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 **chrUSrc,
211 const int16_t **chrVSrc, int chrFilterSize,
212 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
213 int dstW, int chrDstW, int big_endian, int output_bits)
215 //FIXME Optimize (just quickly written not optimized..)
217 int shift = 11 + 16 - output_bits;
219 #define output_pixel(pos, val) \
221 if (output_bits == 16) { \
222 AV_WB16(pos, av_clip_uint16(val >> shift)); \
224 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
227 if (output_bits == 16) { \
228 AV_WL16(pos, av_clip_uint16(val >> shift)); \
230 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
233 for (i = 0; i < dstW; i++) {
234 int val = 1 << (26-output_bits);
237 for (j = 0; j < lumFilterSize; j++)
238 val += lumSrc[j][i] * lumFilter[j];
240 output_pixel(&dest[i], val);
244 for (i = 0; i < chrDstW; i++) {
245 int u = 1 << (26-output_bits);
246 int v = 1 << (26-output_bits);
249 for (j = 0; j < chrFilterSize; j++) {
250 u += chrUSrc[j][i] * chrFilter[j];
251 v += chrVSrc[j][i] * chrFilter[j];
254 output_pixel(&uDest[i], u);
255 output_pixel(&vDest[i], v);
259 if (CONFIG_SWSCALE_ALPHA && aDest) {
260 for (i = 0; i < dstW; i++) {
261 int val = 1 << (26-output_bits);
264 for (j = 0; j < lumFilterSize; j++)
265 val += alpSrc[j][i] * lumFilter[j];
267 output_pixel(&aDest[i], val);
272 #define yuv2NBPS(bits, BE_LE, is_be) \
273 static void yuv2yuvX ## bits ## BE_LE ## _c(const int16_t *lumFilter, \
274 const int16_t **lumSrc, int lumFilterSize, \
275 const int16_t *chrFilter, const int16_t **chrUSrc, \
276 const int16_t **chrVSrc, \
277 int chrFilterSize, const int16_t **alpSrc, \
278 uint16_t *dest, uint16_t *uDest, uint16_t *vDest, \
279 uint16_t *aDest, int dstW, int chrDstW) \
281 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \
282 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
284 dest, uDest, vDest, aDest, \
285 dstW, chrDstW, is_be, bits); \
294 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
295 const int16_t *chrFilter, const int16_t **chrUSrc, const int16_t **chrVSrc, int chrFilterSize,
296 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
297 enum PixelFormat dstFormat)
299 #define conv16(bits) \
300 if (isBE(dstFormat)) { \
301 yuv2yuvX ## bits ## BE_c(lumFilter, lumSrc, lumFilterSize, \
302 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
304 dest, uDest, vDest, aDest, \
307 yuv2yuvX ## bits ## LE_c(lumFilter, lumSrc, lumFilterSize, \
308 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
310 dest, uDest, vDest, aDest, \
313 if (is16BPS(dstFormat)) {
315 } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
323 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
324 const int16_t *chrFilter, const int16_t **chrUSrc,
325 const int16_t **chrVSrc, int chrFilterSize,
326 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
328 //FIXME Optimize (just quickly written not optimized..)
330 for (i=0; i<dstW; i++) {
333 for (j=0; j<lumFilterSize; j++)
334 val += lumSrc[j][i] * lumFilter[j];
336 dest[i]= av_clip_uint8(val>>19);
340 for (i=0; i<chrDstW; i++) {
344 for (j=0; j<chrFilterSize; j++) {
345 u += chrUSrc[j][i] * chrFilter[j];
346 v += chrVSrc[j][i] * chrFilter[j];
349 uDest[i]= av_clip_uint8(u>>19);
350 vDest[i]= av_clip_uint8(v>>19);
353 if (CONFIG_SWSCALE_ALPHA && aDest)
354 for (i=0; i<dstW; i++) {
357 for (j=0; j<lumFilterSize; j++)
358 val += alpSrc[j][i] * lumFilter[j];
360 aDest[i]= av_clip_uint8(val>>19);
365 static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
366 const int16_t *chrFilter, const int16_t **chrUSrc,
367 const int16_t **chrVSrc, int chrFilterSize,
368 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
370 //FIXME Optimize (just quickly written not optimized..)
372 for (i=0; i<dstW; i++) {
375 for (j=0; j<lumFilterSize; j++)
376 val += lumSrc[j][i] * lumFilter[j];
378 dest[i]= av_clip_uint8(val>>19);
384 if (dstFormat == PIX_FMT_NV12)
385 for (i=0; i<chrDstW; i++) {
389 for (j=0; j<chrFilterSize; j++) {
390 u += chrUSrc[j][i] * chrFilter[j];
391 v += chrVSrc[j][i] * chrFilter[j];
394 uDest[2*i]= av_clip_uint8(u>>19);
395 uDest[2*i+1]= av_clip_uint8(v>>19);
398 for (i=0; i<chrDstW; i++) {
402 for (j=0; j<chrFilterSize; j++) {
403 u += chrUSrc[j][i] * chrFilter[j];
404 v += chrVSrc[j][i] * chrFilter[j];
407 uDest[2*i]= av_clip_uint8(v>>19);
408 uDest[2*i+1]= av_clip_uint8(u>>19);
412 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
413 for (i=0; i<(dstW>>1); i++) {\
419 int av_unused A1, A2;\
420 type av_unused *r, *b, *g;\
423 for (j=0; j<lumFilterSize; j++) {\
424 Y1 += lumSrc[j][i2] * lumFilter[j];\
425 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
427 for (j=0; j<chrFilterSize; j++) {\
428 U += chrUSrc[j][i] * chrFilter[j];\
429 V += chrVSrc[j][i] * chrFilter[j];\
438 for (j=0; j<lumFilterSize; j++) {\
439 A1 += alpSrc[j][i2 ] * lumFilter[j];\
440 A2 += alpSrc[j][i2+1] * lumFilter[j];\
446 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
447 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
448 if ((Y1|Y2|U|V)&256) {\
449 if (Y1>255) Y1=255; \
450 else if (Y1<0)Y1=0; \
451 if (Y2>255) Y2=255; \
452 else if (Y2<0)Y2=0; \
458 if (alpha && ((A1|A2)&256)) {\
459 A1=av_clip_uint8(A1);\
460 A2=av_clip_uint8(A2);\
463 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
464 for (i=0; i<dstW; i++) {\
472 for (j=0; j<lumFilterSize; j++) {\
473 Y += lumSrc[j][i ] * lumFilter[j];\
475 for (j=0; j<chrFilterSize; j++) {\
476 U += chrUSrc[j][i] * chrFilter[j];\
477 V += chrVSrc[j][i] * chrFilter[j];\
484 for (j=0; j<lumFilterSize; j++)\
485 A += alpSrc[j][i ] * lumFilter[j];\
488 A = av_clip_uint8(A);\
491 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
492 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
493 Y-= c->yuv2rgb_y_offset;\
494 Y*= c->yuv2rgb_y_coeff;\
496 R= Y + V*c->yuv2rgb_v2r_coeff;\
497 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
498 B= Y + U*c->yuv2rgb_u2b_coeff;\
499 if ((R|G|B)&(0xC0000000)) {\
500 if (R>=(256<<22)) R=(256<<22)-1; \
502 if (G>=(256<<22)) G=(256<<22)-1; \
504 if (B>=(256<<22)) B=(256<<22)-1; \
508 #define YSCALE_YUV_2_GRAY16_C \
509 for (i=0; i<(dstW>>1); i++) {\
518 for (j=0; j<lumFilterSize; j++) {\
519 Y1 += lumSrc[j][i2] * lumFilter[j];\
520 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
524 if ((Y1|Y2|U|V)&65536) {\
525 if (Y1>65535) Y1=65535; \
526 else if (Y1<0)Y1=0; \
527 if (Y2>65535) Y2=65535; \
528 else if (Y2<0)Y2=0; \
531 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
532 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
533 r = (type *)c->table_rV[V]; \
534 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
535 b = (type *)c->table_bU[U];
537 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
538 for (i=0; i<(dstW>>1); i++) { \
540 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
541 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
542 int U= (ubuf0[i]*uvalpha1+ubuf1[i]*uvalpha)>>19; \
543 int V= (vbuf0[i]*uvalpha1+vbuf1[i]*uvalpha)>>19; \
544 type av_unused *r, *b, *g; \
545 int av_unused A1, A2; \
547 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
548 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
551 #define YSCALE_YUV_2_GRAY16_2_C \
552 for (i=0; i<(dstW>>1); i++) { \
554 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
555 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
557 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
558 YSCALE_YUV_2_PACKED2_C(type,alpha)\
559 r = (type *)c->table_rV[V];\
560 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
561 b = (type *)c->table_bU[U];
563 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
564 for (i=0; i<(dstW>>1); i++) {\
566 int Y1= buf0[i2 ]>>7;\
567 int Y2= buf0[i2+1]>>7;\
568 int U= (ubuf1[i])>>7;\
569 int V= (vbuf1[i])>>7;\
570 type av_unused *r, *b, *g;\
571 int av_unused A1, A2;\
577 #define YSCALE_YUV_2_GRAY16_1_C \
578 for (i=0; i<(dstW>>1); i++) {\
580 int Y1= buf0[i2 ]<<1;\
581 int Y2= buf0[i2+1]<<1;
583 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
584 YSCALE_YUV_2_PACKED1_C(type,alpha)\
585 r = (type *)c->table_rV[V];\
586 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
587 b = (type *)c->table_bU[U];
589 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
590 for (i=0; i<(dstW>>1); i++) {\
592 int Y1= buf0[i2 ]>>7;\
593 int Y2= buf0[i2+1]>>7;\
594 int U= (ubuf0[i] + ubuf1[i])>>8;\
595 int V= (vbuf0[i] + vbuf1[i])>>8;\
596 type av_unused *r, *b, *g;\
597 int av_unused A1, A2;\
603 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
604 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
605 r = (type *)c->table_rV[V];\
606 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
607 b = (type *)c->table_bU[U];
609 #define YSCALE_YUV_2_MONO2_C \
610 const uint8_t * const d128=dither_8x8_220[y&7];\
611 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
612 for (i=0; i<dstW-7; i+=8) {\
614 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
615 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
616 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
617 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
618 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
619 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
620 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
621 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
622 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
626 #define YSCALE_YUV_2_MONOX_C \
627 const uint8_t * const d128=dither_8x8_220[y&7];\
628 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
630 for (i=0; i<dstW-1; i+=2) {\
635 for (j=0; j<lumFilterSize; j++) {\
636 Y1 += lumSrc[j][i] * lumFilter[j];\
637 Y2 += lumSrc[j][i+1] * lumFilter[j];\
647 acc+= acc + g[Y1+d128[(i+0)&7]];\
648 acc+= acc + g[Y2+d128[(i+1)&7]];\
650 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
655 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
656 switch(c->dstFormat) {\
657 case PIX_FMT_RGB48BE:\
658 case PIX_FMT_RGB48LE:\
660 ((uint8_t*)dest)[ 0]= r[Y1];\
661 ((uint8_t*)dest)[ 1]= r[Y1];\
662 ((uint8_t*)dest)[ 2]= g[Y1];\
663 ((uint8_t*)dest)[ 3]= g[Y1];\
664 ((uint8_t*)dest)[ 4]= b[Y1];\
665 ((uint8_t*)dest)[ 5]= b[Y1];\
666 ((uint8_t*)dest)[ 6]= r[Y2];\
667 ((uint8_t*)dest)[ 7]= r[Y2];\
668 ((uint8_t*)dest)[ 8]= g[Y2];\
669 ((uint8_t*)dest)[ 9]= g[Y2];\
670 ((uint8_t*)dest)[10]= b[Y2];\
671 ((uint8_t*)dest)[11]= b[Y2];\
675 case PIX_FMT_BGR48BE:\
676 case PIX_FMT_BGR48LE:\
678 ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
679 ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
680 ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
681 ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
682 ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
683 ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
690 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
691 func(uint32_t,needAlpha)\
692 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
693 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
696 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
698 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
699 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
703 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
704 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
712 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
713 func(uint32_t,needAlpha)\
714 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
715 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
718 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
720 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
721 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
725 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
726 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
733 ((uint8_t*)dest)[0]= r[Y1];\
734 ((uint8_t*)dest)[1]= g[Y1];\
735 ((uint8_t*)dest)[2]= b[Y1];\
736 ((uint8_t*)dest)[3]= r[Y2];\
737 ((uint8_t*)dest)[4]= g[Y2];\
738 ((uint8_t*)dest)[5]= b[Y2];\
744 ((uint8_t*)dest)[0]= b[Y1];\
745 ((uint8_t*)dest)[1]= g[Y1];\
746 ((uint8_t*)dest)[2]= r[Y1];\
747 ((uint8_t*)dest)[3]= b[Y2];\
748 ((uint8_t*)dest)[4]= g[Y2];\
749 ((uint8_t*)dest)[5]= r[Y2];\
753 case PIX_FMT_RGB565BE:\
754 case PIX_FMT_RGB565LE:\
755 case PIX_FMT_BGR565BE:\
756 case PIX_FMT_BGR565LE:\
758 const int dr1= dither_2x2_8[y&1 ][0];\
759 const int dg1= dither_2x2_4[y&1 ][0];\
760 const int db1= dither_2x2_8[(y&1)^1][0];\
761 const int dr2= dither_2x2_8[y&1 ][1];\
762 const int dg2= dither_2x2_4[y&1 ][1];\
763 const int db2= dither_2x2_8[(y&1)^1][1];\
765 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
766 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
770 case PIX_FMT_RGB555BE:\
771 case PIX_FMT_RGB555LE:\
772 case PIX_FMT_BGR555BE:\
773 case PIX_FMT_BGR555LE:\
775 const int dr1= dither_2x2_8[y&1 ][0];\
776 const int dg1= dither_2x2_8[y&1 ][1];\
777 const int db1= dither_2x2_8[(y&1)^1][0];\
778 const int dr2= dither_2x2_8[y&1 ][1];\
779 const int dg2= dither_2x2_8[y&1 ][0];\
780 const int db2= dither_2x2_8[(y&1)^1][1];\
782 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
783 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
787 case PIX_FMT_RGB444BE:\
788 case PIX_FMT_RGB444LE:\
789 case PIX_FMT_BGR444BE:\
790 case PIX_FMT_BGR444LE:\
792 const int dr1= dither_4x4_16[y&3 ][0];\
793 const int dg1= dither_4x4_16[y&3 ][1];\
794 const int db1= dither_4x4_16[(y&3)^3][0];\
795 const int dr2= dither_4x4_16[y&3 ][1];\
796 const int dg2= dither_4x4_16[y&3 ][0];\
797 const int db2= dither_4x4_16[(y&3)^3][1];\
799 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
800 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
807 const uint8_t * const d64= dither_8x8_73[y&7];\
808 const uint8_t * const d32= dither_8x8_32[y&7];\
810 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
811 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
818 const uint8_t * const d64= dither_8x8_73 [y&7];\
819 const uint8_t * const d128=dither_8x8_220[y&7];\
821 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
822 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
826 case PIX_FMT_RGB4_BYTE:\
827 case PIX_FMT_BGR4_BYTE:\
829 const uint8_t * const d64= dither_8x8_73 [y&7];\
830 const uint8_t * const d128=dither_8x8_220[y&7];\
832 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
833 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
837 case PIX_FMT_MONOBLACK:\
838 case PIX_FMT_MONOWHITE:\
843 case PIX_FMT_YUYV422:\
845 ((uint8_t*)dest)[2*i2+0]= Y1;\
846 ((uint8_t*)dest)[2*i2+1]= U;\
847 ((uint8_t*)dest)[2*i2+2]= Y2;\
848 ((uint8_t*)dest)[2*i2+3]= V;\
851 case PIX_FMT_UYVY422:\
853 ((uint8_t*)dest)[2*i2+0]= U;\
854 ((uint8_t*)dest)[2*i2+1]= Y1;\
855 ((uint8_t*)dest)[2*i2+2]= V;\
856 ((uint8_t*)dest)[2*i2+3]= Y2;\
859 case PIX_FMT_GRAY16BE:\
861 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
862 ((uint8_t*)dest)[2*i2+1]= Y1;\
863 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
864 ((uint8_t*)dest)[2*i2+3]= Y2;\
867 case PIX_FMT_GRAY16LE:\
869 ((uint8_t*)dest)[2*i2+0]= Y1;\
870 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
871 ((uint8_t*)dest)[2*i2+2]= Y2;\
872 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
877 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter,
878 const int16_t **lumSrc, int lumFilterSize,
879 const int16_t *chrFilter, const int16_t **chrUSrc,
880 const int16_t **chrVSrc, int chrFilterSize,
881 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
884 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)
887 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter,
888 const int16_t **lumSrc, int lumFilterSize,
889 const int16_t *chrFilter, const int16_t **chrUSrc,
890 const int16_t **chrVSrc, int chrFilterSize,
891 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
894 int step= c->dstFormatBpp/8;
897 switch(c->dstFormat) {
905 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
906 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
907 dest[aidx]= needAlpha ? A : 255;
914 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
915 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
923 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
940 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
941 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
942 dest[aidx]= needAlpha ? A : 255;
949 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
950 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
958 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
973 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
976 uint8_t *ptr = plane + stride*y;
977 for (i=0; i<height; i++) {
978 memset(ptr, val, width);
983 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width,
987 for (i = 0; i < width; i++) {
992 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
996 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
997 const uint8_t *src1, const uint8_t *src2,
998 int width, uint32_t *unused)
1002 for (i = 0; i < width; i++) {
1003 int r = src1[6*i + 0];
1004 int g = src1[6*i + 2];
1005 int b = src1[6*i + 4];
1007 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1008 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1012 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1013 const uint8_t *src1, const uint8_t *src2,
1014 int width, uint32_t *unused)
1018 for (i = 0; i < width; i++) {
1019 int r= src1[12*i + 0] + src1[12*i + 6];
1020 int g= src1[12*i + 2] + src1[12*i + 8];
1021 int b= src1[12*i + 4] + src1[12*i + 10];
1023 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1024 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1028 static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, int width,
1032 for (i = 0; i < width; i++) {
1037 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1041 static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1042 const uint8_t *src1, const uint8_t *src2,
1043 int width, uint32_t *unused)
1046 for (i = 0; i < width; i++) {
1047 int b = src1[6*i + 0];
1048 int g = src1[6*i + 2];
1049 int r = src1[6*i + 4];
1051 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1052 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1056 static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1057 const uint8_t *src1, const uint8_t *src2,
1058 int width, uint32_t *unused)
1061 for (i = 0; i < width; i++) {
1062 int b= src1[12*i + 0] + src1[12*i + 6];
1063 int g= src1[12*i + 2] + src1[12*i + 8];
1064 int r= src1[12*i + 4] + src1[12*i + 10];
1066 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1067 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1071 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1072 static inline void name(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)\
1075 for (i=0; i<width; i++) {\
1076 int b= (((const type*)src)[i]>>shb)&maskb;\
1077 int g= (((const type*)src)[i]>>shg)&maskg;\
1078 int r= (((const type*)src)[i]>>shr)&maskr;\
1080 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1084 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1085 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1086 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1087 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1088 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1089 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1090 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1091 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1093 static inline void abgrToA(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
1096 for (i=0; i<width; i++) {
1101 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1102 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, int width, uint32_t *unused)\
1105 for (i=0; i<width; i++) {\
1106 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1107 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1108 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1110 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1111 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1114 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, int width, uint32_t *unused)\
1117 for (i=0; i<width; i++) {\
1118 int pix0= ((const type*)src)[2*i+0]>>shp;\
1119 int pix1= ((const type*)src)[2*i+1]>>shp;\
1120 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1121 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1122 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1123 g&= maskg|(2*maskg);\
1127 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1128 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1132 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1133 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1134 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1135 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1136 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1137 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1138 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1139 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1141 static inline void palToY(uint8_t *dst, const uint8_t *src, int width, uint32_t *pal)
1144 for (i=0; i<width; i++) {
1147 dst[i]= pal[d] & 0xFF;
1151 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1152 const uint8_t *src1, const uint8_t *src2,
1153 int width, uint32_t *pal)
1156 assert(src1 == src2);
1157 for (i=0; i<width; i++) {
1158 int p= pal[src1[i]];
1165 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
1168 for (i=0; i<width/8; i++) {
1171 dst[8*i+j]= ((d>>(7-j))&1)*255;
1175 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
1178 for (i=0; i<width/8; i++) {
1181 dst[8*i+j]= ((d>>(7-j))&1)*255;
1185 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1188 #define COMPILE_TEMPLATE_MMX2 0
1190 #include "swscale_template.c"
1194 #define RENAME(a) a ## _altivec
1195 #include "ppc/swscale_template.c"
1201 #undef COMPILE_TEMPLATE_MMX2
1202 #define COMPILE_TEMPLATE_MMX2 0
1203 #define RENAME(a) a ## _MMX
1204 #include "x86/swscale_template.c"
1210 #undef COMPILE_TEMPLATE_MMX2
1211 #define COMPILE_TEMPLATE_MMX2 1
1212 #define RENAME(a) a ## _MMX2
1213 #include "x86/swscale_template.c"
1216 SwsFunc ff_getSwsFunc(SwsContext *c)
1218 int cpu_flags = av_get_cpu_flags();
1220 sws_init_swScale_c(c);
1223 if (cpu_flags & AV_CPU_FLAG_MMX)
1224 sws_init_swScale_MMX(c);
1227 if (cpu_flags & AV_CPU_FLAG_MMX2)
1228 sws_init_swScale_MMX2(c);
1231 if (cpu_flags & AV_CPU_FLAG_ALTIVEC)
1232 sws_init_swScale_altivec(c);
1238 static void copyPlane(const uint8_t *src, int srcStride,
1239 int srcSliceY, int srcSliceH, int width,
1240 uint8_t *dst, int dstStride)
1242 dst += dstStride * srcSliceY;
1243 if (dstStride == srcStride && srcStride > 0) {
1244 memcpy(dst, src, srcSliceH * dstStride);
1247 for (i=0; i<srcSliceH; i++) {
1248 memcpy(dst, src, width);
1255 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1256 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1258 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1260 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1261 dstParam[0], dstStride[0]);
1263 if (c->dstFormat == PIX_FMT_NV12)
1264 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1266 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1271 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1272 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1274 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1276 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1281 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1282 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1284 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1286 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1291 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1292 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1294 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1296 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1301 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1302 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1304 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1306 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1311 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1312 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1314 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1315 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1316 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1318 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1321 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1326 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1327 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1329 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1330 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1331 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1333 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1338 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1339 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1341 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1342 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1343 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1345 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1348 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1353 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1354 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1356 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1357 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1358 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1360 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1365 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)
1368 for (i=0; i<num_pixels; i++)
1369 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1372 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)
1376 for (i=0; i<num_pixels; i++)
1377 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1380 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)
1384 for (i=0; i<num_pixels; i++) {
1386 dst[0]= palette[src[i<<1]*4+0];
1387 dst[1]= palette[src[i<<1]*4+1];
1388 dst[2]= palette[src[i<<1]*4+2];
1393 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1394 int srcSliceH, uint8_t* dst[], int dstStride[])
1396 const enum PixelFormat srcFormat= c->srcFormat;
1397 const enum PixelFormat dstFormat= c->dstFormat;
1398 void (*conv)(const uint8_t *src, uint8_t *dst, int num_pixels,
1399 const uint8_t *palette)=NULL;
1401 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1402 const uint8_t *srcPtr= src[0];
1404 if (srcFormat == PIX_FMT_Y400A) {
1405 switch (dstFormat) {
1406 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1407 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1408 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1409 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1410 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1411 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1413 } else if (usePal(srcFormat)) {
1414 switch (dstFormat) {
1415 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1416 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1417 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1418 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1419 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1420 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1425 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1426 sws_format_name(srcFormat), sws_format_name(dstFormat));
1428 for (i=0; i<srcSliceH; i++) {
1429 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1430 srcPtr+= srcStride[0];
1431 dstPtr+= dstStride[0];
1438 #define isRGBA32(x) ( \
1439 (x) == PIX_FMT_ARGB \
1440 || (x) == PIX_FMT_RGBA \
1441 || (x) == PIX_FMT_BGRA \
1442 || (x) == PIX_FMT_ABGR \
1445 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1446 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1447 int srcSliceH, uint8_t* dst[], int dstStride[])
1449 const enum PixelFormat srcFormat= c->srcFormat;
1450 const enum PixelFormat dstFormat= c->dstFormat;
1451 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1452 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1453 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1454 const int dstId= c->dstFormatBpp >> 2;
1455 void (*conv)(const uint8_t *src, uint8_t *dst, int src_size)=NULL;
1457 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1459 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1460 if ( CONV_IS(ABGR, RGBA)
1461 || CONV_IS(ARGB, BGRA)
1462 || CONV_IS(BGRA, ARGB)
1463 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1464 else if (CONV_IS(ABGR, ARGB)
1465 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1466 else if (CONV_IS(ABGR, BGRA)
1467 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1468 else if (CONV_IS(BGRA, RGBA)
1469 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1470 else if (CONV_IS(BGRA, ABGR)
1471 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1474 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1475 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1476 switch(srcId | (dstId<<4)) {
1477 case 0x34: conv= rgb16to15; break;
1478 case 0x36: conv= rgb24to15; break;
1479 case 0x38: conv= rgb32to15; break;
1480 case 0x43: conv= rgb15to16; break;
1481 case 0x46: conv= rgb24to16; break;
1482 case 0x48: conv= rgb32to16; break;
1483 case 0x63: conv= rgb15to24; break;
1484 case 0x64: conv= rgb16to24; break;
1485 case 0x68: conv= rgb32to24; break;
1486 case 0x83: conv= rgb15to32; break;
1487 case 0x84: conv= rgb16to32; break;
1488 case 0x86: conv= rgb24to32; break;
1490 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1491 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1492 switch(srcId | (dstId<<4)) {
1493 case 0x33: conv= rgb15tobgr15; break;
1494 case 0x34: conv= rgb16tobgr15; break;
1495 case 0x36: conv= rgb24tobgr15; break;
1496 case 0x38: conv= rgb32tobgr15; break;
1497 case 0x43: conv= rgb15tobgr16; break;
1498 case 0x44: conv= rgb16tobgr16; break;
1499 case 0x46: conv= rgb24tobgr16; break;
1500 case 0x48: conv= rgb32tobgr16; break;
1501 case 0x63: conv= rgb15tobgr24; break;
1502 case 0x64: conv= rgb16tobgr24; break;
1503 case 0x66: conv= rgb24tobgr24; break;
1504 case 0x68: conv= rgb32tobgr24; break;
1505 case 0x83: conv= rgb15tobgr32; break;
1506 case 0x84: conv= rgb16tobgr32; break;
1507 case 0x86: conv= rgb24tobgr32; break;
1512 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1513 sws_format_name(srcFormat), sws_format_name(dstFormat));
1515 const uint8_t *srcPtr= src[0];
1516 uint8_t *dstPtr= dst[0];
1517 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1518 srcPtr += ALT32_CORR;
1520 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1521 dstPtr += ALT32_CORR;
1523 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1524 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1527 dstPtr += dstStride[0]*srcSliceY;
1529 for (i=0; i<srcSliceH; i++) {
1530 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1531 srcPtr+= srcStride[0];
1532 dstPtr+= dstStride[0];
1539 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1540 int srcSliceH, uint8_t* dst[], int dstStride[])
1544 dst[0]+ srcSliceY *dstStride[0],
1545 dst[1]+(srcSliceY>>1)*dstStride[1],
1546 dst[2]+(srcSliceY>>1)*dstStride[2],
1548 dstStride[0], dstStride[1], srcStride[0]);
1550 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1554 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1555 int srcSliceH, uint8_t* dst[], int dstStride[])
1557 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1558 dst[0], dstStride[0]);
1560 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1561 srcSliceH >> 2, srcStride[1], dstStride[1]);
1562 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1563 srcSliceH >> 2, srcStride[2], dstStride[2]);
1565 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1569 /* unscaled copy like stuff (assumes nearly identical formats) */
1570 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1571 int srcSliceH, uint8_t* dst[], int dstStride[])
1573 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1574 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1577 const uint8_t *srcPtr= src[0];
1578 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1581 /* universal length finder */
1582 while(length+c->srcW <= FFABS(dstStride[0])
1583 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1586 for (i=0; i<srcSliceH; i++) {
1587 memcpy(dstPtr, srcPtr, length);
1588 srcPtr+= srcStride[0];
1589 dstPtr+= dstStride[0];
1595 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1596 int srcSliceH, uint8_t* dst[], int dstStride[])
1599 for (plane=0; plane<4; plane++) {
1600 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1601 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1602 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1603 const uint8_t *srcPtr= src[plane];
1604 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1606 if (!dst[plane]) continue;
1607 // ignore palette for GRAY8
1608 if (plane == 1 && !dst[2]) continue;
1609 if (!src[plane] || (plane == 1 && !src[2])) {
1610 if(is16BPS(c->dstFormat))
1612 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1614 if(is9_OR_10BPS(c->srcFormat)) {
1615 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
1616 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1617 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1619 if (is16BPS(c->dstFormat)) {
1620 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1621 #define COPY9_OR_10TO16(rfunc, wfunc) \
1622 for (i = 0; i < height; i++) { \
1623 for (j = 0; j < length; j++) { \
1624 int srcpx = rfunc(&srcPtr2[j]); \
1625 wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \
1627 dstPtr2 += dstStride[plane]/2; \
1628 srcPtr2 += srcStride[plane]/2; \
1630 if (isBE(c->dstFormat)) {
1631 if (isBE(c->srcFormat)) {
1632 COPY9_OR_10TO16(AV_RB16, AV_WB16);
1634 COPY9_OR_10TO16(AV_RL16, AV_WB16);
1637 if (isBE(c->srcFormat)) {
1638 COPY9_OR_10TO16(AV_RB16, AV_WL16);
1640 COPY9_OR_10TO16(AV_RL16, AV_WL16);
1643 } else if (is9_OR_10BPS(c->dstFormat)) {
1644 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1645 #define COPY9_OR_10TO9_OR_10(loop) \
1646 for (i = 0; i < height; i++) { \
1647 for (j = 0; j < length; j++) { \
1650 dstPtr2 += dstStride[plane]/2; \
1651 srcPtr2 += srcStride[plane]/2; \
1653 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
1654 if (dst_depth > src_depth) { \
1655 COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
1656 wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
1657 } else if (dst_depth < src_depth) { \
1658 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \
1660 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
1662 if (isBE(c->dstFormat)) {
1663 if (isBE(c->srcFormat)) {
1664 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
1666 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
1669 if (isBE(c->srcFormat)) {
1670 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
1672 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
1676 // FIXME Maybe dither instead.
1677 #define COPY9_OR_10TO8(rfunc) \
1678 for (i = 0; i < height; i++) { \
1679 for (j = 0; j < length; j++) { \
1680 dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \
1682 dstPtr += dstStride[plane]; \
1683 srcPtr2 += srcStride[plane]/2; \
1685 if (isBE(c->srcFormat)) {
1686 COPY9_OR_10TO8(AV_RB16);
1688 COPY9_OR_10TO8(AV_RL16);
1691 } else if(is9_OR_10BPS(c->dstFormat)) {
1692 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1693 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1695 if (is16BPS(c->srcFormat)) {
1696 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1697 #define COPY16TO9_OR_10(rfunc, wfunc) \
1698 for (i = 0; i < height; i++) { \
1699 for (j = 0; j < length; j++) { \
1700 wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \
1702 dstPtr2 += dstStride[plane]/2; \
1703 srcPtr2 += srcStride[plane]/2; \
1705 if (isBE(c->dstFormat)) {
1706 if (isBE(c->srcFormat)) {
1707 COPY16TO9_OR_10(AV_RB16, AV_WB16);
1709 COPY16TO9_OR_10(AV_RL16, AV_WB16);
1712 if (isBE(c->srcFormat)) {
1713 COPY16TO9_OR_10(AV_RB16, AV_WL16);
1715 COPY16TO9_OR_10(AV_RL16, AV_WL16);
1719 #define COPY8TO9_OR_10(wfunc) \
1720 for (i = 0; i < height; i++) { \
1721 for (j = 0; j < length; j++) { \
1722 const int srcpx = srcPtr[j]; \
1723 wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \
1725 dstPtr2 += dstStride[plane]/2; \
1726 srcPtr += srcStride[plane]; \
1728 if (isBE(c->dstFormat)) {
1729 COPY8TO9_OR_10(AV_WB16);
1731 COPY8TO9_OR_10(AV_WL16);
1734 } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1735 if (!isBE(c->srcFormat)) srcPtr++;
1736 for (i=0; i<height; i++) {
1737 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1738 srcPtr+= srcStride[plane];
1739 dstPtr+= dstStride[plane];
1741 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1742 for (i=0; i<height; i++) {
1743 for (j=0; j<length; j++) {
1744 dstPtr[ j<<1 ] = srcPtr[j];
1745 dstPtr[(j<<1)+1] = srcPtr[j];
1747 srcPtr+= srcStride[plane];
1748 dstPtr+= dstStride[plane];
1750 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1751 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1753 for (i=0; i<height; i++) {
1754 for (j=0; j<length; j++)
1755 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1756 srcPtr+= srcStride[plane];
1757 dstPtr+= dstStride[plane];
1759 } else if (dstStride[plane] == srcStride[plane] &&
1760 srcStride[plane] > 0 && srcStride[plane] == length) {
1761 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1762 height*dstStride[plane]);
1764 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1766 for (i=0; i<height; i++) {
1767 memcpy(dstPtr, srcPtr, length);
1768 srcPtr+= srcStride[plane];
1769 dstPtr+= dstStride[plane];
1777 void ff_get_unscaled_swscale(SwsContext *c)
1779 const enum PixelFormat srcFormat = c->srcFormat;
1780 const enum PixelFormat dstFormat = c->dstFormat;
1781 const int flags = c->flags;
1782 const int dstH = c->dstH;
1785 needsDither= isAnyRGB(dstFormat)
1786 && c->dstFormatBpp < 24
1787 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1790 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1791 c->swScale= planarToNv12Wrapper;
1794 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1795 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1796 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1799 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1800 c->swScale= yvu9ToYv12Wrapper;
1804 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1805 c->swScale= bgr24ToYv12Wrapper;
1807 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1808 if ( isAnyRGB(srcFormat)
1809 && isAnyRGB(dstFormat)
1810 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1811 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1812 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1813 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1814 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1815 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1816 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1817 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1818 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1819 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1820 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
1821 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
1822 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1823 c->swScale= rgbToRgbWrapper;
1825 if ((usePal(srcFormat) && (
1826 dstFormat == PIX_FMT_RGB32 ||
1827 dstFormat == PIX_FMT_RGB32_1 ||
1828 dstFormat == PIX_FMT_RGB24 ||
1829 dstFormat == PIX_FMT_BGR32 ||
1830 dstFormat == PIX_FMT_BGR32_1 ||
1831 dstFormat == PIX_FMT_BGR24)))
1832 c->swScale= palToRgbWrapper;
1834 if (srcFormat == PIX_FMT_YUV422P) {
1835 if (dstFormat == PIX_FMT_YUYV422)
1836 c->swScale= yuv422pToYuy2Wrapper;
1837 else if (dstFormat == PIX_FMT_UYVY422)
1838 c->swScale= yuv422pToUyvyWrapper;
1841 /* LQ converters if -sws 0 or -sws 4*/
1842 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1844 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1845 if (dstFormat == PIX_FMT_YUYV422)
1846 c->swScale= planarToYuy2Wrapper;
1847 else if (dstFormat == PIX_FMT_UYVY422)
1848 c->swScale= planarToUyvyWrapper;
1851 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1852 c->swScale= yuyvToYuv420Wrapper;
1853 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1854 c->swScale= uyvyToYuv420Wrapper;
1855 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1856 c->swScale= yuyvToYuv422Wrapper;
1857 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1858 c->swScale= uyvyToYuv422Wrapper;
1861 if ((av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) &&
1862 !(c->flags & SWS_BITEXACT) &&
1863 srcFormat == PIX_FMT_YUV420P) {
1864 // unscaled YV12 -> packed YUV, we want speed
1865 if (dstFormat == PIX_FMT_YUYV422)
1866 c->swScale= yv12toyuy2_unscaled_altivec;
1867 else if (dstFormat == PIX_FMT_UYVY422)
1868 c->swScale= yv12touyvy_unscaled_altivec;
1873 if ( srcFormat == dstFormat
1874 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1875 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1876 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1877 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1878 || (isGray(dstFormat) && isGray(srcFormat))
1879 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1880 && c->chrDstHSubSample == c->chrSrcHSubSample
1881 && c->chrDstVSubSample == c->chrSrcVSubSample
1882 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1883 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1885 if (isPacked(c->srcFormat))
1886 c->swScale= packedCopyWrapper;
1887 else /* Planar YUV or gray */
1888 c->swScale= planarCopyWrapper;
1891 ff_bfin_get_unscaled_swscale (c);
1895 static void reset_ptr(const uint8_t* src[], int format)
1897 if(!isALPHA(format))
1899 if(!isPlanarYUV(format)) {
1902 if (!usePal(format))
1907 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
1908 const int linesizes[4])
1910 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
1913 for (i = 0; i < 4; i++) {
1914 int plane = desc->comp[i].plane;
1915 if (!data[plane] || !linesizes[plane])
1923 * swscale wrapper, so we don't need to export the SwsContext.
1924 * Assumes planar YUV to be in YUV order instead of YVU.
1926 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
1927 int srcSliceH, uint8_t* const dst[], const int dstStride[])
1930 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
1931 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
1933 // do not mess up sliceDir if we have a "trailing" 0-size slice
1937 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
1938 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
1941 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
1942 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
1946 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
1947 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
1950 if (c->sliceDir == 0) {
1951 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
1954 if (usePal(c->srcFormat)) {
1955 for (i=0; i<256; i++) {
1956 int p, r, g, b,y,u,v;
1957 if(c->srcFormat == PIX_FMT_PAL8) {
1958 p=((const uint32_t*)(src[1]))[i];
1962 } else if(c->srcFormat == PIX_FMT_RGB8) {
1966 } else if(c->srcFormat == PIX_FMT_BGR8) {
1970 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
1974 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
1977 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
1982 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1983 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1984 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
1985 c->pal_yuv[i]= y + (u<<8) + (v<<16);
1987 switch(c->dstFormat) {
1992 c->pal_rgb[i]= r + (g<<8) + (b<<16);
1994 case PIX_FMT_BGR32_1:
1998 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2000 case PIX_FMT_RGB32_1:
2004 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2011 c->pal_rgb[i]= b + (g<<8) + (r<<16);
2016 // copy strides, so they can safely be modified
2017 if (c->sliceDir == 1) {
2018 // slices go from top to bottom
2019 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2020 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2022 reset_ptr(src2, c->srcFormat);
2023 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2025 /* reset slice direction at end of frame */
2026 if (srcSliceY + srcSliceH == c->srcH)
2029 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2031 // slices go from bottom to top => we flip the image internally
2032 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2033 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2035 src2[0] += (srcSliceH-1)*srcStride[0];
2036 if (!usePal(c->srcFormat))
2037 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2038 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2039 src2[3] += (srcSliceH-1)*srcStride[3];
2040 dst2[0] += ( c->dstH -1)*dstStride[0];
2041 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2042 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2043 dst2[3] += ( c->dstH -1)*dstStride[3];
2045 reset_ptr(src2, c->srcFormat);
2046 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2048 /* reset slice direction at end of frame */
2052 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2056 /* Convert the palette to the same packed 32-bit format as the palette */
2057 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)
2061 for (i=0; i<num_pixels; i++)
2062 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2065 /* Palette format: ABCD -> dst format: ABC */
2066 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, int num_pixels, const uint8_t *palette)
2070 for (i=0; i<num_pixels; i++) {
2072 dst[0]= palette[src[i]*4+0];
2073 dst[1]= palette[src[i]*4+1];
2074 dst[2]= palette[src[i]*4+2];