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/avutil.h"
66 #include "libavutil/mathematics.h"
67 #include "libavutil/bswap.h"
68 #include "libavutil/pixdesc.h"
74 //#define HAVE_AMD3DNOW
79 #define isPacked(x) ( \
81 || (x)==PIX_FMT_YUYV422 \
82 || (x)==PIX_FMT_UYVY422 \
83 || (x)==PIX_FMT_Y400A \
87 #define RGB2YUV_SHIFT 15
88 #define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
89 #define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
90 #define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
91 #define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
92 #define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
93 #define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
94 #define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
95 #define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
96 #define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
98 static const double rgb2yuv_table[8][9]={
99 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
100 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5}, //ITU709
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 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
104 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
105 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //DEFAULT / ITU601 / ITU624 / SMPTE 170M
106 {0.701 , 0.087 , 0.212 , -0.384, 0.5, -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
111 Special versions: fast Y 1:1 scaling (no interpolation in y direction)
114 more intelligent misalignment avoidance for the horizontal scaler
115 write special vertical cubic upscale version
116 optimize C code (YV12 / minmax)
117 add support for packed pixel YUV input & output
118 add support for Y8 output
119 optimize BGR24 & BGR32
120 add BGR4 output support
121 write special BGR->BGR scaler
124 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4)[2][8]={
125 { 1, 3, 1, 3, 1, 3, 1, 3, },
126 { 2, 0, 2, 0, 2, 0, 2, 0, },
129 DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8)[2][8]={
130 { 6, 2, 6, 2, 6, 2, 6, 2, },
131 { 0, 4, 0, 4, 0, 4, 0, 4, },
134 DECLARE_ALIGNED(8, const uint8_t, dither_4x4_16)[4][8]={
135 { 8, 4, 11, 7, 8, 4, 11, 7, },
136 { 2, 14, 1, 13, 2, 14, 1, 13, },
137 { 10, 6, 9, 5, 10, 6, 9, 5, },
138 { 0, 12, 3, 15, 0, 12, 3, 15, },
141 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32)[8][8]={
142 { 17, 9, 23, 15, 16, 8, 22, 14, },
143 { 5, 29, 3, 27, 4, 28, 2, 26, },
144 { 21, 13, 19, 11, 20, 12, 18, 10, },
145 { 0, 24, 6, 30, 1, 25, 7, 31, },
146 { 16, 8, 22, 14, 17, 9, 23, 15, },
147 { 4, 28, 2, 26, 5, 29, 3, 27, },
148 { 20, 12, 18, 10, 21, 13, 19, 11, },
149 { 1, 25, 7, 31, 0, 24, 6, 30, },
152 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73)[8][8]={
153 { 0, 55, 14, 68, 3, 58, 17, 72, },
154 { 37, 18, 50, 32, 40, 22, 54, 35, },
155 { 9, 64, 5, 59, 13, 67, 8, 63, },
156 { 46, 27, 41, 23, 49, 31, 44, 26, },
157 { 2, 57, 16, 71, 1, 56, 15, 70, },
158 { 39, 21, 52, 34, 38, 19, 51, 33, },
159 { 11, 66, 7, 62, 10, 65, 6, 60, },
160 { 48, 30, 43, 25, 47, 29, 42, 24, },
164 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
165 {117, 62, 158, 103, 113, 58, 155, 100, },
166 { 34, 199, 21, 186, 31, 196, 17, 182, },
167 {144, 89, 131, 76, 141, 86, 127, 72, },
168 { 0, 165, 41, 206, 10, 175, 52, 217, },
169 {110, 55, 151, 96, 120, 65, 162, 107, },
170 { 28, 193, 14, 179, 38, 203, 24, 189, },
171 {138, 83, 124, 69, 148, 93, 134, 79, },
172 { 7, 172, 48, 213, 3, 168, 45, 210, },
175 // tries to correct a gamma of 1.5
176 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
177 { 0, 143, 18, 200, 2, 156, 25, 215, },
178 { 78, 28, 125, 64, 89, 36, 138, 74, },
179 { 10, 180, 3, 161, 16, 195, 8, 175, },
180 {109, 51, 93, 38, 121, 60, 105, 47, },
181 { 1, 152, 23, 210, 0, 147, 20, 205, },
182 { 85, 33, 134, 71, 81, 30, 130, 67, },
183 { 14, 190, 6, 171, 12, 185, 5, 166, },
184 {117, 57, 101, 44, 113, 54, 97, 41, },
187 // tries to correct a gamma of 2.0
188 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
189 { 0, 124, 8, 193, 0, 140, 12, 213, },
190 { 55, 14, 104, 42, 66, 19, 119, 52, },
191 { 3, 168, 1, 145, 6, 187, 3, 162, },
192 { 86, 31, 70, 21, 99, 39, 82, 28, },
193 { 0, 134, 11, 206, 0, 129, 9, 200, },
194 { 62, 17, 114, 48, 58, 16, 109, 45, },
195 { 5, 181, 2, 157, 4, 175, 1, 151, },
196 { 95, 36, 78, 26, 90, 34, 74, 24, },
199 // tries to correct a gamma of 2.5
200 DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
201 { 0, 107, 3, 187, 0, 125, 6, 212, },
202 { 39, 7, 86, 28, 49, 11, 102, 36, },
203 { 1, 158, 0, 131, 3, 180, 1, 151, },
204 { 68, 19, 52, 12, 81, 25, 64, 17, },
205 { 0, 119, 5, 203, 0, 113, 4, 195, },
206 { 45, 9, 96, 33, 42, 8, 91, 30, },
207 { 2, 172, 1, 144, 2, 165, 0, 137, },
208 { 77, 23, 60, 15, 72, 21, 56, 14, },
212 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
213 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
214 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
215 int dstW, int chrDstW, int big_endian, int output_bits)
217 //FIXME Optimize (just quickly written not optimized..)
219 int shift = 11 + 16 - output_bits;
221 #define output_pixel(pos, val) \
223 if (output_bits == 16) { \
224 AV_WB16(pos, av_clip_uint16(val >> shift)); \
226 AV_WB16(pos, av_clip_uintp2(val >> shift, output_bits)); \
229 if (output_bits == 16) { \
230 AV_WL16(pos, av_clip_uint16(val >> shift)); \
232 AV_WL16(pos, av_clip_uintp2(val >> shift, output_bits)); \
235 for (i = 0; i < dstW; i++) {
239 for (j = 0; j < lumFilterSize; j++)
240 val += lumSrc[j][i] * lumFilter[j];
242 output_pixel(&dest[i], val);
246 for (i = 0; i < chrDstW; i++) {
251 for (j = 0; j < chrFilterSize; j++) {
252 u += chrSrc[j][i ] * chrFilter[j];
253 v += chrSrc[j][i + VOFW] * chrFilter[j];
256 output_pixel(&uDest[i], u);
257 output_pixel(&vDest[i], v);
261 if (CONFIG_SWSCALE_ALPHA && aDest) {
262 for (i = 0; i < dstW; i++) {
266 for (j = 0; j < lumFilterSize; j++)
267 val += alpSrc[j][i] * lumFilter[j];
269 output_pixel(&aDest[i], val);
274 #define yuv2NBPS(bits, BE_LE, is_be) \
275 static void yuv2yuvX ## bits ## BE_LE ## _c(const int16_t *lumFilter, \
276 const int16_t **lumSrc, int lumFilterSize, \
277 const int16_t *chrFilter, const int16_t **chrSrc, \
278 int chrFilterSize, const int16_t **alpSrc, \
279 uint16_t *dest, uint16_t *uDest, uint16_t *vDest, \
280 uint16_t *aDest, int dstW, int chrDstW) \
282 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \
283 chrFilter, chrSrc, chrFilterSize, \
285 dest, uDest, vDest, aDest, \
286 dstW, chrDstW, is_be, bits); \
295 static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
296 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
297 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
298 enum PixelFormat dstFormat)
300 #define conv16(bits) \
301 if (isBE(dstFormat)) { \
302 yuv2yuvX ## bits ## BE_c(lumFilter, lumSrc, lumFilterSize, \
303 chrFilter, chrSrc, chrFilterSize, \
305 dest, uDest, vDest, aDest, \
308 yuv2yuvX ## bits ## LE_c(lumFilter, lumSrc, lumFilterSize, \
309 chrFilter, chrSrc, chrFilterSize, \
311 dest, uDest, vDest, aDest, \
314 if (is16BPS(dstFormat)) {
316 } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
324 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
325 const int16_t *chrFilter, const int16_t **chrSrc, 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 += chrSrc[j][i] * chrFilter[j];
346 v += chrSrc[j][i + VOFW] * 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 **chrSrc, int chrFilterSize,
367 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
369 //FIXME Optimize (just quickly written not optimized..)
371 for (i=0; i<dstW; i++) {
374 for (j=0; j<lumFilterSize; j++)
375 val += lumSrc[j][i] * lumFilter[j];
377 dest[i]= av_clip_uint8(val>>19);
383 if (dstFormat == PIX_FMT_NV12)
384 for (i=0; i<chrDstW; i++) {
388 for (j=0; j<chrFilterSize; j++) {
389 u += chrSrc[j][i] * chrFilter[j];
390 v += chrSrc[j][i + VOFW] * chrFilter[j];
393 uDest[2*i]= av_clip_uint8(u>>19);
394 uDest[2*i+1]= av_clip_uint8(v>>19);
397 for (i=0; i<chrDstW; i++) {
401 for (j=0; j<chrFilterSize; j++) {
402 u += chrSrc[j][i] * chrFilter[j];
403 v += chrSrc[j][i + VOFW] * chrFilter[j];
406 uDest[2*i]= av_clip_uint8(v>>19);
407 uDest[2*i+1]= av_clip_uint8(u>>19);
411 #define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
412 for (i=0; i<(dstW>>1); i++) {\
418 int av_unused A1, A2;\
419 type av_unused *r, *b, *g;\
422 for (j=0; j<lumFilterSize; j++) {\
423 Y1 += lumSrc[j][i2] * lumFilter[j];\
424 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
426 for (j=0; j<chrFilterSize; j++) {\
427 U += chrSrc[j][i] * chrFilter[j];\
428 V += chrSrc[j][i+VOFW] * chrFilter[j];\
437 for (j=0; j<lumFilterSize; j++) {\
438 A1 += alpSrc[j][i2 ] * lumFilter[j];\
439 A2 += alpSrc[j][i2+1] * lumFilter[j];\
445 #define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
446 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
447 if ((Y1|Y2|U|V)&256) {\
448 if (Y1>255) Y1=255; \
449 else if (Y1<0)Y1=0; \
450 if (Y2>255) Y2=255; \
451 else if (Y2<0)Y2=0; \
457 if (alpha && ((A1|A2)&256)) {\
458 A1=av_clip_uint8(A1);\
459 A2=av_clip_uint8(A2);\
462 #define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
463 for (i=0; i<dstW; i++) {\
471 for (j=0; j<lumFilterSize; j++) {\
472 Y += lumSrc[j][i ] * lumFilter[j];\
474 for (j=0; j<chrFilterSize; j++) {\
475 U += chrSrc[j][i ] * chrFilter[j];\
476 V += chrSrc[j][i+VOFW] * chrFilter[j];\
483 for (j=0; j<lumFilterSize; j++)\
484 A += alpSrc[j][i ] * lumFilter[j];\
487 A = av_clip_uint8(A);\
490 #define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
491 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
492 Y-= c->yuv2rgb_y_offset;\
493 Y*= c->yuv2rgb_y_coeff;\
495 R= Y + V*c->yuv2rgb_v2r_coeff;\
496 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
497 B= Y + U*c->yuv2rgb_u2b_coeff;\
498 if ((R|G|B)&(0xC0000000)) {\
499 if (R>=(256<<22)) R=(256<<22)-1; \
501 if (G>=(256<<22)) G=(256<<22)-1; \
503 if (B>=(256<<22)) B=(256<<22)-1; \
507 #define YSCALE_YUV_2_GRAY16_C \
508 for (i=0; i<(dstW>>1); i++) {\
517 for (j=0; j<lumFilterSize; j++) {\
518 Y1 += lumSrc[j][i2] * lumFilter[j];\
519 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
523 if ((Y1|Y2|U|V)&65536) {\
524 if (Y1>65535) Y1=65535; \
525 else if (Y1<0)Y1=0; \
526 if (Y2>65535) Y2=65535; \
527 else if (Y2<0)Y2=0; \
530 #define YSCALE_YUV_2_RGBX_C(type,alpha) \
531 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
532 r = (type *)c->table_rV[V]; \
533 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
534 b = (type *)c->table_bU[U];
536 #define YSCALE_YUV_2_PACKED2_C(type,alpha) \
537 for (i=0; i<(dstW>>1); i++) { \
539 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
540 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
541 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
542 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
543 type av_unused *r, *b, *g; \
544 int av_unused A1, A2; \
546 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
547 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
550 #define YSCALE_YUV_2_GRAY16_2_C \
551 for (i=0; i<(dstW>>1); i++) { \
553 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
554 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
556 #define YSCALE_YUV_2_RGB2_C(type,alpha) \
557 YSCALE_YUV_2_PACKED2_C(type,alpha)\
558 r = (type *)c->table_rV[V];\
559 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
560 b = (type *)c->table_bU[U];
562 #define YSCALE_YUV_2_PACKED1_C(type,alpha) \
563 for (i=0; i<(dstW>>1); i++) {\
565 int Y1= buf0[i2 ]>>7;\
566 int Y2= buf0[i2+1]>>7;\
567 int U= (uvbuf1[i ])>>7;\
568 int V= (uvbuf1[i+VOFW])>>7;\
569 type av_unused *r, *b, *g;\
570 int av_unused A1, A2;\
576 #define YSCALE_YUV_2_GRAY16_1_C \
577 for (i=0; i<(dstW>>1); i++) {\
579 int Y1= buf0[i2 ]<<1;\
580 int Y2= buf0[i2+1]<<1;
582 #define YSCALE_YUV_2_RGB1_C(type,alpha) \
583 YSCALE_YUV_2_PACKED1_C(type,alpha)\
584 r = (type *)c->table_rV[V];\
585 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
586 b = (type *)c->table_bU[U];
588 #define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
589 for (i=0; i<(dstW>>1); i++) {\
591 int Y1= buf0[i2 ]>>7;\
592 int Y2= buf0[i2+1]>>7;\
593 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
594 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
595 type av_unused *r, *b, *g;\
596 int av_unused A1, A2;\
602 #define YSCALE_YUV_2_RGB1B_C(type,alpha) \
603 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
604 r = (type *)c->table_rV[V];\
605 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
606 b = (type *)c->table_bU[U];
608 #define YSCALE_YUV_2_MONO2_C \
609 const uint8_t * const d128=dither_8x8_220[y&7];\
610 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
611 for (i=0; i<dstW-7; i+=8) {\
613 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
614 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
615 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
616 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
617 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
618 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
619 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
620 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
621 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
625 #define YSCALE_YUV_2_MONOX_C \
626 const uint8_t * const d128=dither_8x8_220[y&7];\
627 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
629 for (i=0; i<dstW-1; i+=2) {\
634 for (j=0; j<lumFilterSize; j++) {\
635 Y1 += lumSrc[j][i] * lumFilter[j];\
636 Y2 += lumSrc[j][i+1] * lumFilter[j];\
646 acc+= acc + g[Y1+d128[(i+0)&7]];\
647 acc+= acc + g[Y2+d128[(i+1)&7]];\
649 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
654 #define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
655 switch(c->dstFormat) {\
656 case PIX_FMT_RGB48BE:\
657 case PIX_FMT_RGB48LE:\
659 ((uint8_t*)dest)[ 0]= r[Y1];\
660 ((uint8_t*)dest)[ 1]= r[Y1];\
661 ((uint8_t*)dest)[ 2]= g[Y1];\
662 ((uint8_t*)dest)[ 3]= g[Y1];\
663 ((uint8_t*)dest)[ 4]= b[Y1];\
664 ((uint8_t*)dest)[ 5]= b[Y1];\
665 ((uint8_t*)dest)[ 6]= r[Y2];\
666 ((uint8_t*)dest)[ 7]= r[Y2];\
667 ((uint8_t*)dest)[ 8]= g[Y2];\
668 ((uint8_t*)dest)[ 9]= g[Y2];\
669 ((uint8_t*)dest)[10]= b[Y2];\
670 ((uint8_t*)dest)[11]= b[Y2];\
674 case PIX_FMT_BGR48BE:\
675 case PIX_FMT_BGR48LE:\
677 ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
678 ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
679 ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
680 ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
681 ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
682 ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
689 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
690 func(uint32_t,needAlpha)\
691 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
692 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
695 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
697 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
698 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
702 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
703 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
711 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
712 func(uint32_t,needAlpha)\
713 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
714 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
717 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
719 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
720 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
724 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
725 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
732 ((uint8_t*)dest)[0]= r[Y1];\
733 ((uint8_t*)dest)[1]= g[Y1];\
734 ((uint8_t*)dest)[2]= b[Y1];\
735 ((uint8_t*)dest)[3]= r[Y2];\
736 ((uint8_t*)dest)[4]= g[Y2];\
737 ((uint8_t*)dest)[5]= b[Y2];\
743 ((uint8_t*)dest)[0]= b[Y1];\
744 ((uint8_t*)dest)[1]= g[Y1];\
745 ((uint8_t*)dest)[2]= r[Y1];\
746 ((uint8_t*)dest)[3]= b[Y2];\
747 ((uint8_t*)dest)[4]= g[Y2];\
748 ((uint8_t*)dest)[5]= r[Y2];\
752 case PIX_FMT_RGB565BE:\
753 case PIX_FMT_RGB565LE:\
754 case PIX_FMT_BGR565BE:\
755 case PIX_FMT_BGR565LE:\
757 const int dr1= dither_2x2_8[y&1 ][0];\
758 const int dg1= dither_2x2_4[y&1 ][0];\
759 const int db1= dither_2x2_8[(y&1)^1][0];\
760 const int dr2= dither_2x2_8[y&1 ][1];\
761 const int dg2= dither_2x2_4[y&1 ][1];\
762 const int db2= dither_2x2_8[(y&1)^1][1];\
764 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
765 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
769 case PIX_FMT_RGB555BE:\
770 case PIX_FMT_RGB555LE:\
771 case PIX_FMT_BGR555BE:\
772 case PIX_FMT_BGR555LE:\
774 const int dr1= dither_2x2_8[y&1 ][0];\
775 const int dg1= dither_2x2_8[y&1 ][1];\
776 const int db1= dither_2x2_8[(y&1)^1][0];\
777 const int dr2= dither_2x2_8[y&1 ][1];\
778 const int dg2= dither_2x2_8[y&1 ][0];\
779 const int db2= dither_2x2_8[(y&1)^1][1];\
781 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
782 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
786 case PIX_FMT_RGB444BE:\
787 case PIX_FMT_RGB444LE:\
788 case PIX_FMT_BGR444BE:\
789 case PIX_FMT_BGR444LE:\
791 const int dr1= dither_4x4_16[y&3 ][0];\
792 const int dg1= dither_4x4_16[y&3 ][1];\
793 const int db1= dither_4x4_16[(y&3)^3][0];\
794 const int dr2= dither_4x4_16[y&3 ][1];\
795 const int dg2= dither_4x4_16[y&3 ][0];\
796 const int db2= dither_4x4_16[(y&3)^3][1];\
798 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
799 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
806 const uint8_t * const d64= dither_8x8_73[y&7];\
807 const uint8_t * const d32= dither_8x8_32[y&7];\
809 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
810 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
817 const uint8_t * const d64= dither_8x8_73 [y&7];\
818 const uint8_t * const d128=dither_8x8_220[y&7];\
820 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
821 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
825 case PIX_FMT_RGB4_BYTE:\
826 case PIX_FMT_BGR4_BYTE:\
828 const uint8_t * const d64= dither_8x8_73 [y&7];\
829 const uint8_t * const d128=dither_8x8_220[y&7];\
831 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
832 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
836 case PIX_FMT_MONOBLACK:\
837 case PIX_FMT_MONOWHITE:\
842 case PIX_FMT_YUYV422:\
844 ((uint8_t*)dest)[2*i2+0]= Y1;\
845 ((uint8_t*)dest)[2*i2+1]= U;\
846 ((uint8_t*)dest)[2*i2+2]= Y2;\
847 ((uint8_t*)dest)[2*i2+3]= V;\
850 case PIX_FMT_UYVY422:\
852 ((uint8_t*)dest)[2*i2+0]= U;\
853 ((uint8_t*)dest)[2*i2+1]= Y1;\
854 ((uint8_t*)dest)[2*i2+2]= V;\
855 ((uint8_t*)dest)[2*i2+3]= Y2;\
858 case PIX_FMT_GRAY16BE:\
860 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
861 ((uint8_t*)dest)[2*i2+1]= Y1;\
862 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
863 ((uint8_t*)dest)[2*i2+3]= Y2;\
866 case PIX_FMT_GRAY16LE:\
868 ((uint8_t*)dest)[2*i2+0]= Y1;\
869 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
870 ((uint8_t*)dest)[2*i2+2]= Y2;\
871 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
876 static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
877 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
878 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
881 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)
884 static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
885 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
886 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
889 int step= c->dstFormatBpp/8;
892 switch(c->dstFormat) {
900 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
901 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
902 dest[aidx]= needAlpha ? A : 255;
909 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
910 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
918 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
935 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
936 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
937 dest[aidx]= needAlpha ? A : 255;
944 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
945 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
953 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
968 static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
971 uint8_t *ptr = plane + stride*y;
972 for (i=0; i<height; i++) {
973 memset(ptr, val, width);
978 static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, long width,
982 for (i = 0; i < width; i++) {
987 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
991 static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
992 const uint8_t *src1, const uint8_t *src2,
993 long width, uint32_t *unused)
997 for (i = 0; i < width; i++) {
998 int r = src1[6*i + 0];
999 int g = src1[6*i + 2];
1000 int b = src1[6*i + 4];
1002 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1003 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1007 static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1008 const uint8_t *src1, const uint8_t *src2,
1009 long width, uint32_t *unused)
1013 for (i = 0; i < width; i++) {
1014 int r= src1[12*i + 0] + src1[12*i + 6];
1015 int g= src1[12*i + 2] + src1[12*i + 8];
1016 int b= src1[12*i + 4] + src1[12*i + 10];
1018 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1019 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1023 static inline void bgr48ToY(uint8_t *dst, const uint8_t *src, long width,
1027 for (i = 0; i < width; i++) {
1032 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1036 static inline void bgr48ToUV(uint8_t *dstU, uint8_t *dstV,
1037 const uint8_t *src1, const uint8_t *src2,
1038 long width, uint32_t *unused)
1041 for (i = 0; i < width; i++) {
1042 int b = src1[6*i + 0];
1043 int g = src1[6*i + 2];
1044 int r = src1[6*i + 4];
1046 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1047 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1051 static inline void bgr48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1052 const uint8_t *src1, const uint8_t *src2,
1053 long width, uint32_t *unused)
1056 for (i = 0; i < width; i++) {
1057 int b= src1[12*i + 0] + src1[12*i + 6];
1058 int g= src1[12*i + 2] + src1[12*i + 8];
1059 int r= src1[12*i + 4] + src1[12*i + 10];
1061 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1062 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1066 #define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1067 static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1070 for (i=0; i<width; i++) {\
1071 int b= (((const type*)src)[i]>>shb)&maskb;\
1072 int g= (((const type*)src)[i]>>shg)&maskg;\
1073 int r= (((const type*)src)[i]>>shr)&maskr;\
1075 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1079 BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1080 BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1081 BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1082 BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY , GY<<8, BY , RGB2YUV_SHIFT+8)
1083 BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1084 BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1085 BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1086 BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1088 static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1091 for (i=0; i<width; i++) {
1096 #define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
1097 static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1100 for (i=0; i<width; i++) {\
1101 int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
1102 int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
1103 int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
1105 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1106 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1109 static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1112 for (i=0; i<width; i++) {\
1113 int pix0= ((const type*)src)[2*i+0]>>shp;\
1114 int pix1= ((const type*)src)[2*i+1]>>shp;\
1115 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1116 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1117 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1118 g&= maskg|(2*maskg);\
1122 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1123 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1127 BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1128 BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1129 BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1130 BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1131 BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1132 BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1133 BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1134 BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1136 static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1139 for (i=0; i<width; i++) {
1142 dst[i]= pal[d] & 0xFF;
1146 static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1147 const uint8_t *src1, const uint8_t *src2,
1148 long width, uint32_t *pal)
1151 assert(src1 == src2);
1152 for (i=0; i<width; i++) {
1153 int p= pal[src1[i]];
1160 static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1163 for (i=0; i<width/8; i++) {
1166 dst[8*i+j]= ((d>>(7-j))&1)*255;
1170 static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1173 for (i=0; i<width/8; i++) {
1176 dst[8*i+j]= ((d>>(7-j))&1)*255;
1180 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
1182 #if CONFIG_RUNTIME_CPUDETECT
1183 # define COMPILE_C 1
1185 # define COMPILE_MMX HAVE_MMX
1186 # define COMPILE_MMX2 HAVE_MMX2
1187 # define COMPILE_3DNOW HAVE_AMD3DNOW
1189 # define COMPILE_ALTIVEC HAVE_ALTIVEC
1191 #else /* CONFIG_RUNTIME_CPUDETECT */
1194 # define COMPILE_MMX2 1
1195 # elif HAVE_AMD3DNOW
1196 # define COMPILE_3DNOW 1
1198 # define COMPILE_MMX 1
1200 # define COMPILE_C 1
1202 # elif ARCH_PPC && HAVE_ALTIVEC
1203 # define COMPILE_ALTIVEC 1
1205 # define COMPILE_C 1
1210 # define COMPILE_C 0
1213 # define COMPILE_MMX 0
1215 #ifndef COMPILE_MMX2
1216 # define COMPILE_MMX2 0
1218 #ifndef COMPILE_3DNOW
1219 # define COMPILE_3DNOW 0
1221 #ifndef COMPILE_ALTIVEC
1222 # define COMPILE_ALTIVEC 0
1225 #define COMPILE_TEMPLATE_MMX 0
1226 #define COMPILE_TEMPLATE_MMX2 0
1227 #define COMPILE_TEMPLATE_AMD3DNOW 0
1228 #define COMPILE_TEMPLATE_ALTIVEC 0
1230 #include "swscale_template.c"
1234 #undef COMPILE_TEMPLATE_ALTIVEC
1235 #define COMPILE_TEMPLATE_ALTIVEC 1
1236 #define RENAME(a) a ## _altivec
1237 #include "ppc/swscale_template.c"
1245 #undef COMPILE_TEMPLATE_MMX
1246 #undef COMPILE_TEMPLATE_MMX2
1247 #undef COMPILE_TEMPLATE_AMD3DNOW
1248 #define COMPILE_TEMPLATE_MMX 1
1249 #define COMPILE_TEMPLATE_MMX2 0
1250 #define COMPILE_TEMPLATE_AMD3DNOW 0
1251 #define RENAME(a) a ## _MMX
1252 #include "x86/swscale_template.c"
1258 #undef COMPILE_TEMPLATE_MMX
1259 #undef COMPILE_TEMPLATE_MMX2
1260 #undef COMPILE_TEMPLATE_AMD3DNOW
1261 #define COMPILE_TEMPLATE_MMX 1
1262 #define COMPILE_TEMPLATE_MMX2 1
1263 #define COMPILE_TEMPLATE_AMD3DNOW 0
1264 #define RENAME(a) a ## _MMX2
1265 #include "x86/swscale_template.c"
1271 #undef COMPILE_TEMPLATE_MMX
1272 #undef COMPILE_TEMPLATE_MMX2
1273 #undef COMPILE_TEMPLATE_AMD3DNOW
1274 #define COMPILE_TEMPLATE_MMX 1
1275 #define COMPILE_TEMPLATE_MMX2 0
1276 #define COMPILE_TEMPLATE_AMD3DNOW 1
1277 #define RENAME(a) a ## _3DNow
1278 #include "x86/swscale_template.c"
1283 SwsFunc ff_getSwsFunc(SwsContext *c)
1285 sws_init_swScale_c(c);
1287 #if CONFIG_RUNTIME_CPUDETECT
1289 // ordered per speed fastest first
1290 if (c->flags & SWS_CPU_CAPS_MMX2) {
1291 sws_init_swScale_MMX2(c);
1292 return swScale_MMX2;
1293 } else if (c->flags & SWS_CPU_CAPS_3DNOW) {
1294 sws_init_swScale_3DNow(c);
1295 return swScale_3DNow;
1296 } else if (c->flags & SWS_CPU_CAPS_MMX) {
1297 sws_init_swScale_MMX(c);
1303 if (c->flags & SWS_CPU_CAPS_ALTIVEC) {
1304 sws_init_swScale_altivec(c);
1305 return swScale_altivec;
1308 #endif /* ARCH_X86 */
1309 #else //CONFIG_RUNTIME_CPUDETECT
1310 #if COMPILE_TEMPLATE_MMX2
1311 sws_init_swScale_MMX2(c);
1312 return swScale_MMX2;
1313 #elif COMPILE_TEMPLATE_AMD3DNOW
1314 sws_init_swScale_3DNow(c);
1315 return swScale_3DNow;
1316 #elif COMPILE_TEMPLATE_MMX
1317 sws_init_swScale_MMX(c);
1319 #elif COMPILE_TEMPLATE_ALTIVEC
1320 sws_init_swScale_altivec(c);
1321 return swScale_altivec;
1323 #endif //!CONFIG_RUNTIME_CPUDETECT
1328 static void copyPlane(const uint8_t *src, int srcStride,
1329 int srcSliceY, int srcSliceH, int width,
1330 uint8_t *dst, int dstStride)
1332 dst += dstStride * srcSliceY;
1333 if (dstStride == srcStride && srcStride > 0) {
1334 memcpy(dst, src, srcSliceH * dstStride);
1337 for (i=0; i<srcSliceH; i++) {
1338 memcpy(dst, src, width);
1345 static int planarToNv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1346 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1348 uint8_t *dst = dstParam[1] + dstStride[1]*srcSliceY/2;
1350 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1351 dstParam[0], dstStride[0]);
1353 if (c->dstFormat == PIX_FMT_NV12)
1354 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
1356 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
1361 static int planarToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1362 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1364 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1366 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1371 static int planarToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1372 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1374 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1376 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
1381 static int yuv422pToYuy2Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1382 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1384 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1386 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1391 static int yuv422pToUyvyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1392 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1394 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1396 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
1401 static int yuyvToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1402 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1404 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1405 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1406 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1408 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1411 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1416 static int yuyvToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1417 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1419 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1420 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1421 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1423 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1428 static int uyvyToYuv420Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1429 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1431 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1432 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
1433 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
1435 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1438 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1443 static int uyvyToYuv422Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1444 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1446 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
1447 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
1448 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
1450 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
1455 static void gray8aToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1458 for (i=0; i<num_pixels; i++)
1459 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | (src[(i<<1)+1] << 24);
1462 static void gray8aToPacked32_1(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1466 for (i=0; i<num_pixels; i++)
1467 ((uint32_t *) dst)[i] = ((const uint32_t *)palette)[src[i<<1]] | src[(i<<1)+1];
1470 static void gray8aToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
1474 for (i=0; i<num_pixels; i++) {
1476 dst[0]= palette[src[i<<1]*4+0];
1477 dst[1]= palette[src[i<<1]*4+1];
1478 dst[2]= palette[src[i<<1]*4+2];
1483 static int palToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1484 int srcSliceH, uint8_t* dst[], int dstStride[])
1486 const enum PixelFormat srcFormat= c->srcFormat;
1487 const enum PixelFormat dstFormat= c->dstFormat;
1488 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
1489 const uint8_t *palette)=NULL;
1491 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1492 const uint8_t *srcPtr= src[0];
1494 if (srcFormat == PIX_FMT_Y400A) {
1495 switch (dstFormat) {
1496 case PIX_FMT_RGB32 : conv = gray8aToPacked32; break;
1497 case PIX_FMT_BGR32 : conv = gray8aToPacked32; break;
1498 case PIX_FMT_BGR32_1: conv = gray8aToPacked32_1; break;
1499 case PIX_FMT_RGB32_1: conv = gray8aToPacked32_1; break;
1500 case PIX_FMT_RGB24 : conv = gray8aToPacked24; break;
1501 case PIX_FMT_BGR24 : conv = gray8aToPacked24; break;
1503 } else if (usePal(srcFormat)) {
1504 switch (dstFormat) {
1505 case PIX_FMT_RGB32 : conv = sws_convertPalette8ToPacked32; break;
1506 case PIX_FMT_BGR32 : conv = sws_convertPalette8ToPacked32; break;
1507 case PIX_FMT_BGR32_1: conv = sws_convertPalette8ToPacked32; break;
1508 case PIX_FMT_RGB32_1: conv = sws_convertPalette8ToPacked32; break;
1509 case PIX_FMT_RGB24 : conv = sws_convertPalette8ToPacked24; break;
1510 case PIX_FMT_BGR24 : conv = sws_convertPalette8ToPacked24; break;
1515 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1516 sws_format_name(srcFormat), sws_format_name(dstFormat));
1518 for (i=0; i<srcSliceH; i++) {
1519 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
1520 srcPtr+= srcStride[0];
1521 dstPtr+= dstStride[0];
1528 #define isRGBA32(x) ( \
1529 (x) == PIX_FMT_ARGB \
1530 || (x) == PIX_FMT_RGBA \
1531 || (x) == PIX_FMT_BGRA \
1532 || (x) == PIX_FMT_ABGR \
1535 /* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
1536 static int rgbToRgbWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1537 int srcSliceH, uint8_t* dst[], int dstStride[])
1539 const enum PixelFormat srcFormat= c->srcFormat;
1540 const enum PixelFormat dstFormat= c->dstFormat;
1541 const int srcBpp= (c->srcFormatBpp + 7) >> 3;
1542 const int dstBpp= (c->dstFormatBpp + 7) >> 3;
1543 const int srcId= c->srcFormatBpp >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
1544 const int dstId= c->dstFormatBpp >> 2;
1545 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
1547 #define CONV_IS(src, dst) (srcFormat == PIX_FMT_##src && dstFormat == PIX_FMT_##dst)
1549 if (isRGBA32(srcFormat) && isRGBA32(dstFormat)) {
1550 if ( CONV_IS(ABGR, RGBA)
1551 || CONV_IS(ARGB, BGRA)
1552 || CONV_IS(BGRA, ARGB)
1553 || CONV_IS(RGBA, ABGR)) conv = shuffle_bytes_3210;
1554 else if (CONV_IS(ABGR, ARGB)
1555 || CONV_IS(ARGB, ABGR)) conv = shuffle_bytes_0321;
1556 else if (CONV_IS(ABGR, BGRA)
1557 || CONV_IS(ARGB, RGBA)) conv = shuffle_bytes_1230;
1558 else if (CONV_IS(BGRA, RGBA)
1559 || CONV_IS(RGBA, BGRA)) conv = shuffle_bytes_2103;
1560 else if (CONV_IS(BGRA, ABGR)
1561 || CONV_IS(RGBA, ARGB)) conv = shuffle_bytes_3012;
1564 if ( (isBGRinInt(srcFormat) && isBGRinInt(dstFormat))
1565 || (isRGBinInt(srcFormat) && isRGBinInt(dstFormat))) {
1566 switch(srcId | (dstId<<4)) {
1567 case 0x34: conv= rgb16to15; break;
1568 case 0x36: conv= rgb24to15; break;
1569 case 0x38: conv= rgb32to15; break;
1570 case 0x43: conv= rgb15to16; break;
1571 case 0x46: conv= rgb24to16; break;
1572 case 0x48: conv= rgb32to16; break;
1573 case 0x63: conv= rgb15to24; break;
1574 case 0x64: conv= rgb16to24; break;
1575 case 0x68: conv= rgb32to24; break;
1576 case 0x83: conv= rgb15to32; break;
1577 case 0x84: conv= rgb16to32; break;
1578 case 0x86: conv= rgb24to32; break;
1580 } else if ( (isBGRinInt(srcFormat) && isRGBinInt(dstFormat))
1581 || (isRGBinInt(srcFormat) && isBGRinInt(dstFormat))) {
1582 switch(srcId | (dstId<<4)) {
1583 case 0x33: conv= rgb15tobgr15; break;
1584 case 0x34: conv= rgb16tobgr15; break;
1585 case 0x36: conv= rgb24tobgr15; break;
1586 case 0x38: conv= rgb32tobgr15; break;
1587 case 0x43: conv= rgb15tobgr16; break;
1588 case 0x44: conv= rgb16tobgr16; break;
1589 case 0x46: conv= rgb24tobgr16; break;
1590 case 0x48: conv= rgb32tobgr16; break;
1591 case 0x63: conv= rgb15tobgr24; break;
1592 case 0x64: conv= rgb16tobgr24; break;
1593 case 0x66: conv= rgb24tobgr24; break;
1594 case 0x68: conv= rgb32tobgr24; break;
1595 case 0x83: conv= rgb15tobgr32; break;
1596 case 0x84: conv= rgb16tobgr32; break;
1597 case 0x86: conv= rgb24tobgr32; break;
1602 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
1603 sws_format_name(srcFormat), sws_format_name(dstFormat));
1605 const uint8_t *srcPtr= src[0];
1606 uint8_t *dstPtr= dst[0];
1607 if ((srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1) && !isRGBA32(dstFormat))
1608 srcPtr += ALT32_CORR;
1610 if ((dstFormat == PIX_FMT_RGB32_1 || dstFormat == PIX_FMT_BGR32_1) && !isRGBA32(srcFormat))
1611 dstPtr += ALT32_CORR;
1613 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
1614 conv(srcPtr, dstPtr + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
1617 dstPtr += dstStride[0]*srcSliceY;
1619 for (i=0; i<srcSliceH; i++) {
1620 conv(srcPtr, dstPtr, c->srcW*srcBpp);
1621 srcPtr+= srcStride[0];
1622 dstPtr+= dstStride[0];
1629 static int bgr24ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1630 int srcSliceH, uint8_t* dst[], int dstStride[])
1634 dst[0]+ srcSliceY *dstStride[0],
1635 dst[1]+(srcSliceY>>1)*dstStride[1],
1636 dst[2]+(srcSliceY>>1)*dstStride[2],
1638 dstStride[0], dstStride[1], srcStride[0]);
1640 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1644 static int yvu9ToYv12Wrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1645 int srcSliceH, uint8_t* dst[], int dstStride[])
1647 copyPlane(src[0], srcStride[0], srcSliceY, srcSliceH, c->srcW,
1648 dst[0], dstStride[0]);
1650 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
1651 srcSliceH >> 2, srcStride[1], dstStride[1]);
1652 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
1653 srcSliceH >> 2, srcStride[2], dstStride[2]);
1655 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
1659 /* unscaled copy like stuff (assumes nearly identical formats) */
1660 static int packedCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1661 int srcSliceH, uint8_t* dst[], int dstStride[])
1663 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1664 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
1667 const uint8_t *srcPtr= src[0];
1668 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
1671 /* universal length finder */
1672 while(length+c->srcW <= FFABS(dstStride[0])
1673 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
1676 for (i=0; i<srcSliceH; i++) {
1677 memcpy(dstPtr, srcPtr, length);
1678 srcPtr+= srcStride[0];
1679 dstPtr+= dstStride[0];
1685 static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[], int srcSliceY,
1686 int srcSliceH, uint8_t* dst[], int dstStride[])
1689 for (plane=0; plane<4; plane++) {
1690 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
1691 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
1692 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
1693 const uint8_t *srcPtr= src[plane];
1694 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
1696 if (!dst[plane]) continue;
1697 // ignore palette for GRAY8
1698 if (plane == 1 && !dst[2]) continue;
1699 if (!src[plane] || (plane == 1 && !src[2])) {
1700 if(is16BPS(c->dstFormat))
1702 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
1704 if(is9_OR_10BPS(c->srcFormat)) {
1705 const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
1706 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1707 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1709 if (is16BPS(c->dstFormat)) {
1710 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1711 #define COPY9_OR_10TO16(rfunc, wfunc) \
1712 for (i = 0; i < height; i++) { \
1713 for (j = 0; j < length; j++) { \
1714 int srcpx = rfunc(&srcPtr2[j]); \
1715 wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \
1717 dstPtr2 += dstStride[plane]/2; \
1718 srcPtr2 += srcStride[plane]/2; \
1720 if (isBE(c->dstFormat)) {
1721 if (isBE(c->srcFormat)) {
1722 COPY9_OR_10TO16(AV_RB16, AV_WB16);
1724 COPY9_OR_10TO16(AV_RL16, AV_WB16);
1727 if (isBE(c->srcFormat)) {
1728 COPY9_OR_10TO16(AV_RB16, AV_WL16);
1730 COPY9_OR_10TO16(AV_RL16, AV_WL16);
1733 } else if (is9_OR_10BPS(c->dstFormat)) {
1734 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1735 #define COPY9_OR_10TO9_OR_10(loop) \
1736 for (i = 0; i < height; i++) { \
1737 for (j = 0; j < length; j++) { \
1740 dstPtr2 += dstStride[plane]/2; \
1741 srcPtr2 += srcStride[plane]/2; \
1743 #define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
1744 if (dst_depth > src_depth) { \
1745 COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
1746 wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
1747 } else if (dst_depth < src_depth) { \
1748 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \
1750 COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
1752 if (isBE(c->dstFormat)) {
1753 if (isBE(c->srcFormat)) {
1754 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
1756 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
1759 if (isBE(c->srcFormat)) {
1760 COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
1762 COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
1766 // FIXME Maybe dither instead.
1767 #define COPY9_OR_10TO8(rfunc) \
1768 for (i = 0; i < height; i++) { \
1769 for (j = 0; j < length; j++) { \
1770 dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \
1772 dstPtr += dstStride[plane]; \
1773 srcPtr2 += srcStride[plane]/2; \
1775 if (isBE(c->srcFormat)) {
1776 COPY9_OR_10TO8(AV_RB16);
1778 COPY9_OR_10TO8(AV_RL16);
1781 } else if(is9_OR_10BPS(c->dstFormat)) {
1782 const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
1783 uint16_t *dstPtr2 = (uint16_t*)dstPtr;
1785 if (is16BPS(c->srcFormat)) {
1786 const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
1787 #define COPY16TO9_OR_10(rfunc, wfunc) \
1788 for (i = 0; i < height; i++) { \
1789 for (j = 0; j < length; j++) { \
1790 wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \
1792 dstPtr2 += dstStride[plane]/2; \
1793 srcPtr2 += srcStride[plane]/2; \
1795 if (isBE(c->dstFormat)) {
1796 if (isBE(c->srcFormat)) {
1797 COPY16TO9_OR_10(AV_RB16, AV_WB16);
1799 COPY16TO9_OR_10(AV_RL16, AV_WB16);
1802 if (isBE(c->srcFormat)) {
1803 COPY16TO9_OR_10(AV_RB16, AV_WL16);
1805 COPY16TO9_OR_10(AV_RL16, AV_WL16);
1809 #define COPY8TO9_OR_10(wfunc) \
1810 for (i = 0; i < height; i++) { \
1811 for (j = 0; j < length; j++) { \
1812 const int srcpx = srcPtr[j]; \
1813 wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \
1815 dstPtr2 += dstStride[plane]/2; \
1816 srcPtr += srcStride[plane]; \
1818 if (isBE(c->dstFormat)) {
1819 COPY8TO9_OR_10(AV_WB16);
1821 COPY8TO9_OR_10(AV_WL16);
1824 } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
1825 if (!isBE(c->srcFormat)) srcPtr++;
1826 for (i=0; i<height; i++) {
1827 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
1828 srcPtr+= srcStride[plane];
1829 dstPtr+= dstStride[plane];
1831 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
1832 for (i=0; i<height; i++) {
1833 for (j=0; j<length; j++) {
1834 dstPtr[ j<<1 ] = srcPtr[j];
1835 dstPtr[(j<<1)+1] = srcPtr[j];
1837 srcPtr+= srcStride[plane];
1838 dstPtr+= dstStride[plane];
1840 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
1841 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
1843 for (i=0; i<height; i++) {
1844 for (j=0; j<length; j++)
1845 ((uint16_t*)dstPtr)[j] = av_bswap16(((const uint16_t*)srcPtr)[j]);
1846 srcPtr+= srcStride[plane];
1847 dstPtr+= dstStride[plane];
1849 } else if (dstStride[plane] == srcStride[plane] &&
1850 srcStride[plane] > 0 && srcStride[plane] == length) {
1851 memcpy(dst[plane] + dstStride[plane]*y, src[plane],
1852 height*dstStride[plane]);
1854 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
1856 for (i=0; i<height; i++) {
1857 memcpy(dstPtr, srcPtr, length);
1858 srcPtr+= srcStride[plane];
1859 dstPtr+= dstStride[plane];
1867 int ff_hardcodedcpuflags(void)
1870 #if COMPILE_TEMPLATE_MMX2
1871 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
1872 #elif COMPILE_TEMPLATE_AMD3DNOW
1873 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
1874 #elif COMPILE_TEMPLATE_MMX
1875 flags |= SWS_CPU_CAPS_MMX;
1876 #elif COMPILE_TEMPLATE_ALTIVEC
1877 flags |= SWS_CPU_CAPS_ALTIVEC;
1879 flags |= SWS_CPU_CAPS_BFIN;
1884 void ff_get_unscaled_swscale(SwsContext *c)
1886 const enum PixelFormat srcFormat = c->srcFormat;
1887 const enum PixelFormat dstFormat = c->dstFormat;
1888 const int flags = c->flags;
1889 const int dstH = c->dstH;
1892 needsDither= isAnyRGB(dstFormat)
1893 && c->dstFormatBpp < 24
1894 && (c->dstFormatBpp < c->srcFormatBpp || (!isAnyRGB(srcFormat)));
1897 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
1898 c->swScale= planarToNv12Wrapper;
1901 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && isAnyRGB(dstFormat)
1902 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
1903 c->swScale= ff_yuv2rgb_get_func_ptr(c);
1906 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
1907 c->swScale= yvu9ToYv12Wrapper;
1911 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
1912 c->swScale= bgr24ToYv12Wrapper;
1914 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
1915 if ( isAnyRGB(srcFormat)
1916 && isAnyRGB(dstFormat)
1917 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
1918 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
1919 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
1920 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
1921 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
1922 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
1923 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
1924 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
1925 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
1926 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
1927 && srcFormat != PIX_FMT_BGR48LE && dstFormat != PIX_FMT_BGR48LE
1928 && srcFormat != PIX_FMT_BGR48BE && dstFormat != PIX_FMT_BGR48BE
1929 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
1930 c->swScale= rgbToRgbWrapper;
1932 if ((usePal(srcFormat) && (
1933 dstFormat == PIX_FMT_RGB32 ||
1934 dstFormat == PIX_FMT_RGB32_1 ||
1935 dstFormat == PIX_FMT_RGB24 ||
1936 dstFormat == PIX_FMT_BGR32 ||
1937 dstFormat == PIX_FMT_BGR32_1 ||
1938 dstFormat == PIX_FMT_BGR24)))
1939 c->swScale= palToRgbWrapper;
1941 if (srcFormat == PIX_FMT_YUV422P) {
1942 if (dstFormat == PIX_FMT_YUYV422)
1943 c->swScale= yuv422pToYuy2Wrapper;
1944 else if (dstFormat == PIX_FMT_UYVY422)
1945 c->swScale= yuv422pToUyvyWrapper;
1948 /* LQ converters if -sws 0 or -sws 4*/
1949 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
1951 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
1952 if (dstFormat == PIX_FMT_YUYV422)
1953 c->swScale= planarToYuy2Wrapper;
1954 else if (dstFormat == PIX_FMT_UYVY422)
1955 c->swScale= planarToUyvyWrapper;
1958 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1959 c->swScale= yuyvToYuv420Wrapper;
1960 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
1961 c->swScale= uyvyToYuv420Wrapper;
1962 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
1963 c->swScale= yuyvToYuv422Wrapper;
1964 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
1965 c->swScale= uyvyToYuv422Wrapper;
1968 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
1969 !(c->flags & SWS_BITEXACT) &&
1970 srcFormat == PIX_FMT_YUV420P) {
1971 // unscaled YV12 -> packed YUV, we want speed
1972 if (dstFormat == PIX_FMT_YUYV422)
1973 c->swScale= yv12toyuy2_unscaled_altivec;
1974 else if (dstFormat == PIX_FMT_UYVY422)
1975 c->swScale= yv12touyvy_unscaled_altivec;
1980 if ( srcFormat == dstFormat
1981 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
1982 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
1983 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
1984 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
1985 || (isGray(dstFormat) && isGray(srcFormat))
1986 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
1987 && c->chrDstHSubSample == c->chrSrcHSubSample
1988 && c->chrDstVSubSample == c->chrSrcVSubSample
1989 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
1990 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
1992 if (isPacked(c->srcFormat))
1993 c->swScale= packedCopyWrapper;
1994 else /* Planar YUV or gray */
1995 c->swScale= planarCopyWrapper;
1998 if (flags & SWS_CPU_CAPS_BFIN)
1999 ff_bfin_get_unscaled_swscale (c);
2003 static void reset_ptr(const uint8_t* src[], int format)
2005 if(!isALPHA(format))
2007 if(!isPlanarYUV(format)) {
2010 if (!usePal(format))
2015 static int check_image_pointers(uint8_t *data[4], enum PixelFormat pix_fmt,
2016 const int linesizes[4])
2018 const AVPixFmtDescriptor *desc = &av_pix_fmt_descriptors[pix_fmt];
2021 for (i = 0; i < 4; i++) {
2022 int plane = desc->comp[i].plane;
2023 if (!data[plane] || !linesizes[plane])
2031 * swscale wrapper, so we don't need to export the SwsContext.
2032 * Assumes planar YUV to be in YUV order instead of YVU.
2034 int sws_scale(SwsContext *c, const uint8_t* const src[], const int srcStride[], int srcSliceY,
2035 int srcSliceH, uint8_t* const dst[], const int dstStride[])
2038 const uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
2039 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
2041 // do not mess up sliceDir if we have a "trailing" 0-size slice
2045 if (!check_image_pointers(src, c->srcFormat, srcStride)) {
2046 av_log(c, AV_LOG_ERROR, "bad src image pointers\n");
2049 if (!check_image_pointers(dst, c->dstFormat, dstStride)) {
2050 av_log(c, AV_LOG_ERROR, "bad dst image pointers\n");
2054 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
2055 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
2058 if (c->sliceDir == 0) {
2059 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
2062 if (usePal(c->srcFormat)) {
2063 for (i=0; i<256; i++) {
2064 int p, r, g, b,y,u,v;
2065 if(c->srcFormat == PIX_FMT_PAL8) {
2066 p=((const uint32_t*)(src[1]))[i];
2070 } else if(c->srcFormat == PIX_FMT_RGB8) {
2074 } else if(c->srcFormat == PIX_FMT_BGR8) {
2078 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
2082 } else if(c->srcFormat == PIX_FMT_GRAY8 || c->srcFormat == PIX_FMT_Y400A) {
2085 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
2090 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2091 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2092 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
2093 c->pal_yuv[i]= y + (u<<8) + (v<<16);
2095 switch(c->dstFormat) {
2100 c->pal_rgb[i]= r + (g<<8) + (b<<16);
2102 case PIX_FMT_BGR32_1:
2106 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
2108 case PIX_FMT_RGB32_1:
2112 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
2119 c->pal_rgb[i]= b + (g<<8) + (r<<16);
2124 // copy strides, so they can safely be modified
2125 if (c->sliceDir == 1) {
2126 // slices go from top to bottom
2127 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
2128 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
2130 reset_ptr(src2, c->srcFormat);
2131 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2133 /* reset slice direction at end of frame */
2134 if (srcSliceY + srcSliceH == c->srcH)
2137 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
2139 // slices go from bottom to top => we flip the image internally
2140 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
2141 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
2143 src2[0] += (srcSliceH-1)*srcStride[0];
2144 if (!usePal(c->srcFormat))
2145 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
2146 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
2147 src2[3] += (srcSliceH-1)*srcStride[3];
2148 dst2[0] += ( c->dstH -1)*dstStride[0];
2149 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
2150 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
2151 dst2[3] += ( c->dstH -1)*dstStride[3];
2153 reset_ptr(src2, c->srcFormat);
2154 reset_ptr((const uint8_t**)dst2, c->dstFormat);
2156 /* reset slice direction at end of frame */
2160 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
2164 /* Convert the palette to the same packed 32-bit format as the palette */
2165 void sws_convertPalette8ToPacked32(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2169 for (i=0; i<num_pixels; i++)
2170 ((uint32_t *) dst)[i] = ((const uint32_t *) palette)[src[i]];
2173 /* Palette format: ABCD -> dst format: ABC */
2174 void sws_convertPalette8ToPacked24(const uint8_t *src, uint8_t *dst, long num_pixels, const uint8_t *palette)
2178 for (i=0; i<num_pixels; i++) {
2180 dst[0]= palette[src[i]*4+0];
2181 dst[1]= palette[src[i]*4+1];
2182 dst[2]= palette[src[i]*4+2];