1333 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1334 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1336 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1337 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1338 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1340 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1343 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1348 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1349 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1351 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1352 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1353 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1355 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1360 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1363 for (i=0; i<num_pixels; i++)
1364 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1367 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1371 for (i=0; i<num_pixels; i++)
1372 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1375 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1379 for (i=0; i<num_pixels; i++) {
1381 dst[0]= palette[src[i<<1]*4+0];
1382 dst[1]= palette[src[i<<1]*4+1];
1383 dst[2]= palette[src[i<<1]*4+2];
1388 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1389 int srcSliceH, uint8_t* dst[], int dstStride[])
1391 const enum PixelFormat srcFormat= c->srcFormat;
1392 const enum PixelFormat dstFormat= c->dstFormat;
1393 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1394 const uint8_t *palette)=NULL;
1396 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1397 const uint8_t *srcPtr= src[0];
1399 if (srcFormat == PIX_FMT_Y400A) {
1400 switch (dstFormat) {
1401 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1402 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1403 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1404 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1405 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1406 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1408 } else if (usePal(srcFormat)) {
1409 switch (dstFormat) {
1410 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1411 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1412 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1413 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1414 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1415 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1420 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1421 sws_format_name(srcFormat), sws_format_name(dstFormat));
1423 for (i=0; i<srcSliceH; i++) {
1424 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1425 srcPtr+= srcStride[0];
1426 dstPtr+= dstStride[0];
1433 #define isRGBA32(x) ( \
1434 (x) == PIX_FMT_ARGB \
1435 || (x) == PIX_FMT_RGBA \
1436 || (x) == PIX_FMT_BGRA \
1437 || (x) == PIX_FMT_ABGR \
1440 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1441 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1442 int srcSliceH, uint8_t* dst[], int dstStride[])
1444 const enum PixelFormat srcFormat= c->srcFormat;
1445 const enum PixelFormat dstFormat= c->dstFormat;
1446 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1447 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1448 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1449 const int dstId= c->dstFormatBpp >> 2;
1450 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1452 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1454 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1455 if ( CONV_IS(ABGR, RGBA)
1456 || CONV_IS(ARGB, BGRA)
1457 || CONV_IS(BGRA, ARGB)
1458 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1459 else if (CONV_IS(ABGR, ARGB)
1460 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1461 else if (CONV_IS(ABGR, BGRA)
1462 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1463 else if (CONV_IS(BGRA, RGBA)
1464 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1465 else if (CONV_IS(BGRA, ABGR)
1466 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1469 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1470 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1471 switch(srcId | (dstId<<4)) {
1472 case 0x34: conv= rgb16to15; break;
1473 case 0x36: conv= rgb24to15; break;
1474 case 0x38: conv= rgb32to15; break;
1475 case 0x43: conv= rgb15to16; break;
1476 case 0x46: conv= rgb24to16; break;
1477 case 0x48: conv= rgb32to16; break;
1478 case 0x63: conv= rgb15to24; break;
1479 case 0x64: conv= rgb16to24; break;
1480 case 0x68: conv= rgb32to24; break;
1481 case 0x83: conv= rgb15to32; break;
1482 case 0x84: conv= rgb16to32; break;
1483 case 0x86: conv= rgb24to32; break;
1485 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1486 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1487 switch(srcId | (dstId<<4)) {
1488 case 0x33: conv= rgb15tobgr15; break;
1489 case 0x34: conv= rgb16tobgr15; break;
1490 case 0x36: conv= rgb24tobgr15; break;
1491 case 0x38: conv= rgb32tobgr15; break;
1492 case 0x43: conv= rgb15tobgr16; break;
1493 case 0x44: conv= rgb16tobgr16; break;
1494 case 0x46: conv= rgb24tobgr16; break;
1495 case 0x48: conv= rgb32tobgr16; break;
1496 case 0x63: conv= rgb15tobgr24; break;
1497 case 0x64: conv= rgb16tobgr24; break;
1498 case 0x66: conv= rgb24tobgr24; break;
1499 case 0x68: conv= rgb32tobgr24; break;
1500 case 0x83: conv= rgb15tobgr32; break;
1501 case 0x84: conv= rgb16tobgr32; break;
1502 case 0x86: conv= rgb24tobgr32; break;
1507 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1508 sws_format_name(srcFormat), sws_format_name(dstFormat));
1510 const uint8_t *srcPtr= src[0];
1511 uint8_t *dstPtr= dst[0];
1512 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1513 srcPtr += ALT32_CORR;
1515 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1516 dstPtr += ALT32_CORR;
1518 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1519 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1522 dstPtr += dstStride[0]*srcSliceY;
1524 for (i=0; i<srcSliceH; i++) {
1525 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1526 srcPtr+= srcStride[0];
1527 dstPtr+= dstStride[0];
1534 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1535 int srcSliceH, uint8_t* dst[], int dstStride[])
1539 dst[0]+ srcSliceY *dstStride[0],
1540 dst[1]+(srcSliceY>>1)*dstStride[1],
1541 dst[2]+(srcSliceY>>1)*dstStride[2],
1543 dstStride[0], dstStride[1], srcStride[0]);
1545 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1549 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1550 int srcSliceH, uint8_t* dst[], int dstStride[])
1552 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1553 dst[0], dstStride[0]);
1555 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1556 srcSliceH >> 2, srcStride[1], dstStride[1]);
1557 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1558 srcSliceH >> 2, srcStride[2], dstStride[2]);
1560 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1564 /* unscaled copy like stuff (assumes nearly identical formats) */
1565 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1566 int srcSliceH, uint8_t* dst[], int dstStride[])
1568 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1569 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1572 const uint8_t *srcPtr= src[0];
1573 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1576 /* universal length finder */
1577 while(length+c->srcW <= FFABS(dstStride[0])
1578 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1581 for (i=0; i<srcSliceH; i++) {
1582 memcpy(dstPtr, srcPtr, length);
1583 srcPtr+= srcStride[0];
1584 dstPtr+= dstStride[0];
1590 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1591 int srcSliceH, uint8_t* dst[], int dstStride[])
1594 for (plane=0; plane<4; plane++) {
1595 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1596 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1597 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1598 const uint8_t *srcPtr= src[plane];
1599 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1601 if (!dst[plane]) continue;
1602 // ignore palette for GRAY8
1603 if (plane == 1 && !dst[2]) continue;
1604 if (!src[plane] || (plane == 1 && !src[2])) {
1605 if(is16BPS(c->dstFormat))
1607 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1609 if(is9_OR_10BPS(c->srcFormat)) {
1610 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
1611 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1612 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1614 if (is16BPS(c->dstFormat)) {
1615 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1616 #define COPY9_OR_10TO16(rfunc, wfunc) \
1617 for (i = 0; i < height; i++) { \
1618 for (j = 0; j < length; j++) { \
1619 int srcpx = rfunc(&srcPtr2[j]); \
1620 wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \
1622 dstPtr2 += dstStride[plane]/2; \
1623 srcPtr2 += srcStride[plane]/2; \
1625 if (isBE(c->dstFormat)) {
1626 if (isBE(c->srcFormat)) {
1627 COPY9_OR_10TO16(AV_RB16, AV_WB16);
1629 COPY9_OR_10TO16(AV_RL16, AV_WB16);
1632 if (isBE(c->srcFormat)) {
1633 COPY9_OR_10TO16(AV_RB16, AV_WL16);
1635 COPY9_OR_10TO16(AV_RL16, AV_WL16);
1638 } else if (is9_OR_10BPS(c->dstFormat)) {
1639 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1640 #define COPY9_OR_10TO9_OR_10(loop) \
1641 for (i = 0; i < height; i++) { \
1642 for (j = 0; j < length; j++) { \
1645 dstPtr2 += dstStride[plane]/2; \
1646 srcPtr2 += srcStride[plane]/2; \
1648 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
1649 if (dst_depth > src_depth) { \
1650 COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
1651 wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
1652 } else if (dst_depth < src_depth) { \
1653 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \
1655 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
1657 if (isBE(c->dstFormat)) {
1658 if (isBE(c->srcFormat)) {
1659 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
1661 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
1664 if (isBE(c->srcFormat)) {
1665 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
1667 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
1671 // FIXME Maybe dither instead.
1672 #define COPY9_OR_10TO8(rfunc) \
1673 for (i = 0; i < height; i++) { \
1674 for (j = 0; j < length; j++) { \
1675 dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \
1677 dstPtr += dstStride[plane]; \
1678 srcPtr2 += srcStride[plane]/2; \
1680 if (isBE(c->srcFormat)) {
1681 COPY9_OR_10TO8(AV_RB16);
1683 COPY9_OR_10TO8(AV_RL16);
1686 } else if(is9_OR_10BPS(c->dstFormat)) {
1687 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1688 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1690 if (is16BPS(c->srcFormat)) {
1691 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1692 #define COPY16TO9_OR_10(rfunc, wfunc) \
1693 for (i = 0; i < height; i++) { \
1694 for (j = 0; j < length; j++) { \
1695 wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \
1697 dstPtr2 += dstStride[plane]/2; \
1698 srcPtr2 += srcStride[plane]/2; \
1700 if (isBE(c->dstFormat)) {
1701 if (isBE(c->srcFormat)) {
1702 COPY16TO9_OR_10(AV_RB16, AV_WB16);
1704 COPY16TO9_OR_10(AV_RL16, AV_WB16);
1707 if (isBE(c->srcFormat)) {
1708 COPY16TO9_OR_10(AV_RB16, AV_WL16);
1710 COPY16TO9_OR_10(AV_RL16, AV_WL16);
1714 #define COPY8TO9_OR_10(wfunc) \
1715 for (i = 0; i < height; i++) { \
1716 for (j = 0; j < length; j++) { \
1717 const int srcpx = srcPtr[j]; \
1718 wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \
1720 dstPtr2 += dstStride[plane]/2; \
1721 srcPtr += srcStride[plane]; \
1723 if (isBE(c->dstFormat)) {
1724 COPY8TO9_OR_10(AV_WB16);
1726 COPY8TO9_OR_10(AV_WL16);
1729 } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1730 if (!isBE(c->srcFormat)) srcPtr++;
1731 for (i=0; i<height; i++) {
1732 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1733 srcPtr+= srcStride[plane];
1734 dstPtr+= dstStride[plane];
1736 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1737 for (i=0; i<height; i++) {
1738 for (j=0; j<length; j++) {
1739 dstPtr[ j<<1 ] = srcPtr[j];
1740 dstPtr[(j<<1)+1] = srcPtr[j];
1742 srcPtr+= srcStride[plane];
1743 dstPtr+= dstStride[plane];
1745 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1746 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1748 for (i=0; i<height; i++) {
1749 for (j=0; j<length; j++)
1750 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1751 srcPtr+= srcStride[plane];
1752 dstPtr+= dstStride[plane];
1754 } else if (dstStride[plane] == srcStride[plane] &&
1755 srcStride[plane] > 0 && srcStride[plane] == length) {
1756 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1757 height*dstStride[plane]);
1759 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1761 for (i=0; i<height; i++) {
1762 memcpy(dstPtr, srcPtr, length);
1763 srcPtr+= srcStride[plane];
1764 dstPtr+= dstStride[plane];
1772 void ff_get_unscaled_swscale(SwsContext *c)
1774 const enum PixelFormat srcFormat = c->srcFormat;
1775 const enum PixelFormat dstFormat = c->dstFormat;
1776 const int flags = c->flags;
1777 const int dstH = c->dstH;
1780 needsDither= isAnyRGB(dstFormat)
1781 && c->dstFormatBpp < 24
1782 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1785 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1786 c->swScale= planarToNv12Wrapper;
1789 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1790 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1791 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1794 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1795 c->swScale= yvu9ToYv12Wrapper;
1799 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1800 c->swScale= bgr24ToYv12Wrapper;
1802 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1803 if ( isAnyRGB(srcFormat)
1804 && isAnyRGB(dstFormat)
1805 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1806 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1807 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1808 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1809 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1810 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1811 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1812 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1813 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1814 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1815 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
1816 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
1817 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1818 c->swScale= rgbToRgbWrapper;
1820 if ((usePal(srcFormat) && (
1821 dstFormat == PIX_FMT_RGB32 ||
1822 dstFormat == PIX_FMT_RGB32_1 ||
1823 dstFormat == PIX_FMT_RGB24 ||
1824 dstFormat == PIX_FMT_BGR32 ||
1825 dstFormat == PIX_FMT_BGR32_1 ||
1826 dstFormat == PIX_FMT_BGR24)))
1827 c->swScale= palToRgbWrapper;
1829 if (srcFormat == PIX_FMT_YUV422P) {
1830 if (dstFormat == PIX_FMT_YUYV422)
1831 c->swScale= yuv422pToYuy2Wrapper;
1832 else if (dstFormat == PIX_FMT_UYVY422)
1833 c->swScale= yuv422pToUyvyWrapper;
1836 /* LQ converters if -sws 0 or -sws 4*/
1837 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1839 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1840 if (dstFormat == PIX_FMT_YUYV422)
1841 c->swScale= planarToYuy2Wrapper;
1842 else if (dstFormat == PIX_FMT_UYVY422)
1843 c->swScale= planarToUyvyWrapper;
1846 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1847 c->swScale= yuyvToYuv420Wrapper;
1848 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1849 c->swScale= uyvyToYuv420Wrapper;
1850 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1851 c->swScale= yuyvToYuv422Wrapper;
1852 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1853 c->swScale= uyvyToYuv422Wrapper;
1856 if ((av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) &&
1857 !(c->flags & SWS_BITEXACT) &&
1858 srcFormat == PIX_FMT_YUV420P) {
1859 // unscaled YV12 -> packed YUV, we want speed
1860 if (dstFormat == PIX_FMT_YUYV422)
1861 c->swScale= yv12toyuy2_unscaled_altivec;
1862 else if (dstFormat == PIX_FMT_UYVY422)
1863 c->swScale= yv12touyvy_unscaled_altivec;
1868 if ( srcFormat == dstFormat
1869 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1870 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1871 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1872 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1873 || (isGray(dstFormat) && isGray(srcFormat))
1874 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1875 && c->chrDstHSubSample == c->chrSrcHSubSample
1876 && c->chrDstVSubSample == c->chrSrcVSubSample
1877 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1878 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1880 if (isPacked(c->srcFormat))
1881 c->swScale= packedCopyWrapper;
1882 else /* Planar YUV or gray */
1883 c->swScale= planarCopyWrapper;
1886 ff_bfin_get_unscaled_swscale (c);
1890 static void reset_ptr(const uint8_t* src[], int format)
1892 if(!isALPHA(format))
1894 if(!isPlanarYUV(format)) {
1897 if (!usePal(format))
1902 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1903 const int linesizes[4])
1905 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1908 for (i = 0; i < 4; i++) {
1909 int plane = desc->comp[i].plane;
1910 if (!data[plane] || !linesizes[plane])
1918 * swscale wrapper, so we don't need to export the SwsContext.
1919 * Assumes planar YUV to be in YUV order instead of YVU.
1921 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1922 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1925 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1926 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1928 // do not mess up sliceDir if we have a "trailing" 0-size slice
1932 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1933 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1936 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1937 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1941 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1942 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1945 if (c->sliceDir == 0) {
1946 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1949 if (usePal(c->srcFormat)) {
1950 for (i=0; i<256; i++) {
1951 int p, r, g, b,y,u,v;
1952 if(c->srcFormat == PIX_FMT_PAL8) {
1953 p=((const uint32_t*)(src[1]))[i];
1957 } else if(c->srcFormat == PIX_FMT_RGB8) {
1961 } else if(c->srcFormat == PIX_FMT_BGR8) {
1965 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1969 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
1972 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1977 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1978 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1979 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1980 c->pal_yuv[i]= y + (u<<8) + (v<<16);
1982 switch(c->dstFormat) {
1987 c->pal_rgb[i]= r + (g<<8) + (b<<16);
1989 case PIX_FMT_BGR32_1:
1993 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
1995 case PIX_FMT_RGB32_1:
1999 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2006 c->pal_rgb[i]= b + (g<<8) + (r<<16);
2011 // copy strides, so they can safely be modified
2012 if (c->sliceDir == 1) {
2013 // slices go from top to bottom
2014 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2015 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2017 reset_ptr(src2, c->srcFormat);
2018 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2020 /* reset slice direction at end of frame */
2021 if (srcSliceY + srcSliceH == c->srcH)
2024 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2026 // slices go from bottom to top => we flip the image internally
2027 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2028 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2030 src2[0] += (srcSliceH-1)*srcStride[0];
2031 if (!usePal(c->srcFormat))
2032 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2033 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2034 src2[3] += (srcSliceH-1)*srcStride[3];
2035 dst2[0] += ( c->dstH -1)*dstStride[0];
2036 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2037 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2038 dst2[3] += ( c->dstH -1)*dstStride[3];
2040 reset_ptr(src2, c->srcFormat);
2041 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2043 /* reset slice direction at end of frame */
2047 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2051 /* Convert the palette to the same packed 32-bit format as the palette */
2052 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2056 for (i=0; i<num_pixels; i++)
2057 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2060 /* Palette format: ABCD -> dst format: ABC */
2061 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2065 for (i=0; i<num_pixels; i++) {
2067 dst[0]= palette[src[i]*4+0];
2068 dst[1]= palette[src[i]*4+1];
2069 dst[2]= palette[src[i]*4+2];