Add CHECKED_ALLOC macro.
[libav.git] / libswscale / swscale.c
CommitLineData
fe8054c0 1/*
d026b45e
DB
2 * Copyright (C) 2001-2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License
17 * along with FFmpeg; if not, write to the Free Software
b19bcbaa 18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
d026b45e 19 *
807e0c66
LA
20 * the C code (not assembly, mmx, ...) of this file can be used
21 * under the LGPL license too
d026b45e 22 */
783e9cc9 23
28bf81c9 24/*
9990e426 25 supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR32_1, BGR24, BGR16, BGR15, RGB32, RGB32_1, RGB24, Y8/Y800, YVU9/IF09, PAL8
caeaabe7 26 supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
e09d12f4 27 {BGR,RGB}{1,4,8,15,16} support dithering
6a4970ab 28
e09d12f4
MN
29 unscaled special converters (YV12=I420=IYUV, Y800=Y8)
30 YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
31 x -> x
32 YUV9 -> YV12
33 YUV9/YV12 -> Y800
34 Y800 -> YUV9/YV12
b935781b
MN
35 BGR24 -> BGR32 & RGB24 -> RGB32
36 BGR32 -> BGR24 & RGB32 -> RGB24
4bb3fa5e 37 BGR15 -> BGR16
b935781b
MN
38*/
39
6a4970ab 40/*
a6f6b237 41tested special converters (most are tested actually, but I did not write it down ...)
e09d12f4 42 YV12 -> BGR16
b935781b 43 YV12 -> YV12
4bb3fa5e 44 BGR15 -> BGR16
1e1c4fe9 45 BGR16 -> BGR16
e09d12f4 46 YVU9 -> YV12
b935781b
MN
47
48untested special converters
f40c7dbb 49 YV12/I420 -> BGR15/BGR24/BGR32 (it is the yuv2rgb stuff, so it should be OK)
1e1c4fe9
MN
50 YV12/I420 -> YV12/I420
51 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format
b935781b
MN
52 BGR24 -> BGR32 & RGB24 -> RGB32
53 BGR32 -> BGR24 & RGB32 -> RGB24
ec22603f 54 BGR24 -> YV12
28bf81c9
MN
55*/
56
d63a2cb1 57#define _SVID_SOURCE //needed for MAP_ANONYMOUS
d3f41512 58#include <inttypes.h>
dda87e9f 59#include <string.h>
077ea8a7 60#include <math.h>
c1b0bfb4 61#include <stdio.h>
b2d374c9 62#include "config.h"
81b7c056 63#include <assert.h>
b63f641e 64#if HAVE_SYS_MMAN_H
38d5c282 65#include <sys/mman.h>
113ef149
RD
66#if defined(MAP_ANON) && !defined(MAP_ANONYMOUS)
67#define MAP_ANONYMOUS MAP_ANON
68#endif
38d5c282 69#endif
dd35beb2
RP
70#if HAVE_VIRTUALALLOC
71#define WIN32_LEAN_AND_MEAN
72#include <windows.h>
73#endif
d604bab9 74#include "swscale.h"
5427e242 75#include "swscale_internal.h"
37079906 76#include "rgb2rgb.h"
52154148 77#include "libavutil/intreadwrite.h"
83da2c6f
DB
78#include "libavutil/x86_cpu.h"
79#include "libavutil/bswap.h"
0d9f3d85 80
b3e03fa7
SS
81unsigned swscale_version(void)
82{
83 return LIBSWSCALE_VERSION_INT;
84}
85
541c4eb9 86#undef MOVNTQ
7d7f78b5 87#undef PAVGB
d3f41512 88
783e9cc9 89//#undef HAVE_MMX2
f4406ec1 90//#define HAVE_AMD3DNOW
d3f41512 91//#undef HAVE_MMX
783e9cc9 92//#undef ARCH_X86
d604bab9 93#define DITHER1XBPP
d3f41512 94
f40c7dbb 95#define FAST_BGR2YV12 // use 7 bit coefficients instead of 15 bit
ac6a2e45 96
f40c7dbb 97#define RET 0xC3 //near return opcode for x86
c1b0bfb4 98
28bf81c9
MN
99#ifdef M_PI
100#define PI M_PI
101#else
102#define PI 3.14159265358979323846
103#endif
c1b0bfb4 104
9d9de37d
IP
105#define isSupportedIn(x) ( \
106 (x)==PIX_FMT_YUV420P \
79973335 107 || (x)==PIX_FMT_YUVA420P \
9d9de37d
IP
108 || (x)==PIX_FMT_YUYV422 \
109 || (x)==PIX_FMT_UYVY422 \
e8417235
KS
110 || (x)==PIX_FMT_RGB48BE \
111 || (x)==PIX_FMT_RGB48LE \
9d9de37d 112 || (x)==PIX_FMT_RGB32 \
9990e426 113 || (x)==PIX_FMT_RGB32_1 \
9d9de37d
IP
114 || (x)==PIX_FMT_BGR24 \
115 || (x)==PIX_FMT_BGR565 \
116 || (x)==PIX_FMT_BGR555 \
117 || (x)==PIX_FMT_BGR32 \
9990e426 118 || (x)==PIX_FMT_BGR32_1 \
9d9de37d
IP
119 || (x)==PIX_FMT_RGB24 \
120 || (x)==PIX_FMT_RGB565 \
121 || (x)==PIX_FMT_RGB555 \
122 || (x)==PIX_FMT_GRAY8 \
123 || (x)==PIX_FMT_YUV410P \
6c80eb16 124 || (x)==PIX_FMT_YUV440P \
9d9de37d
IP
125 || (x)==PIX_FMT_GRAY16BE \
126 || (x)==PIX_FMT_GRAY16LE \
127 || (x)==PIX_FMT_YUV444P \
128 || (x)==PIX_FMT_YUV422P \
129 || (x)==PIX_FMT_YUV411P \
130 || (x)==PIX_FMT_PAL8 \
131 || (x)==PIX_FMT_BGR8 \
132 || (x)==PIX_FMT_RGB8 \
133 || (x)==PIX_FMT_BGR4_BYTE \
134 || (x)==PIX_FMT_RGB4_BYTE \
9ba7fe6d 135 || (x)==PIX_FMT_YUV440P \
3d05e078
MN
136 || (x)==PIX_FMT_MONOWHITE \
137 || (x)==PIX_FMT_MONOBLACK \
de1275d5
MN
138 || (x)==PIX_FMT_YUV420PLE \
139 || (x)==PIX_FMT_YUV422PLE \
140 || (x)==PIX_FMT_YUV444PLE \
141 || (x)==PIX_FMT_YUV420PBE \
142 || (x)==PIX_FMT_YUV422PBE \
143 || (x)==PIX_FMT_YUV444PBE \
9d9de37d
IP
144 )
145#define isSupportedOut(x) ( \
146 (x)==PIX_FMT_YUV420P \
6268f55b 147 || (x)==PIX_FMT_YUVA420P \
9d9de37d
IP
148 || (x)==PIX_FMT_YUYV422 \
149 || (x)==PIX_FMT_UYVY422 \
150 || (x)==PIX_FMT_YUV444P \
151 || (x)==PIX_FMT_YUV422P \
152 || (x)==PIX_FMT_YUV411P \
153 || isRGB(x) \
154 || isBGR(x) \
155 || (x)==PIX_FMT_NV12 \
156 || (x)==PIX_FMT_NV21 \
157 || (x)==PIX_FMT_GRAY16BE \
158 || (x)==PIX_FMT_GRAY16LE \
159 || (x)==PIX_FMT_GRAY8 \
160 || (x)==PIX_FMT_YUV410P \
6c80eb16 161 || (x)==PIX_FMT_YUV440P \
de1275d5
MN
162 || (x)==PIX_FMT_YUV420PLE \
163 || (x)==PIX_FMT_YUV422PLE \
164 || (x)==PIX_FMT_YUV444PLE \
165 || (x)==PIX_FMT_YUV420PBE \
166 || (x)==PIX_FMT_YUV422PBE \
167 || (x)==PIX_FMT_YUV444PBE \
9d9de37d
IP
168 )
169#define isPacked(x) ( \
170 (x)==PIX_FMT_PAL8 \
171 || (x)==PIX_FMT_YUYV422 \
172 || (x)==PIX_FMT_UYVY422 \
173 || isRGB(x) \
174 || isBGR(x) \
175 )
49004617
VS
176#define usePal(x) ( \
177 (x)==PIX_FMT_PAL8 \
178 || (x)==PIX_FMT_BGR4_BYTE \
179 || (x)==PIX_FMT_RGB4_BYTE \
180 || (x)==PIX_FMT_BGR8 \
181 || (x)==PIX_FMT_RGB8 \
182 )
6ff0ad6b 183
6b79dbce 184#define RGB2YUV_SHIFT 15
7b5d7b9e
MN
185#define BY ( (int)(0.114*219/255*(1<<RGB2YUV_SHIFT)+0.5))
186#define BV (-(int)(0.081*224/255*(1<<RGB2YUV_SHIFT)+0.5))
187#define BU ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
188#define GY ( (int)(0.587*219/255*(1<<RGB2YUV_SHIFT)+0.5))
189#define GV (-(int)(0.419*224/255*(1<<RGB2YUV_SHIFT)+0.5))
190#define GU (-(int)(0.331*224/255*(1<<RGB2YUV_SHIFT)+0.5))
191#define RY ( (int)(0.299*219/255*(1<<RGB2YUV_SHIFT)+0.5))
192#define RV ( (int)(0.500*224/255*(1<<RGB2YUV_SHIFT)+0.5))
193#define RU (-(int)(0.169*224/255*(1<<RGB2YUV_SHIFT)+0.5))
6c7506de 194
fa58ba15 195extern const int32_t ff_yuv2rgb_coeffs[8][4];
0481412a 196
0f5d4aa8
MN
197static const double rgb2yuv_table[8][9]={
198 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
199 {0.7152, 0.0722, 0.2126, -0.386, 0.5, -0.115, -0.454, -0.046, 0.5},
200 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
201 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
202 {0.59 , 0.11 , 0.30 , -0.331, 0.5, -0.169, -0.421, -0.079, 0.5}, //FCC
203 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5},
204 {0.587 , 0.114 , 0.299 , -0.331, 0.5, -0.169, -0.419, -0.081, 0.5}, //SMPTE 170M
205 {0.701 , 0.087 , 0.212 , -0.384, 0.5 -0.116, -0.445, -0.055, 0.5}, //SMPTE 240M
206};
207
783e9cc9
MN
208/*
209NOTES
d604bab9 210Special versions: fast Y 1:1 scaling (no interpolation in y direction)
31190492 211
783e9cc9 212TODO
bd7c6fd5 213more intelligent misalignment avoidance for the horizontal scaler
c1b0bfb4 214write special vertical cubic upscale version
f40c7dbb
DB
215optimize C code (YV12 / minmax)
216add support for packed pixel YUV input & output
6ff0ad6b 217add support for Y8 output
f40c7dbb 218optimize BGR24 & BGR32
ff7ba856 219add BGR4 output support
1e621b18 220write special BGR->BGR scaler
783e9cc9 221*/
31190492 222
b63f641e 223#if ARCH_X86 && CONFIG_GPL
d334c7c2
RD
224DECLARE_ASM_CONST(8, uint64_t, bF8)= 0xF8F8F8F8F8F8F8F8LL;
225DECLARE_ASM_CONST(8, uint64_t, bFC)= 0xFCFCFCFCFCFCFCFCLL;
226DECLARE_ASM_CONST(8, uint64_t, w10)= 0x0010001000100010LL;
227DECLARE_ASM_CONST(8, uint64_t, w02)= 0x0002000200020002LL;
228DECLARE_ASM_CONST(8, uint64_t, bm00001111)=0x00000000FFFFFFFFLL;
229DECLARE_ASM_CONST(8, uint64_t, bm00000111)=0x0000000000FFFFFFLL;
230DECLARE_ASM_CONST(8, uint64_t, bm11111000)=0xFFFFFFFFFF000000LL;
231DECLARE_ASM_CONST(8, uint64_t, bm01010101)=0x00FF00FF00FF00FFLL;
d604bab9 232
0cb25594 233const DECLARE_ALIGNED(8, uint64_t, ff_dither4[2]) = {
221b804f
DB
234 0x0103010301030103LL,
235 0x0200020002000200LL,};
d8fa3c54 236
0cb25594 237const DECLARE_ALIGNED(8, uint64_t, ff_dither8[2]) = {
221b804f
DB
238 0x0602060206020602LL,
239 0x0004000400040004LL,};
d604bab9 240
d334c7c2
RD
241DECLARE_ASM_CONST(8, uint64_t, b16Mask)= 0x001F001F001F001FLL;
242DECLARE_ASM_CONST(8, uint64_t, g16Mask)= 0x07E007E007E007E0LL;
243DECLARE_ASM_CONST(8, uint64_t, r16Mask)= 0xF800F800F800F800LL;
244DECLARE_ASM_CONST(8, uint64_t, b15Mask)= 0x001F001F001F001FLL;
245DECLARE_ASM_CONST(8, uint64_t, g15Mask)= 0x03E003E003E003E0LL;
246DECLARE_ASM_CONST(8, uint64_t, r15Mask)= 0x7C007C007C007C00LL;
d604bab9 247
5802683a
RD
248DECLARE_ALIGNED(8, const uint64_t, ff_M24A) = 0x00FF0000FF0000FFLL;
249DECLARE_ALIGNED(8, const uint64_t, ff_M24B) = 0xFF0000FF0000FF00LL;
250DECLARE_ALIGNED(8, const uint64_t, ff_M24C) = 0x0000FF0000FF0000LL;
99d2cb72 251
ac6a2e45 252#ifdef FAST_BGR2YV12
5802683a
RD
253DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000000210041000DULL;
254DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000FFEEFFDC0038ULL;
255DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00000038FFD2FFF8ULL;
ac6a2e45 256#else
5802683a
RD
257DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YCoeff) = 0x000020E540830C8BULL;
258DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UCoeff) = 0x0000ED0FDAC23831ULL;
259DECLARE_ALIGNED(8, const uint64_t, ff_bgr2VCoeff) = 0x00003831D0E6F6EAULL;
69796008 260#endif /* FAST_BGR2YV12 */
5802683a
RD
261DECLARE_ALIGNED(8, const uint64_t, ff_bgr2YOffset) = 0x1010101010101010ULL;
262DECLARE_ALIGNED(8, const uint64_t, ff_bgr2UVOffset) = 0x8080808080808080ULL;
263DECLARE_ALIGNED(8, const uint64_t, ff_w1111) = 0x0001000100010001ULL;
dfb09bd1 264
b5c44b1a
DP
265DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY1Coeff) = 0x0C88000040870C88ULL;
266DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toY2Coeff) = 0x20DE4087000020DEULL;
267DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY1Coeff) = 0x20DE0000408720DEULL;
268DECLARE_ASM_CONST(8, uint64_t, ff_rgb24toY2Coeff) = 0x0C88408700000C88ULL;
269DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toYOffset) = 0x0008400000084000ULL;
dfb09bd1 270
b5c44b1a 271DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUV[2][4]) = {
dfb09bd1
MN
272 {0x38380000DAC83838ULL, 0xECFFDAC80000ECFFULL, 0xF6E40000D0E3F6E4ULL, 0x3838D0E300003838ULL},
273 {0xECFF0000DAC8ECFFULL, 0x3838DAC800003838ULL, 0x38380000D0E33838ULL, 0xF6E4D0E30000F6E4ULL},
274};
275
b5c44b1a 276DECLARE_ASM_CONST(8, uint64_t, ff_bgr24toUVOffset)= 0x0040400000404000ULL;
dfb09bd1 277
7a24ec50 278#endif /* ARCH_X86 && CONFIG_GPL */
783e9cc9
MN
279
280// clipping helper table for C implementations:
281static unsigned char clip_table[768];
282
d4e24275 283static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
6a4970ab 284
92db6235 285DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_4[2][8])={
45e18be8
MN
286{ 1, 3, 1, 3, 1, 3, 1, 3, },
287{ 2, 0, 2, 0, 2, 0, 2, 0, },
288};
289
92db6235 290DECLARE_ALIGNED(8, static const uint8_t, dither_2x2_8[2][8])={
45e18be8
MN
291{ 6, 2, 6, 2, 6, 2, 6, 2, },
292{ 0, 4, 0, 4, 0, 4, 0, 4, },
293};
294
92db6235 295DECLARE_ALIGNED(8, const uint8_t, dither_8x8_32[8][8])={
45e18be8
MN
296{ 17, 9, 23, 15, 16, 8, 22, 14, },
297{ 5, 29, 3, 27, 4, 28, 2, 26, },
298{ 21, 13, 19, 11, 20, 12, 18, 10, },
299{ 0, 24, 6, 30, 1, 25, 7, 31, },
300{ 16, 8, 22, 14, 17, 9, 23, 15, },
301{ 4, 28, 2, 26, 5, 29, 3, 27, },
302{ 20, 12, 18, 10, 21, 13, 19, 11, },
303{ 1, 25, 7, 31, 0, 24, 6, 30, },
304};
305
92db6235 306DECLARE_ALIGNED(8, const uint8_t, dither_8x8_73[8][8])={
45e18be8
MN
307{ 0, 55, 14, 68, 3, 58, 17, 72, },
308{ 37, 18, 50, 32, 40, 22, 54, 35, },
309{ 9, 64, 5, 59, 13, 67, 8, 63, },
310{ 46, 27, 41, 23, 49, 31, 44, 26, },
311{ 2, 57, 16, 71, 1, 56, 15, 70, },
312{ 39, 21, 52, 34, 38, 19, 51, 33, },
313{ 11, 66, 7, 62, 10, 65, 6, 60, },
314{ 48, 30, 43, 25, 47, 29, 42, 24, },
315};
316
45e18be8 317#if 1
92db6235 318DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
45e18be8
MN
319{117, 62, 158, 103, 113, 58, 155, 100, },
320{ 34, 199, 21, 186, 31, 196, 17, 182, },
321{144, 89, 131, 76, 141, 86, 127, 72, },
322{ 0, 165, 41, 206, 10, 175, 52, 217, },
323{110, 55, 151, 96, 120, 65, 162, 107, },
324{ 28, 193, 14, 179, 38, 203, 24, 189, },
325{138, 83, 124, 69, 148, 93, 134, 79, },
326{ 7, 172, 48, 213, 3, 168, 45, 210, },
327};
328#elif 1
329// tries to correct a gamma of 1.5
92db6235 330DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
45e18be8
MN
331{ 0, 143, 18, 200, 2, 156, 25, 215, },
332{ 78, 28, 125, 64, 89, 36, 138, 74, },
333{ 10, 180, 3, 161, 16, 195, 8, 175, },
334{109, 51, 93, 38, 121, 60, 105, 47, },
335{ 1, 152, 23, 210, 0, 147, 20, 205, },
336{ 85, 33, 134, 71, 81, 30, 130, 67, },
337{ 14, 190, 6, 171, 12, 185, 5, 166, },
338{117, 57, 101, 44, 113, 54, 97, 41, },
339};
340#elif 1
341// tries to correct a gamma of 2.0
92db6235 342DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
45e18be8
MN
343{ 0, 124, 8, 193, 0, 140, 12, 213, },
344{ 55, 14, 104, 42, 66, 19, 119, 52, },
345{ 3, 168, 1, 145, 6, 187, 3, 162, },
346{ 86, 31, 70, 21, 99, 39, 82, 28, },
347{ 0, 134, 11, 206, 0, 129, 9, 200, },
348{ 62, 17, 114, 48, 58, 16, 109, 45, },
349{ 5, 181, 2, 157, 4, 175, 1, 151, },
350{ 95, 36, 78, 26, 90, 34, 74, 24, },
351};
352#else
353// tries to correct a gamma of 2.5
92db6235 354DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220[8][8])={
45e18be8
MN
355{ 0, 107, 3, 187, 0, 125, 6, 212, },
356{ 39, 7, 86, 28, 49, 11, 102, 36, },
357{ 1, 158, 0, 131, 3, 180, 1, 151, },
358{ 68, 19, 52, 12, 81, 25, 64, 17, },
359{ 0, 119, 5, 203, 0, 113, 4, 195, },
360{ 45, 9, 96, 33, 42, 8, 91, 30, },
361{ 2, 172, 1, 144, 2, 165, 0, 137, },
362{ 77, 23, 60, 15, 72, 21, 56, 14, },
363};
364#endif
5cebb24b 365
8055ede6 366const char *sws_format_name(enum PixelFormat format)
94c4def2 367{
e9e12f0e 368 switch (format) {
9b734d44
RP
369 case PIX_FMT_YUV420P:
370 return "yuv420p";
371 case PIX_FMT_YUVA420P:
372 return "yuva420p";
373 case PIX_FMT_YUYV422:
374 return "yuyv422";
375 case PIX_FMT_RGB24:
376 return "rgb24";
377 case PIX_FMT_BGR24:
378 return "bgr24";
379 case PIX_FMT_YUV422P:
380 return "yuv422p";
381 case PIX_FMT_YUV444P:
382 return "yuv444p";
383 case PIX_FMT_RGB32:
384 return "rgb32";
385 case PIX_FMT_YUV410P:
386 return "yuv410p";
387 case PIX_FMT_YUV411P:
388 return "yuv411p";
389 case PIX_FMT_RGB565:
390 return "rgb565";
391 case PIX_FMT_RGB555:
392 return "rgb555";
393 case PIX_FMT_GRAY16BE:
394 return "gray16be";
395 case PIX_FMT_GRAY16LE:
396 return "gray16le";
397 case PIX_FMT_GRAY8:
398 return "gray8";
399 case PIX_FMT_MONOWHITE:
400 return "mono white";
401 case PIX_FMT_MONOBLACK:
402 return "mono black";
403 case PIX_FMT_PAL8:
404 return "Palette";
405 case PIX_FMT_YUVJ420P:
406 return "yuvj420p";
407 case PIX_FMT_YUVJ422P:
408 return "yuvj422p";
409 case PIX_FMT_YUVJ444P:
410 return "yuvj444p";
411 case PIX_FMT_XVMC_MPEG2_MC:
412 return "xvmc_mpeg2_mc";
413 case PIX_FMT_XVMC_MPEG2_IDCT:
414 return "xvmc_mpeg2_idct";
415 case PIX_FMT_UYVY422:
416 return "uyvy422";
417 case PIX_FMT_UYYVYY411:
418 return "uyyvyy411";
419 case PIX_FMT_RGB32_1:
420 return "rgb32x";
421 case PIX_FMT_BGR32_1:
422 return "bgr32x";
423 case PIX_FMT_BGR32:
424 return "bgr32";
425 case PIX_FMT_BGR565:
426 return "bgr565";
427 case PIX_FMT_BGR555:
428 return "bgr555";
429 case PIX_FMT_BGR8:
430 return "bgr8";
431 case PIX_FMT_BGR4:
432 return "bgr4";
433 case PIX_FMT_BGR4_BYTE:
434 return "bgr4 byte";
435 case PIX_FMT_RGB8:
436 return "rgb8";
437 case PIX_FMT_RGB4:
438 return "rgb4";
439 case PIX_FMT_RGB4_BYTE:
440 return "rgb4 byte";
441 case PIX_FMT_RGB48BE:
442 return "rgb48be";
443 case PIX_FMT_RGB48LE:
444 return "rgb48le";
445 case PIX_FMT_NV12:
446 return "nv12";
447 case PIX_FMT_NV21:
448 return "nv21";
449 case PIX_FMT_YUV440P:
450 return "yuv440p";
451 case PIX_FMT_VDPAU_H264:
452 return "vdpau_h264";
453 case PIX_FMT_VDPAU_MPEG1:
454 return "vdpau_mpeg1";
455 case PIX_FMT_VDPAU_MPEG2:
456 return "vdpau_mpeg2";
457 case PIX_FMT_VDPAU_WMV3:
458 return "vdpau_wmv3";
459 case PIX_FMT_VDPAU_VC1:
460 return "vdpau_vc1";
461 case PIX_FMT_YUV420PLE:
462 return "yuv420ple";
463 case PIX_FMT_YUV422PLE:
464 return "yuv422ple";
465 case PIX_FMT_YUV444PLE:
466 return "yuv444ple";
467 case PIX_FMT_YUV420PBE:
468 return "yuv420pbe";
469 case PIX_FMT_YUV422PBE:
470 return "yuv422pbe";
471 case PIX_FMT_YUV444PBE:
472 return "yuv444pbe";
473 default:
474 return "Unknown format";
e9e12f0e 475 }
94c4def2
LA
476}
477
52154148
RP
478static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
479 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
480 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
481 int dstW, int chrDstW, int big_endian)
482{
483 //FIXME Optimize (just quickly written not optimized..)
484 int i;
485
486 for (i = 0; i < dstW; i++) {
487 int val = 1 << 10;
488 int j;
489
490 for (j = 0; j < lumFilterSize; j++)
491 val += lumSrc[j][i] * lumFilter[j];
492
493 if (big_endian) {
494 AV_WB16(&dest[i], av_clip_uint16(val >> 11));
495 } else {
496 AV_WL16(&dest[i], av_clip_uint16(val >> 11));
497 }
498 }
499
500 if (uDest) {
501 for (i = 0; i < chrDstW; i++) {
502 int u = 1 << 10;
503 int v = 1 << 10;
504 int j;
505
506 for (j = 0; j < chrFilterSize; j++) {
507 u += chrSrc[j][i ] * chrFilter[j];
508 v += chrSrc[j][i + VOFW] * chrFilter[j];
509 }
510
511 if (big_endian) {
512 AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
513 AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
514 } else {
515 AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
516 AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
517 }
518 }
519 }
520
521 if (CONFIG_SWSCALE_ALPHA && aDest) {
522 for (i = 0; i < dstW; i++) {
523 int val = 1 << 10;
524 int j;
525
526 for (j = 0; j < lumFilterSize; j++)
527 val += alpSrc[j][i] * lumFilter[j];
528
529 if (big_endian) {
530 AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
531 } else {
532 AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
533 }
534 }
535 }
536}
537
538static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
539 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
540 const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
541 enum PixelFormat dstFormat)
542{
543 if (isBE(dstFormat)) {
544 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
545 chrFilter, chrSrc, chrFilterSize,
546 alpSrc,
547 dest, uDest, vDest, aDest,
548 dstW, chrDstW, 1);
549 } else {
550 yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
551 chrFilter, chrSrc, chrFilterSize,
552 alpSrc,
553 dest, uDest, vDest, aDest,
554 dstW, chrDstW, 0);
555 }
556}
557
7ac40327
RP
558static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
559 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
560 const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW)
e3d2500f 561{
f40c7dbb 562 //FIXME Optimize (just quickly written not optimized..)
221b804f 563 int i;
dd68318c 564 for (i=0; i<dstW; i++) {
221b804f
DB
565 int val=1<<18;
566 int j;
567 for (j=0; j<lumFilterSize; j++)
568 val += lumSrc[j][i] * lumFilter[j];
569
570 dest[i]= av_clip_uint8(val>>19);
571 }
572
1b0a4572 573 if (uDest)
dd68318c 574 for (i=0; i<chrDstW; i++) {
221b804f
DB
575 int u=1<<18;
576 int v=1<<18;
577 int j;
dd68318c 578 for (j=0; j<chrFilterSize; j++) {
221b804f 579 u += chrSrc[j][i] * chrFilter[j];
8b2fce0d 580 v += chrSrc[j][i + VOFW] * chrFilter[j];
221b804f
DB
581 }
582
583 uDest[i]= av_clip_uint8(u>>19);
584 vDest[i]= av_clip_uint8(v>>19);
585 }
6858492e
CS
586
587 if (CONFIG_SWSCALE_ALPHA && aDest)
dd68318c 588 for (i=0; i<dstW; i++) {
6858492e
CS
589 int val=1<<18;
590 int j;
591 for (j=0; j<lumFilterSize; j++)
592 val += alpSrc[j][i] * lumFilter[j];
593
594 aDest[i]= av_clip_uint8(val>>19);
595 }
596
e3d2500f
MN
597}
598
7ac40327
RP
599static inline void yuv2nv12XinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
600 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
221b804f 601 uint8_t *dest, uint8_t *uDest, int dstW, int chrDstW, int dstFormat)
6118e52e 602{
f40c7dbb 603 //FIXME Optimize (just quickly written not optimized..)
221b804f 604 int i;
dd68318c 605 for (i=0; i<dstW; i++) {
221b804f
DB
606 int val=1<<18;
607 int j;
608 for (j=0; j<lumFilterSize; j++)
609 val += lumSrc[j][i] * lumFilter[j];
610
611 dest[i]= av_clip_uint8(val>>19);
612 }
613
1b0a4572 614 if (!uDest)
221b804f
DB
615 return;
616
617 if (dstFormat == PIX_FMT_NV12)
dd68318c 618 for (i=0; i<chrDstW; i++) {
221b804f
DB
619 int u=1<<18;
620 int v=1<<18;
621 int j;
dd68318c 622 for (j=0; j<chrFilterSize; j++) {
221b804f 623 u += chrSrc[j][i] * chrFilter[j];
8b2fce0d 624 v += chrSrc[j][i + VOFW] * chrFilter[j];
221b804f
DB
625 }
626
627 uDest[2*i]= av_clip_uint8(u>>19);
628 uDest[2*i+1]= av_clip_uint8(v>>19);
629 }
630 else
dd68318c 631 for (i=0; i<chrDstW; i++) {
221b804f
DB
632 int u=1<<18;
633 int v=1<<18;
634 int j;
dd68318c 635 for (j=0; j<chrFilterSize; j++) {
221b804f 636 u += chrSrc[j][i] * chrFilter[j];
8b2fce0d 637 v += chrSrc[j][i + VOFW] * chrFilter[j];
221b804f
DB
638 }
639
640 uDest[2*i]= av_clip_uint8(v>>19);
641 uDest[2*i+1]= av_clip_uint8(u>>19);
642 }
6118e52e 643}
46de8b73 644
6858492e 645#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
dd68318c 646 for (i=0; i<(dstW>>1); i++) {\
221b804f
DB
647 int j;\
648 int Y1 = 1<<18;\
649 int Y2 = 1<<18;\
650 int U = 1<<18;\
651 int V = 1<<18;\
6858492e 652 int av_unused A1, A2;\
2db27aad 653 type av_unused *r, *b, *g;\
221b804f
DB
654 const int i2= 2*i;\
655 \
dd68318c 656 for (j=0; j<lumFilterSize; j++) {\
221b804f
DB
657 Y1 += lumSrc[j][i2] * lumFilter[j];\
658 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
659 }\
dd68318c 660 for (j=0; j<chrFilterSize; j++) {\
221b804f 661 U += chrSrc[j][i] * chrFilter[j];\
8b2fce0d 662 V += chrSrc[j][i+VOFW] * chrFilter[j];\
221b804f
DB
663 }\
664 Y1>>=19;\
665 Y2>>=19;\
666 U >>=19;\
667 V >>=19;\
dd68318c 668 if (alpha) {\
6858492e
CS
669 A1 = 1<<18;\
670 A2 = 1<<18;\
dd68318c 671 for (j=0; j<lumFilterSize; j++) {\
6858492e
CS
672 A1 += alpSrc[j][i2 ] * lumFilter[j];\
673 A2 += alpSrc[j][i2+1] * lumFilter[j];\
674 }\
675 A1>>=19;\
676 A2>>=19;\
677 }\
bdf397ba 678
6858492e
CS
679#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
680 YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
dd68318c 681 if ((Y1|Y2|U|V)&256) {\
221b804f
DB
682 if (Y1>255) Y1=255; \
683 else if (Y1<0)Y1=0; \
684 if (Y2>255) Y2=255; \
685 else if (Y2<0)Y2=0; \
686 if (U>255) U=255; \
687 else if (U<0) U=0; \
688 if (V>255) V=255; \
689 else if (V<0) V=0; \
6858492e 690 }\
dd68318c 691 if (alpha && ((A1|A2)&256)) {\
6858492e
CS
692 A1=av_clip_uint8(A1);\
693 A2=av_clip_uint8(A2);\
221b804f 694 }
6a4970ab 695
6858492e 696#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
dd68318c 697 for (i=0; i<dstW; i++) {\
f0faee4c
MN
698 int j;\
699 int Y = 0;\
700 int U = -128<<19;\
701 int V = -128<<19;\
6858492e 702 int av_unused A;\
f0faee4c
MN
703 int R,G,B;\
704 \
dd68318c 705 for (j=0; j<lumFilterSize; j++) {\
f0faee4c
MN
706 Y += lumSrc[j][i ] * lumFilter[j];\
707 }\
dd68318c 708 for (j=0; j<chrFilterSize; j++) {\
f0faee4c
MN
709 U += chrSrc[j][i ] * chrFilter[j];\
710 V += chrSrc[j][i+VOFW] * chrFilter[j];\
711 }\
712 Y >>=10;\
713 U >>=10;\
714 V >>=10;\
dd68318c 715 if (alpha) {\
6858492e
CS
716 A = rnd;\
717 for (j=0; j<lumFilterSize; j++)\
718 A += alpSrc[j][i ] * lumFilter[j];\
719 A >>=19;\
720 if (A&256)\
721 A = av_clip_uint8(A);\
722 }\
f0faee4c 723
6858492e
CS
724#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
725 YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
43c16478
MN
726 Y-= c->yuv2rgb_y_offset;\
727 Y*= c->yuv2rgb_y_coeff;\
f0faee4c 728 Y+= rnd;\
43c16478
MN
729 R= Y + V*c->yuv2rgb_v2r_coeff;\
730 G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
731 B= Y + U*c->yuv2rgb_u2b_coeff;\
dd68318c 732 if ((R|G|B)&(0xC0000000)) {\
f0faee4c
MN
733 if (R>=(256<<22)) R=(256<<22)-1; \
734 else if (R<0)R=0; \
735 if (G>=(256<<22)) G=(256<<22)-1; \
736 else if (G<0)G=0; \
737 if (B>=(256<<22)) B=(256<<22)-1; \
738 else if (B<0)B=0; \
739 }\
740
741
e69bd294 742#define YSCALE_YUV_2_GRAY16_C \
dd68318c 743 for (i=0; i<(dstW>>1); i++) {\
b0880d5d
MN
744 int j;\
745 int Y1 = 1<<18;\
746 int Y2 = 1<<18;\
747 int U = 1<<18;\
748 int V = 1<<18;\
e69bd294 749 \
b0880d5d
MN
750 const int i2= 2*i;\
751 \
dd68318c 752 for (j=0; j<lumFilterSize; j++) {\
b0880d5d
MN
753 Y1 += lumSrc[j][i2] * lumFilter[j];\
754 Y2 += lumSrc[j][i2+1] * lumFilter[j];\
755 }\
756 Y1>>=11;\
757 Y2>>=11;\
dd68318c 758 if ((Y1|Y2|U|V)&65536) {\
b0880d5d
MN
759 if (Y1>65535) Y1=65535; \
760 else if (Y1<0)Y1=0; \
761 if (Y2>65535) Y2=65535; \
762 else if (Y2<0)Y2=0; \
763 }
764
6858492e
CS
765#define YSCALE_YUV_2_RGBX_C(type,alpha) \
766 YSCALE_YUV_2_PACKEDX_C(type,alpha) /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
221b804f
DB
767 r = (type *)c->table_rV[V]; \
768 g = (type *)(c->table_gU[U] + c->table_gV[V]); \
769 b = (type *)c->table_bU[U]; \
770
6858492e 771#define YSCALE_YUV_2_PACKED2_C(type,alpha) \
dd68318c 772 for (i=0; i<(dstW>>1); i++) { \
221b804f
DB
773 const int i2= 2*i; \
774 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19; \
775 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19; \
776 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19; \
8b2fce0d 777 int V= (uvbuf0[i+VOFW]*uvalpha1+uvbuf1[i+VOFW]*uvalpha)>>19; \
6858492e
CS
778 type av_unused *r, *b, *g; \
779 int av_unused A1, A2; \
dd68318c 780 if (alpha) {\
6858492e
CS
781 A1= (abuf0[i2 ]*yalpha1+abuf1[i2 ]*yalpha)>>19; \
782 A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19; \
783 }\
46de8b73 784
b0880d5d 785#define YSCALE_YUV_2_GRAY16_2_C \
dd68318c 786 for (i=0; i<(dstW>>1); i++) { \
b0880d5d
MN
787 const int i2= 2*i; \
788 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>11; \
789 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11; \
790
6858492e
CS
791#define YSCALE_YUV_2_RGB2_C(type,alpha) \
792 YSCALE_YUV_2_PACKED2_C(type,alpha)\
221b804f
DB
793 r = (type *)c->table_rV[V];\
794 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
795 b = (type *)c->table_bU[U];\
cf7d1c1a 796
6858492e 797#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
dd68318c 798 for (i=0; i<(dstW>>1); i++) {\
221b804f
DB
799 const int i2= 2*i;\
800 int Y1= buf0[i2 ]>>7;\
801 int Y2= buf0[i2+1]>>7;\
802 int U= (uvbuf1[i ])>>7;\
8b2fce0d 803 int V= (uvbuf1[i+VOFW])>>7;\
6858492e
CS
804 type av_unused *r, *b, *g;\
805 int av_unused A1, A2;\
dd68318c 806 if (alpha) {\
6858492e
CS
807 A1= abuf0[i2 ]>>7;\
808 A2= abuf0[i2+1]>>7;\
809 }\
46de8b73 810
b0880d5d 811#define YSCALE_YUV_2_GRAY16_1_C \
dd68318c 812 for (i=0; i<(dstW>>1); i++) {\
b0880d5d
MN
813 const int i2= 2*i;\
814 int Y1= buf0[i2 ]<<1;\
815 int Y2= buf0[i2+1]<<1;\
816
6858492e
CS
817#define YSCALE_YUV_2_RGB1_C(type,alpha) \
818 YSCALE_YUV_2_PACKED1_C(type,alpha)\
221b804f
DB
819 r = (type *)c->table_rV[V];\
820 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
821 b = (type *)c->table_bU[U];\
cf7d1c1a 822
6858492e 823#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
dd68318c 824 for (i=0; i<(dstW>>1); i++) {\
221b804f
DB
825 const int i2= 2*i;\
826 int Y1= buf0[i2 ]>>7;\
827 int Y2= buf0[i2+1]>>7;\
828 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\
8b2fce0d 829 int V= (uvbuf0[i+VOFW] + uvbuf1[i+VOFW])>>8;\
6858492e
CS
830 type av_unused *r, *b, *g;\
831 int av_unused A1, A2;\
dd68318c 832 if (alpha) {\
6858492e
CS
833 A1= abuf0[i2 ]>>7;\
834 A2= abuf0[i2+1]>>7;\
835 }\
46de8b73 836
6858492e
CS
837#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
838 YSCALE_YUV_2_PACKED1B_C(type,alpha)\
221b804f
DB
839 r = (type *)c->table_rV[V];\
840 g = (type *)(c->table_gU[U] + c->table_gV[V]);\
841 b = (type *)c->table_bU[U];\
cf7d1c1a 842
ec1bca2a 843#define YSCALE_YUV_2_MONO2_C \
e69bd294
MN
844 const uint8_t * const d128=dither_8x8_220[y&7];\
845 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
dd68318c 846 for (i=0; i<dstW-7; i+=8) {\
e69bd294
MN
847 int acc;\
848 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\
849 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
850 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
851 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
852 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
853 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
854 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
855 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
ec1bca2a 856 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
e69bd294
MN
857 dest++;\
858 }\
859
860
ec1bca2a 861#define YSCALE_YUV_2_MONOX_C \
e69bd294
MN
862 const uint8_t * const d128=dither_8x8_220[y&7];\
863 uint8_t *g= c->table_gU[128] + c->table_gV[128];\
864 int acc=0;\
dd68318c 865 for (i=0; i<dstW-1; i+=2) {\
e69bd294
MN
866 int j;\
867 int Y1=1<<18;\
868 int Y2=1<<18;\
869\
dd68318c 870 for (j=0; j<lumFilterSize; j++) {\
e69bd294
MN
871 Y1 += lumSrc[j][i] * lumFilter[j];\
872 Y2 += lumSrc[j][i+1] * lumFilter[j];\
873 }\
874 Y1>>=19;\
875 Y2>>=19;\
dd68318c 876 if ((Y1|Y2)&256) {\
e69bd294
MN
877 if (Y1>255) Y1=255;\
878 else if (Y1<0)Y1=0;\
879 if (Y2>255) Y2=255;\
880 else if (Y2<0)Y2=0;\
881 }\
882 acc+= acc + g[Y1+d128[(i+0)&7]];\
883 acc+= acc + g[Y2+d128[(i+1)&7]];\
dd68318c 884 if ((i&7)==6) {\
ec1bca2a 885 ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
e69bd294
MN
886 dest++;\
887 }\
888 }
889
890
891#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
dd68318c 892 switch(c->dstFormat) {\
68e7f482
KS
893 case PIX_FMT_RGB48BE:\
894 case PIX_FMT_RGB48LE:\
895 func(uint8_t,0)\
896 ((uint8_t*)dest)[ 0]= r[Y1];\
897 ((uint8_t*)dest)[ 1]= r[Y1];\
898 ((uint8_t*)dest)[ 2]= g[Y1];\
899 ((uint8_t*)dest)[ 3]= g[Y1];\
900 ((uint8_t*)dest)[ 4]= b[Y1];\
901 ((uint8_t*)dest)[ 5]= b[Y1];\
902 ((uint8_t*)dest)[ 6]= r[Y2];\
903 ((uint8_t*)dest)[ 7]= r[Y2];\
904 ((uint8_t*)dest)[ 8]= g[Y2];\
905 ((uint8_t*)dest)[ 9]= g[Y2];\
906 ((uint8_t*)dest)[10]= b[Y2];\
907 ((uint8_t*)dest)[11]= b[Y2];\
908 dest+=12;\
909 }\
910 break;\
6858492e
CS
911 case PIX_FMT_RGBA:\
912 case PIX_FMT_BGRA:\
dd68318c 913 if (CONFIG_SMALL) {\
6858492e
CS
914 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
915 func(uint32_t,needAlpha)\
916 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
917 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
918 }\
dd68318c
RP
919 } else {\
920 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
6858492e
CS
921 func(uint32_t,1)\
922 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
923 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
924 }\
dd68318c 925 } else {\
6858492e
CS
926 func(uint32_t,0)\
927 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
928 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
929 }\
930 }\
931 }\
932 break;\
933 case PIX_FMT_ARGB:\
934 case PIX_FMT_ABGR:\
dd68318c 935 if (CONFIG_SMALL) {\
6858492e
CS
936 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
937 func(uint32_t,needAlpha)\
938 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
939 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
940 }\
dd68318c
RP
941 } else {\
942 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
6858492e
CS
943 func(uint32_t,1)\
944 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
945 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
946 }\
dd68318c 947 } else {\
6858492e
CS
948 func(uint32_t,0)\
949 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
950 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
951 }\
952 }\
221b804f
DB
953 } \
954 break;\
955 case PIX_FMT_RGB24:\
6858492e 956 func(uint8_t,0)\
221b804f
DB
957 ((uint8_t*)dest)[0]= r[Y1];\
958 ((uint8_t*)dest)[1]= g[Y1];\
959 ((uint8_t*)dest)[2]= b[Y1];\
960 ((uint8_t*)dest)[3]= r[Y2];\
961 ((uint8_t*)dest)[4]= g[Y2];\
962 ((uint8_t*)dest)[5]= b[Y2];\
963 dest+=6;\
964 }\
965 break;\
966 case PIX_FMT_BGR24:\
6858492e 967 func(uint8_t,0)\
221b804f
DB
968 ((uint8_t*)dest)[0]= b[Y1];\
969 ((uint8_t*)dest)[1]= g[Y1];\
970 ((uint8_t*)dest)[2]= r[Y1];\
971 ((uint8_t*)dest)[3]= b[Y2];\
972 ((uint8_t*)dest)[4]= g[Y2];\
973 ((uint8_t*)dest)[5]= r[Y2];\
974 dest+=6;\
975 }\
976 break;\
977 case PIX_FMT_RGB565:\
978 case PIX_FMT_BGR565:\
979 {\
980 const int dr1= dither_2x2_8[y&1 ][0];\
981 const int dg1= dither_2x2_4[y&1 ][0];\
982 const int db1= dither_2x2_8[(y&1)^1][0];\
983 const int dr2= dither_2x2_8[y&1 ][1];\
984 const int dg2= dither_2x2_4[y&1 ][1];\
985 const int db2= dither_2x2_8[(y&1)^1][1];\
6858492e 986 func(uint16_t,0)\
221b804f
DB
987 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
988 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
989 }\
990 }\
991 break;\
992 case PIX_FMT_RGB555:\
993 case PIX_FMT_BGR555:\
994 {\
995 const int dr1= dither_2x2_8[y&1 ][0];\
996 const int dg1= dither_2x2_8[y&1 ][1];\
997 const int db1= dither_2x2_8[(y&1)^1][0];\
998 const int dr2= dither_2x2_8[y&1 ][1];\
999 const int dg2= dither_2x2_8[y&1 ][0];\
1000 const int db2= dither_2x2_8[(y&1)^1][1];\
6858492e 1001 func(uint16_t,0)\
221b804f
DB
1002 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
1003 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
1004 }\
1005 }\
1006 break;\
1007 case PIX_FMT_RGB8:\
1008 case PIX_FMT_BGR8:\
1009 {\
1010 const uint8_t * const d64= dither_8x8_73[y&7];\
1011 const uint8_t * const d32= dither_8x8_32[y&7];\
6858492e 1012 func(uint8_t,0)\
221b804f
DB
1013 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
1014 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
1015 }\
1016 }\
1017 break;\
1018 case PIX_FMT_RGB4:\
1019 case PIX_FMT_BGR4:\
1020 {\
1021 const uint8_t * const d64= dither_8x8_73 [y&7];\
1022 const uint8_t * const d128=dither_8x8_220[y&7];\
6858492e 1023 func(uint8_t,0)\
221b804f
DB
1024 ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
1025 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
1026 }\
1027 }\
1028 break;\
1029 case PIX_FMT_RGB4_BYTE:\
1030 case PIX_FMT_BGR4_BYTE:\
1031 {\
1032 const uint8_t * const d64= dither_8x8_73 [y&7];\
1033 const uint8_t * const d128=dither_8x8_220[y&7];\
6858492e 1034 func(uint8_t,0)\
221b804f
DB
1035 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
1036 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
1037 }\
1038 }\
1039 break;\
1040 case PIX_FMT_MONOBLACK:\
ec1bca2a 1041 case PIX_FMT_MONOWHITE:\
221b804f 1042 {\
e69bd294 1043 func_monoblack\
221b804f
DB
1044 }\
1045 break;\
1046 case PIX_FMT_YUYV422:\
1047 func2\
1048 ((uint8_t*)dest)[2*i2+0]= Y1;\
1049 ((uint8_t*)dest)[2*i2+1]= U;\
1050 ((uint8_t*)dest)[2*i2+2]= Y2;\
1051 ((uint8_t*)dest)[2*i2+3]= V;\
1052 } \
1053 break;\
1054 case PIX_FMT_UYVY422:\
1055 func2\
1056 ((uint8_t*)dest)[2*i2+0]= U;\
1057 ((uint8_t*)dest)[2*i2+1]= Y1;\
1058 ((uint8_t*)dest)[2*i2+2]= V;\
1059 ((uint8_t*)dest)[2*i2+3]= Y2;\
1060 } \
1061 break;\
b0880d5d
MN
1062 case PIX_FMT_GRAY16BE:\
1063 func_g16\
1064 ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
1065 ((uint8_t*)dest)[2*i2+1]= Y1;\
1066 ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
1067 ((uint8_t*)dest)[2*i2+3]= Y2;\
1068 } \
1069 break;\
1070 case PIX_FMT_GRAY16LE:\
1071 func_g16\
1072 ((uint8_t*)dest)[2*i2+0]= Y1;\
1073 ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
1074 ((uint8_t*)dest)[2*i2+2]= Y2;\
1075 ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
1076 } \
1077 break;\
221b804f 1078 }\
cf7d1c1a
MN
1079
1080
7ac40327
RP
1081static inline void yuv2packedXinC(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1082 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1083 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
e3d2500f 1084{
221b804f 1085 int i;
6858492e 1086 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)
e3d2500f
MN
1087}
1088
7ac40327
RP
1089static inline void yuv2rgbXinC_full(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
1090 const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
1091 const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
f0faee4c
MN
1092{
1093 int i;
1094 int step= fmt_depth(c->dstFormat)/8;
d616c8ae 1095 int aidx= 3;
f0faee4c 1096
dd68318c 1097 switch(c->dstFormat) {
f0faee4c
MN
1098 case PIX_FMT_ARGB:
1099 dest++;
a3398feb 1100 aidx= 0;
f0faee4c 1101 case PIX_FMT_RGB24:
d616c8ae 1102 aidx--;
f0faee4c 1103 case PIX_FMT_RGBA:
dd68318c 1104 if (CONFIG_SMALL) {
6858492e
CS
1105 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1106 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1107 dest[aidx]= needAlpha ? A : 255;
1108 dest[0]= R>>22;
1109 dest[1]= G>>22;
1110 dest[2]= B>>22;
1111 dest+= step;
1112 }
dd68318c
RP
1113 } else {
1114 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
6858492e
CS
1115 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1116 dest[aidx]= A;
1117 dest[0]= R>>22;
1118 dest[1]= G>>22;
1119 dest[2]= B>>22;
1120 dest+= step;
1121 }
dd68318c 1122 } else {
6858492e
CS
1123 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1124 dest[aidx]= 255;
1125 dest[0]= R>>22;
1126 dest[1]= G>>22;
1127 dest[2]= B>>22;
1128 dest+= step;
1129 }
1130 }
f0faee4c
MN
1131 }
1132 break;
1133 case PIX_FMT_ABGR:
1134 dest++;
a3398feb 1135 aidx= 0;
f0faee4c 1136 case PIX_FMT_BGR24:
d616c8ae 1137 aidx--;
f0faee4c 1138 case PIX_FMT_BGRA:
dd68318c 1139 if (CONFIG_SMALL) {
6858492e
CS
1140 int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
1141 YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
1142 dest[aidx]= needAlpha ? A : 255;
1143 dest[0]= B>>22;
1144 dest[1]= G>>22;
1145 dest[2]= R>>22;
1146 dest+= step;
1147 }
dd68318c
RP
1148 } else {
1149 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
6858492e
CS
1150 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
1151 dest[aidx]= A;
1152 dest[0]= B>>22;
1153 dest[1]= G>>22;
1154 dest[2]= R>>22;
1155 dest+= step;
1156 }
dd68318c 1157 } else {
6858492e
CS
1158 YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
1159 dest[aidx]= 255;
1160 dest[0]= B>>22;
1161 dest[1]= G>>22;
1162 dest[2]= R>>22;
1163 dest+= step;
1164 }
1165 }
f0faee4c
MN
1166 }
1167 break;
1168 default:
1169 assert(0);
1170 }
1171}
e3d2500f 1172
dd68318c
RP
1173static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
1174{
d4da3e47
CS
1175 int i;
1176 uint8_t *ptr = plane + stride*y;
dd68318c 1177 for (i=0; i<height; i++) {
d4da3e47
CS
1178 memset(ptr, val, width);
1179 ptr += stride;
1180 }
1181}
1182
e8417235
KS
1183static inline void rgb48ToY(uint8_t *dst, const uint8_t *src, int width)
1184{
1185 int i;
1186 for (i = 0; i < width; i++) {
1187 int r = src[i*6+0];
1188 int g = src[i*6+2];
1189 int b = src[i*6+4];
1190
1191 dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1192 }
1193}
1194
1195static inline void rgb48ToUV(uint8_t *dstU, uint8_t *dstV,
1196 uint8_t *src1, uint8_t *src2, int width)
1197{
1198 int i;
1199 assert(src1==src2);
1200 for (i = 0; i < width; i++) {
1201 int r = src1[6*i + 0];
1202 int g = src1[6*i + 2];
1203 int b = src1[6*i + 4];
1204
1205 dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1206 dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
1207 }
1208}
1209
1210static inline void rgb48ToUV_half(uint8_t *dstU, uint8_t *dstV,
1211 uint8_t *src1, uint8_t *src2, int width)
1212{
1213 int i;
1214 assert(src1==src2);
1215 for (i = 0; i < width; i++) {
1216 int r= src1[12*i + 0] + src1[12*i + 6];
1217 int g= src1[12*i + 2] + src1[12*i + 8];
1218 int b= src1[12*i + 4] + src1[12*i + 10];
1219
1220 dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1221 dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
1222 }
1223}
1224
80704c47
KS
1225#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
1226static inline void name(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)\
1227{\
1228 int i;\
dd68318c 1229 for (i=0; i<width; i++) {\
80704c47
KS
1230 int b= (((const type*)src)[i]>>shb)&maskb;\
1231 int g= (((const type*)src)[i]>>shg)&maskg;\
1232 int r= (((const type*)src)[i]>>shr)&maskr;\
1233\
1234 dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
1235 }\
1236}
1237
1238BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1239BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY , BY<< 8, RGB2YUV_SHIFT+8)
1240BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY , RGB2YUV_SHIFT+8)
1241BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY , RGB2YUV_SHIFT+7)
1242BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
1243BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
1244
dd68318c
RP
1245static inline void abgrToA(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1246{
80704c47 1247 int i;
dd68318c 1248 for (i=0; i<width; i++) {
80704c47
KS
1249 dst[i]= src[4*i];
1250 }
1251}
1252
1253#define BGR2UV(type, name, shr, shg, shb, maska, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S)\
1254static inline void name(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1255{\
1256 int i;\
dd68318c 1257 for (i=0; i<width; i++) {\
80704c47
KS
1258 int b= (((const type*)src)[i]&maskb)>>shb;\
1259 int g= (((const type*)src)[i]&maskg)>>shg;\
1260 int r= (((const type*)src)[i]&maskr)>>shr;\
1261\
1262 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
1263 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
1264 }\
1265}\
1266static inline void name ## _half(uint8_t *dstU, uint8_t *dstV, const uint8_t *src, const uint8_t *dummy, long width, uint32_t *unused)\
1267{\
1268 int i;\
dd68318c 1269 for (i=0; i<width; i++) {\
80704c47
KS
1270 int pix0= ((const type*)src)[2*i+0];\
1271 int pix1= ((const type*)src)[2*i+1];\
1272 int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
1273 int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
1274 int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
1275 g&= maskg|(2*maskg);\
1276\
1277 g>>=shg;\
1278\
1279 dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
1280 dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
1281 }\
1282}
1283
1284BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0xFF000000, 0xFF0000, 0xFF00, 0x00FF, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1285BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0xFF000000, 0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU , BU<< 8, RV<< 8, GV , BV<< 8, RGB2YUV_SHIFT+8)
1286BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RU<<11, GU<<5, BU , RV<<11, GV<<5, BV , RGB2YUV_SHIFT+8)
1287BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RU<<10, GU<<5, BU , RV<<10, GV<<5, BV , RGB2YUV_SHIFT+7)
1288BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RU , GU<<5, BU<<11, RV , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
1289BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RU , GU<<5, BU<<10, RV , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
1290
1291static inline void palToY(uint8_t *dst, const uint8_t *src, long width, uint32_t *pal)
1292{
1293 int i;
dd68318c 1294 for (i=0; i<width; i++) {
80704c47
KS
1295 int d= src[i];
1296
1297 dst[i]= pal[d] & 0xFF;
1298 }
1299}
1300
1301static inline void palToUV(uint8_t *dstU, uint8_t *dstV,
1302 const uint8_t *src1, const uint8_t *src2,
1303 long width, uint32_t *pal)
1304{
1305 int i;
1306 assert(src1 == src2);
dd68318c 1307 for (i=0; i<width; i++) {
80704c47
KS
1308 int p= pal[src1[i]];
1309
1310 dstU[i]= p>>8;
1311 dstV[i]= p>>16;
1312 }
1313}
1314
1315static inline void monowhite2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1316{
1317 int i, j;
dd68318c 1318 for (i=0; i<width/8; i++) {
80704c47
KS
1319 int d= ~src[i];
1320 for(j=0; j<8; j++)
1321 dst[8*i+j]= ((d>>(7-j))&1)*255;
1322 }
1323}
1324
1325static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uint32_t *unused)
1326{
1327 int i, j;
dd68318c 1328 for (i=0; i<width/8; i++) {
80704c47
KS
1329 int d= src[i];
1330 for(j=0; j<8; j++)
1331 dst[8*i+j]= ((d>>(7-j))&1)*255;
1332 }
1333}
1334
1335
dd409025 1336//Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
7630f2e0 1337//Plain C versions
8b1a6441 1338#if ((!HAVE_MMX || !CONFIG_GPL) && !HAVE_ALTIVEC) || CONFIG_RUNTIME_CPUDETECT
726a959a
MN
1339#define COMPILE_C
1340#endif
1341
b63f641e 1342#if ARCH_PPC
57f6d52a 1343#if HAVE_ALTIVEC || CONFIG_RUNTIME_CPUDETECT
a2faa401 1344#define COMPILE_ALTIVEC
7a24ec50 1345#endif
cb82a073 1346#endif //ARCH_PPC
a2faa401 1347
b63f641e 1348#if ARCH_X86
726a959a 1349
10a7d216 1350#if ((HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
726a959a
MN
1351#define COMPILE_MMX
1352#endif
1353
10a7d216 1354#if (HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
726a959a
MN
1355#define COMPILE_MMX2
1356#endif
1357
10a7d216 1358#if ((HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT) && CONFIG_GPL
726a959a
MN
1359#define COMPILE_3DNOW
1360#endif
7a24ec50 1361#endif //ARCH_X86
726a959a 1362
94daf2e9
RP
1363#define COMPILE_TEMPLATE_MMX 0
1364#define COMPILE_TEMPLATE_MMX2 0
1365#define COMPILE_TEMPLATE_AMD3DNOW 0
1366#define COMPILE_TEMPLATE_ALTIVEC 0
726a959a
MN
1367
1368#ifdef COMPILE_C
7630f2e0
MN
1369#define RENAME(a) a ## _C
1370#include "swscale_template.c"
726a959a 1371#endif
397c035e 1372
a2faa401
RD
1373#ifdef COMPILE_ALTIVEC
1374#undef RENAME
94daf2e9
RP
1375#undef COMPILE_TEMPLATE_ALTIVEC
1376#define COMPILE_TEMPLATE_ALTIVEC 1
a2faa401
RD
1377#define RENAME(a) a ## _altivec
1378#include "swscale_template.c"
1379#endif
a2faa401 1380
b63f641e 1381#if ARCH_X86
397c035e 1382
7630f2e0 1383//MMX versions
726a959a 1384#ifdef COMPILE_MMX
7630f2e0 1385#undef RENAME
94daf2e9
RP
1386#undef COMPILE_TEMPLATE_MMX
1387#undef COMPILE_TEMPLATE_MMX2
1388#undef COMPILE_TEMPLATE_AMD3DNOW
1389#define COMPILE_TEMPLATE_MMX 1
1390#define COMPILE_TEMPLATE_MMX2 0
1391#define COMPILE_TEMPLATE_AMD3DNOW 0
7630f2e0
MN
1392#define RENAME(a) a ## _MMX
1393#include "swscale_template.c"
726a959a 1394#endif
7630f2e0
MN
1395
1396//MMX2 versions
726a959a 1397#ifdef COMPILE_MMX2
7630f2e0 1398#undef RENAME
94daf2e9
RP
1399#undef COMPILE_TEMPLATE_MMX
1400#undef COMPILE_TEMPLATE_MMX2
1401#undef COMPILE_TEMPLATE_AMD3DNOW
1402#define COMPILE_TEMPLATE_MMX 1
1403#define COMPILE_TEMPLATE_MMX2 1
1404#define COMPILE_TEMPLATE_AMD3DNOW 0
7630f2e0
MN
1405#define RENAME(a) a ## _MMX2
1406#include "swscale_template.c"
726a959a 1407#endif
7630f2e0
MN
1408
1409//3DNOW versions
726a959a 1410#ifdef COMPILE_3DNOW
7630f2e0 1411#undef RENAME
94daf2e9
RP
1412#undef COMPILE_TEMPLATE_MMX
1413#undef COMPILE_TEMPLATE_MMX2
1414#undef COMPILE_TEMPLATE_AMD3DNOW
1415#define COMPILE_TEMPLATE_MMX 1
1416#define COMPILE_TEMPLATE_MMX2 0
1417#define COMPILE_TEMPLATE_AMD3DNOW 1
7630f2e0
MN
1418#define RENAME(a) a ## _3DNow
1419#include "swscale_template.c"
726a959a 1420#endif
7630f2e0 1421
7a24ec50 1422#endif //ARCH_X86
7630f2e0 1423
a86c461c
MN
1424static double getSplineCoeff(double a, double b, double c, double d, double dist)
1425{
221b804f 1426// printf("%f %f %f %f %f\n", a,b,c,d,dist);
9b734d44
RP
1427 if (dist<=1.0) return ((d*dist + c)*dist + b)*dist +a;
1428 else return getSplineCoeff( 0.0,
1429 b+ 2.0*c + 3.0*d,
1430 c + 3.0*d,
1431 -b- 3.0*c - 6.0*d,
1432 dist-1.0);
a86c461c 1433}
6c7506de 1434
bca11e75 1435static inline int initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
221b804f
DB
1436 int srcW, int dstW, int filterAlign, int one, int flags,
1437 SwsVector *srcFilter, SwsVector *dstFilter, double param[2])
28bf81c9 1438{
221b804f
DB
1439 int i;
1440 int filterSize;
1441 int filter2Size;
1442 int minFilterSize;
a64a062f
MN
1443 int64_t *filter=NULL;
1444 int64_t *filter2=NULL;
1445 const int64_t fone= 1LL<<54;
091d3bdc 1446 int ret= -1;
b63f641e 1447#if ARCH_X86
221b804f 1448 if (flags & SWS_CPU_CAPS_MMX)
7ad6469e 1449 __asm__ volatile("emms\n\t"::: "memory"); //FIXME this should not be required but it IS (even for non-MMX versions)
726a959a 1450#endif
31190492 1451
f40c7dbb 1452 // NOTE: the +1 is for the MMX scaler which reads over the end
221b804f
DB
1453 *filterPos = av_malloc((dstW+1)*sizeof(int16_t));
1454
dd68318c 1455 if (FFABS(xInc - 0x10000) <10) { // unscaled
221b804f
DB
1456 int i;
1457 filterSize= 1;
8588e148 1458 filter= av_mallocz(dstW*sizeof(*filter)*filterSize);
221b804f 1459
dd68318c 1460 for (i=0; i<dstW; i++) {
a64a062f 1461 filter[i*filterSize]= fone;
221b804f
DB
1462 (*filterPos)[i]=i;
1463 }
1464
dd68318c 1465 } else if (flags&SWS_POINT) { // lame looking point sampling mode
221b804f
DB
1466 int i;
1467 int xDstInSrc;
1468 filterSize= 1;
8588e148 1469 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
221b804f
DB
1470
1471 xDstInSrc= xInc/2 - 0x8000;
dd68318c 1472 for (i=0; i<dstW; i++) {
221b804f
DB
1473 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1474
1475 (*filterPos)[i]= xx;
a64a062f 1476 filter[i]= fone;
221b804f
DB
1477 xDstInSrc+= xInc;
1478 }
dd68318c 1479 } else if ((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) { // bilinear upscale
221b804f
DB
1480 int i;
1481 int xDstInSrc;
571d9587 1482 filterSize= 2;
8588e148 1483 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
221b804f
DB
1484
1485 xDstInSrc= xInc/2 - 0x8000;
dd68318c 1486 for (i=0; i<dstW; i++) {
221b804f
DB
1487 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;
1488 int j;
1489
1490 (*filterPos)[i]= xx;
9b734d44 1491 //bilinear upscale / linear interpolate / area averaging
dd68318c 1492 for (j=0; j<filterSize; j++) {
9b734d44
RP
1493 int64_t coeff= fone - FFABS((xx<<16) - xDstInSrc)*(fone>>16);
1494 if (coeff<0) coeff=0;
1495 filter[i*filterSize + j]= coeff;
1496 xx++;
1497 }
221b804f
DB
1498 xDstInSrc+= xInc;
1499 }
dd68318c 1500 } else {
a64a062f
MN
1501 int xDstInSrc;
1502 int sizeFactor;
1503
1504 if (flags&SWS_BICUBIC) sizeFactor= 4;
1505 else if (flags&SWS_X) sizeFactor= 8;
1506 else if (flags&SWS_AREA) sizeFactor= 1; //downscale only, for upscale it is bilinear
1507 else if (flags&SWS_GAUSS) sizeFactor= 8; // infinite ;)
1508 else if (flags&SWS_LANCZOS) sizeFactor= param[0] != SWS_PARAM_DEFAULT ? ceil(2*param[0]) : 6;
1509 else if (flags&SWS_SINC) sizeFactor= 20; // infinite ;)
1510 else if (flags&SWS_SPLINE) sizeFactor= 20; // infinite ;)
1511 else if (flags&SWS_BILINEAR) sizeFactor= 2;
221b804f 1512 else {
a64a062f 1513 sizeFactor= 0; //GCC warning killer
fcc402b1 1514 assert(0);
221b804f
DB
1515 }
1516
a64a062f
MN
1517 if (xInc <= 1<<16) filterSize= 1 + sizeFactor; // upscale
1518 else filterSize= 1 + (sizeFactor*srcW + dstW - 1)/ dstW;
221b804f 1519
221b804f
DB
1520 if (filterSize > srcW-2) filterSize=srcW-2;
1521
8588e148 1522 filter= av_malloc(dstW*sizeof(*filter)*filterSize);
221b804f 1523
a64a062f 1524 xDstInSrc= xInc - 0x10000;
dd68318c 1525 for (i=0; i<dstW; i++) {
a64a062f 1526 int xx= (xDstInSrc - ((filterSize-2)<<16)) / (1<<17);
221b804f
DB
1527 int j;
1528 (*filterPos)[i]= xx;
dd68318c 1529 for (j=0; j<filterSize; j++) {
a64a062f
MN
1530 int64_t d= ((int64_t)FFABS((xx<<17) - xDstInSrc))<<13;
1531 double floatd;
1532 int64_t coeff;
1533
1534 if (xInc > 1<<16)
1535 d= d*dstW/srcW;
1536 floatd= d * (1.0/(1<<30));
1537
dd68318c 1538 if (flags & SWS_BICUBIC) {
a64a062f
MN
1539 int64_t B= (param[0] != SWS_PARAM_DEFAULT ? param[0] : 0) * (1<<24);
1540 int64_t C= (param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6) * (1<<24);
1541 int64_t dd = ( d*d)>>30;
1542 int64_t ddd= (dd*d)>>30;
1543
1544 if (d < 1LL<<30)
1545 coeff = (12*(1<<24)-9*B-6*C)*ddd + (-18*(1<<24)+12*B+6*C)*dd + (6*(1<<24)-2*B)*(1<<30);
1546 else if (d < 1LL<<31)
1547 coeff = (-B-6*C)*ddd + (6*B+30*C)*dd + (-12*B-48*C)*d + (8*B+24*C)*(1<<30);
221b804f
DB
1548 else
1549 coeff=0.0;
a64a062f 1550 coeff *= fone>>(30+24);
221b804f 1551 }
dd68318c 1552/* else if (flags & SWS_X) {
221b804f
DB
1553 double p= param ? param*0.01 : 0.3;
1554 coeff = d ? sin(d*PI)/(d*PI) : 1.0;
1555 coeff*= pow(2.0, - p*d*d);
1556 }*/
dd68318c 1557 else if (flags & SWS_X) {
221b804f 1558 double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
a64a062f 1559 double c;
221b804f 1560
a64a062f
MN
1561 if (floatd<1.0)
1562 c = cos(floatd*PI);
221b804f 1563 else
a64a062f
MN
1564 c=-1.0;
1565 if (c<0.0) c= -pow(-c, A);
1566 else c= pow( c, A);
1567 coeff= (c*0.5 + 0.5)*fone;
dd68318c 1568 } else if (flags & SWS_AREA) {
a64a062f
MN
1569 int64_t d2= d - (1<<29);
1570 if (d2*xInc < -(1LL<<(29+16))) coeff= 1.0 * (1LL<<(30+16));
1571 else if (d2*xInc < (1LL<<(29+16))) coeff= -d2*xInc + (1LL<<(29+16));
221b804f 1572 else coeff=0.0;
a64a062f 1573 coeff *= fone>>(30+16);
dd68318c 1574 } else if (flags & SWS_GAUSS) {
221b804f 1575 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
a64a062f 1576 coeff = (pow(2.0, - p*floatd*floatd))*fone;
dd68318c 1577 } else if (flags & SWS_SINC) {
a64a062f 1578 coeff = (d ? sin(floatd*PI)/(floatd*PI) : 1.0)*fone;
dd68318c 1579 } else if (flags & SWS_LANCZOS) {
221b804f 1580 double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
a64a062f
MN
1581 coeff = (d ? sin(floatd*PI)*sin(floatd*PI/p)/(floatd*floatd*PI*PI/p) : 1.0)*fone;
1582 if (floatd>p) coeff=0;
dd68318c 1583 } else if (flags & SWS_BILINEAR) {
a64a062f 1584 coeff= (1<<30) - d;
221b804f 1585 if (coeff<0) coeff=0;
a64a062f 1586 coeff *= fone >> 30;
dd68318c 1587 } else if (flags & SWS_SPLINE) {
221b804f 1588 double p=-2.196152422706632;
f830d824 1589 coeff = getSplineCoeff(1.0, 0.0, p, -p-1.0, floatd) * fone;
dd68318c 1590 } else {
221b804f 1591 coeff= 0.0; //GCC warning killer
fcc402b1 1592 assert(0);
221b804f
DB
1593 }
1594
1595 filter[i*filterSize + j]= coeff;
1596 xx++;
1597 }
a64a062f 1598 xDstInSrc+= 2*xInc;
221b804f
DB
1599 }
1600 }
1601
1602 /* apply src & dst Filter to filter -> filter2
1603 av_free(filter);
1604 */
fcc402b1 1605 assert(filterSize>0);
221b804f
DB
1606 filter2Size= filterSize;
1607 if (srcFilter) filter2Size+= srcFilter->length - 1;
1608 if (dstFilter) filter2Size+= dstFilter->length - 1;
fcc402b1 1609 assert(filter2Size>0);
a64a062f 1610 filter2= av_mallocz(filter2Size*dstW*sizeof(*filter2));
221b804f 1611
dd68318c 1612 for (i=0; i<dstW; i++) {
a64a062f 1613 int j, k;
221b804f 1614
dd68318c
RP
1615 if(srcFilter) {
1616 for (k=0; k<srcFilter->length; k++) {
a64a062f
MN
1617 for (j=0; j<filterSize; j++)
1618 filter2[i*filter2Size + k + j] += srcFilter->coeff[k]*filter[i*filterSize + j];
1619 }
dd68318c 1620 } else {
a64a062f
MN
1621 for (j=0; j<filterSize; j++)
1622 filter2[i*filter2Size + j]= filter[i*filterSize + j];
221b804f 1623 }
a64a062f 1624 //FIXME dstFilter
221b804f
DB
1625
1626 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2;
221b804f 1627 }
47b7382d 1628 av_freep(&filter);
221b804f
DB
1629
1630 /* try to reduce the filter-size (step1 find size and shift left) */
86bdf3fd 1631 // Assume it is near normalized (*0.5 or *2.0 is OK but * 0.001 is not).
221b804f 1632 minFilterSize= 0;
dd68318c 1633 for (i=dstW-1; i>=0; i--) {
221b804f
DB
1634 int min= filter2Size;
1635 int j;
a64a062f 1636 int64_t cutOff=0.0;
221b804f
DB
1637
1638 /* get rid off near zero elements on the left by shifting left */
dd68318c 1639 for (j=0; j<filter2Size; j++) {
221b804f
DB
1640 int k;
1641 cutOff += FFABS(filter2[i*filter2Size]);
1642
a64a062f 1643 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
221b804f 1644
86bdf3fd 1645 /* preserve monotonicity because the core can't handle the filter otherwise */
221b804f
DB
1646 if (i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break;
1647
f40c7dbb 1648 // move filter coefficients left
221b804f
DB
1649 for (k=1; k<filter2Size; k++)
1650 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k];
a64a062f 1651 filter2[i*filter2Size + k - 1]= 0;
221b804f
DB
1652 (*filterPos)[i]++;
1653 }
1654
a64a062f 1655 cutOff=0;
221b804f 1656 /* count near zeros on the right */
dd68318c 1657 for (j=filter2Size-1; j>0; j--) {
221b804f
DB
1658 cutOff += FFABS(filter2[i*filter2Size + j]);
1659
a64a062f 1660 if (cutOff > SWS_MAX_REDUCE_CUTOFF*fone) break;
221b804f
DB
1661 min--;
1662 }
1663
1664 if (min>minFilterSize) minFilterSize= min;
1665 }
1666
1667 if (flags & SWS_CPU_CAPS_ALTIVEC) {
1668 // we can handle the special case 4,
1669 // so we don't want to go to the full 8
1670 if (minFilterSize < 5)
8c266f0c
RD
1671 filterAlign = 4;
1672
f40c7dbb
DB
1673 // We really don't want to waste our time
1674 // doing useless computation, so fall back on
1675 // the scalar C code for very small filters.
1676 // Vectorizing is worth it only if you have a
221b804f
DB
1677 // decent-sized vector.
1678 if (minFilterSize < 3)
8c266f0c 1679 filterAlign = 1;
221b804f
DB
1680 }
1681
1682 if (flags & SWS_CPU_CAPS_MMX) {
1683 // special case for unscaled vertical filtering
1684 if (minFilterSize == 1 && filterAlign == 2)
1685 filterAlign= 1;
1686 }
1687
fcc402b1 1688 assert(minFilterSize > 0);
221b804f 1689 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1));
fcc402b1 1690 assert(filterSize > 0);
8588e148 1691 filter= av_malloc(filterSize*dstW*sizeof(*filter));
1625216e 1692 if (filterSize >= MAX_FILTER_SIZE*16/((flags&SWS_ACCURATE_RND) ? APCK_SIZE : 16) || !filter)
091d3bdc 1693 goto error;
221b804f
DB
1694 *outFilterSize= filterSize;
1695
1696 if (flags&SWS_PRINT_INFO)
1697 av_log(NULL, AV_LOG_VERBOSE, "SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize);
1698 /* try to reduce the filter-size (step2 reduce it) */
dd68318c 1699 for (i=0; i<dstW; i++) {
221b804f
DB
1700 int j;
1701
dd68318c 1702 for (j=0; j<filterSize; j++) {
a64a062f 1703 if (j>=filter2Size) filter[i*filterSize + j]= 0;
221b804f 1704 else filter[i*filterSize + j]= filter2[i*filter2Size + j];
88bc5a64 1705 if((flags & SWS_BITEXACT) && j>=minFilterSize)
a64a062f 1706 filter[i*filterSize + j]= 0;
8c266f0c 1707 }
221b804f 1708 }
221b804f
DB
1709
1710
f40c7dbb 1711 //FIXME try to align filterPos if possible
8c266f0c 1712
221b804f 1713 //fix borders
dd68318c 1714 for (i=0; i<dstW; i++) {
221b804f 1715 int j;
dd68318c 1716 if ((*filterPos)[i] < 0) {
f40c7dbb 1717 // move filter coefficients left to compensate for filterPos
dd68318c 1718 for (j=1; j<filterSize; j++) {
221b804f
DB
1719 int left= FFMAX(j + (*filterPos)[i], 0);
1720 filter[i*filterSize + left] += filter[i*filterSize + j];
1721 filter[i*filterSize + j]=0;
1722 }
1723 (*filterPos)[i]= 0;
bca11e75
MN
1724 }
1725
dd68318c 1726 if ((*filterPos)[i] + filterSize > srcW) {
221b804f 1727 int shift= (*filterPos)[i] + filterSize - srcW;
f40c7dbb 1728 // move filter coefficients right to compensate for filterPos
dd68318c 1729 for (j=filterSize-2; j>=0; j--) {
221b804f
DB
1730 int right= FFMIN(j + shift, filterSize-1);
1731 filter[i*filterSize +right] += filter[i*filterSize +j];
1732 filter[i*filterSize +j]=0;
1733 }
1734 (*filterPos)[i]= srcW - filterSize;
1735 }
1736 }
1737
f40c7dbb 1738 // Note the +1 is for the MMX scaler which reads over the end
221b804f
DB
1739 /* align at 16 for AltiVec (needed by hScale_altivec_real) */
1740 *outFilter= av_mallocz(*outFilterSize*(dstW+1)*sizeof(int16_t));
1741
f40c7dbb 1742 /* normalize & store in outFilter */
dd68318c 1743 for (i=0; i<dstW; i++) {
221b804f 1744 int j;
a64a062f
MN
1745 int64_t error=0;
1746 int64_t sum=0;
221b804f 1747
dd68318c 1748 for (j=0; j<filterSize; j++) {
221b804f
DB
1749 sum+= filter[i*filterSize + j];
1750 }
a64a062f 1751 sum= (sum + one/2)/ one;
dd68318c 1752 for (j=0; j<*outFilterSize; j++) {
a64a062f
MN
1753 int64_t v= filter[i*filterSize + j] + error;
1754 int intV= ROUNDED_DIV(v, sum);
221b804f 1755 (*outFilter)[i*(*outFilterSize) + j]= intV;
a64a062f 1756 error= v - intV*sum;
221b804f
DB
1757 }
1758 }
1759
1760 (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
dd68318c 1761 for (i=0; i<*outFilterSize; i++) {
221b804f
DB
1762 int j= dstW*(*outFilterSize);
1763 (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];
1764 }
1765
091d3bdc
MN
1766 ret=0;
1767error:
221b804f 1768 av_free(filter);
091d3bdc
MN
1769 av_free(filter2);
1770 return ret;
7630f2e0 1771}
31190492 1772
17c613ef 1773#ifdef COMPILE_MMX2
48f4c612 1774static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *filter, int32_t *filterPos, int numSplits)
28bf81c9 1775{
221b804f 1776 uint8_t *fragmentA;
d0ce212a
RP
1777 x86_reg imm8OfPShufW1A;
1778 x86_reg imm8OfPShufW2A;
1779 x86_reg fragmentLengthA;
221b804f 1780 uint8_t *fragmentB;
d0ce212a
RP
1781 x86_reg imm8OfPShufW1B;
1782 x86_reg imm8OfPShufW2B;
1783 x86_reg fragmentLengthB;
221b804f
DB
1784 int fragmentPos;
1785
1786 int xpos, i;
1787
1788 // create an optimized horizontal scaling routine
1789
1790 //code fragment
1791
7ad6469e 1792 __asm__ volatile(
221b804f
DB
1793 "jmp 9f \n\t"
1794 // Begin
1795 "0: \n\t"
1796 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
1797 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
1798 "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t"
1799 "punpcklbw %%mm7, %%mm1 \n\t"
1800 "punpcklbw %%mm7, %%mm0 \n\t"
1801 "pshufw $0xFF, %%mm1, %%mm1 \n\t"
1802 "1: \n\t"
1803 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1804 "2: \n\t"
1805 "psubw %%mm1, %%mm0 \n\t"
1806 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
1807 "pmullw %%mm3, %%mm0 \n\t"
1808 "psllw $7, %%mm1 \n\t"
1809 "paddw %%mm1, %%mm0 \n\t"
1810
1811 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1812
1813 "add $8, %%"REG_a" \n\t"
1814 // End
1815 "9: \n\t"
1816// "int $3 \n\t"
86593486
AS
1817 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
1818 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
1819 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
221b804f
DB
1820 "dec %1 \n\t"
1821 "dec %2 \n\t"
1822 "sub %0, %1 \n\t"
1823 "sub %0, %2 \n\t"
86593486 1824 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
221b804f
DB
1825 "sub %0, %3 \n\t"
1826
1827
1828 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A),
1829 "=r" (fragmentLengthA)
1830 );
1831
7ad6469e 1832 __asm__ volatile(
221b804f
DB
1833 "jmp 9f \n\t"
1834 // Begin
1835 "0: \n\t"
1836 "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t"
1837 "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t"
1838 "punpcklbw %%mm7, %%mm0 \n\t"
1839 "pshufw $0xFF, %%mm0, %%mm1 \n\t"
1840 "1: \n\t"
1841 "pshufw $0xFF, %%mm0, %%mm0 \n\t"
1842 "2: \n\t"
1843 "psubw %%mm1, %%mm0 \n\t"
1844 "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t"
1845 "pmullw %%mm3, %%mm0 \n\t"
1846 "psllw $7, %%mm1 \n\t"
1847 "paddw %%mm1, %%mm0 \n\t"
1848
1849 "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t"
1850
1851 "add $8, %%"REG_a" \n\t"
1852 // End
1853 "9: \n\t"
1854// "int $3 \n\t"
86593486
AS
1855 "lea " LOCAL_MANGLE(0b) ", %0 \n\t"
1856 "lea " LOCAL_MANGLE(1b) ", %1 \n\t"
1857 "lea " LOCAL_MANGLE(2b) ", %2 \n\t"
221b804f
DB
1858 "dec %1 \n\t"
1859 "dec %2 \n\t"
1860 "sub %0, %1 \n\t"
1861 "sub %0, %2 \n\t"
86593486 1862 "lea " LOCAL_MANGLE(9b) ", %3 \n\t"
221b804f
DB
1863 "sub %0, %3 \n\t"
1864
1865
1866 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B),
1867 "=r" (fragmentLengthB)
1868 );
1869
1870 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
1871 fragmentPos=0;
1872
dd68318c 1873 for (i=0; i<dstW/numSplits; i++) {
221b804f
DB
1874 int xx=xpos>>16;
1875
dd68318c 1876 if ((i&3) == 0) {
221b804f
DB
1877 int a=0;
1878 int b=((xpos+xInc)>>16) - xx;
1879 int c=((xpos+xInc*2)>>16) - xx;
1880 int d=((xpos+xInc*3)>>16) - xx;
901775c7 1881 int inc = (d+1<4);
f1c6dfa3
RP
1882 uint8_t *fragment = (d+1<4) ? fragmentB : fragmentA;
1883 x86_reg imm8OfPShufW1 = (d+1<4) ? imm8OfPShufW1B : imm8OfPShufW1A;
1884 x86_reg imm8OfPShufW2 = (d+1<4) ? imm8OfPShufW2B : imm8OfPShufW2A;
1885 x86_reg fragmentLength = (d+1<4) ? fragmentLengthB : fragmentLengthA;
469fd9b0
RP
1886 int maxShift= 3-(d+inc);
1887 int shift=0;
221b804f 1888
48f4c612 1889 if (filterCode) {
737cbcde
RP
1890 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9;
1891 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9;
1892 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
1893 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
1894 filterPos[i/2]= xx;
221b804f 1895
737cbcde 1896 memcpy(filterCode + fragmentPos, fragment, fragmentLength);
221b804f 1897
737cbcde
RP
1898 filterCode[fragmentPos + imm8OfPShufW1]=
1899 (a+inc) | ((b+inc)<<2) | ((c+inc)<<4) | ((d+inc)<<6);
1900 filterCode[fragmentPos + imm8OfPShufW2]=
1901 a | (b<<2) | (c<<4) | (d<<6);
221b804f 1902
737cbcde
RP
1903 if (i+4-inc>=dstW) shift=maxShift; //avoid overread
1904 else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align
221b804f 1905
dd68318c 1906 if (shift && i>=shift) {
737cbcde
RP
1907 filterCode[fragmentPos + imm8OfPShufW1]+= 0x55*shift;
1908 filterCode[fragmentPos + imm8OfPShufW2]+= 0x55*shift;
1909 filterPos[i/2]-=shift;
1910 }
48f4c612 1911 }
221b804f 1912
469fd9b0
RP
1913 fragmentPos+= fragmentLength;
1914
48f4c612 1915 if (filterCode)
737cbcde 1916 filterCode[fragmentPos]= RET;
221b804f
DB
1917 }
1918 xpos+=xInc;
1919 }
48f4c612 1920 if (filterCode)
737cbcde 1921 filterPos[((i/2)+1)&(~1)]= xpos>>16; // needed to jump to the next part
48f4c612
RP
1922
1923 return fragmentPos + 1;
28bf81c9 1924}
17c613ef 1925#endif /* COMPILE_MMX2 */
28bf81c9 1926
dd68318c
RP
1927static void globalInit(void)
1928{
31190492
A
1929 // generating tables:
1930 int i;
dd68318c 1931 for (i=0; i<768; i++) {
221b804f
DB
1932 int c= av_clip_uint8(i-256);
1933 clip_table[i]=c;
b18ea156 1934 }
516b1f82 1935}
c1b0bfb4 1936
40fa5140 1937static SwsFunc getSwsFunc(SwsContext *c)
14d5c18c 1938{
8b1a6441 1939#if CONFIG_RUNTIME_CPUDETECT
40fa5140 1940 int flags = c->flags;
6a4970ab 1941
8b1a6441 1942#if ARCH_X86 && CONFIG_GPL
c14731d8 1943 // ordered per speed fastest first
14d5c18c 1944 if (flags & SWS_CPU_CAPS_MMX2) {
40fa5140 1945 sws_init_swScale_MMX2(c);
221b804f 1946 return swScale_MMX2;
14d5c18c 1947 } else if (flags & SWS_CPU_CAPS_3DNOW) {
40fa5140 1948 sws_init_swScale_3DNow(c);
221b804f 1949 return swScale_3DNow;
14d5c18c 1950 } else if (flags & SWS_CPU_CAPS_MMX) {
40fa5140 1951 sws_init_swScale_MMX(c);
221b804f 1952 return swScale_MMX;
14d5c18c 1953 } else {
40fa5140 1954 sws_init_swScale_C(c);
221b804f 1955 return swScale_C;
14d5c18c 1956 }
28bf81c9
MN
1957
1958#else
b63f641e 1959#if ARCH_PPC
14d5c18c 1960 if (flags & SWS_CPU_CAPS_ALTIVEC) {
40fa5140 1961 sws_init_swScale_altivec(c);
221b804f 1962 return swScale_altivec;
14d5c18c 1963 } else {
40fa5140 1964 sws_init_swScale_C(c);
221b804f 1965 return swScale_C;
14d5c18c 1966 }
a2faa401 1967#endif
40fa5140 1968 sws_init_swScale_C(c);
221b804f 1969 return swScale_C;
8b1a6441 1970#endif /* ARCH_X86 && CONFIG_GPL */
10a7d216 1971#else //CONFIG_RUNTIME_CPUDETECT
94daf2e9 1972#if COMPILE_TEMPLATE_MMX2
40fa5140 1973 sws_init_swScale_MMX2(c);
221b804f 1974 return swScale_MMX2;
94daf2e9 1975#elif COMPILE_TEMPLATE_AMD3DNOW
40fa5140 1976 sws_init_swScale_3DNow(c);
221b804f 1977 return swScale_3DNow;
94daf2e9 1978#elif COMPILE_TEMPLATE_MMX
40fa5140 1979 sws_init_swScale_MMX(c);
221b804f 1980 return swScale_MMX;
94daf2e9 1981#elif COMPILE_TEMPLATE_ALTIVEC
40fa5140 1982 sws_init_swScale_altivec(c);
221b804f 1983 return swScale_altivec;
28bf81c9 1984#else
40fa5140 1985 sws_init_swScale_C(c);
221b804f 1986 return swScale_C;
28bf81c9 1987#endif
10a7d216 1988#endif //!CONFIG_RUNTIME_CPUDETECT
31190492 1989}
7630f2e0 1990
d4e24275 1991static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
1992 int srcSliceH, uint8_t* dstParam[], int dstStride[])
1993{
221b804f
DB
1994 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
1995 /* Copy Y plane */
1996 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
1997 memcpy(dst, src[0], srcSliceH*dstStride[0]);
dd68318c 1998 else {
221b804f 1999 int i;
7ac40327 2000 const uint8_t *srcPtr= src[0];
221b804f 2001 uint8_t *dstPtr= dst;
dd68318c 2002 for (i=0; i<srcSliceH; i++) {
221b804f
DB
2003 memcpy(dstPtr, srcPtr, c->srcW);
2004 srcPtr+= srcStride[0];
2005 dstPtr+= dstStride[0];
2006 }
2007 }
2008 dst = dstParam[1] + dstStride[1]*srcSliceY/2;
2009 if (c->dstFormat == PIX_FMT_NV12)
30c48a0a 2010 interleaveBytes(src[1], src[2], dst, c->srcW/2, srcSliceH/2, srcStride[1], srcStride[2], dstStride[0]);
221b804f 2011 else
30c48a0a 2012 interleaveBytes(src[2], src[1], dst, c->srcW/2, srcSliceH/2, srcStride[2], srcStride[1], dstStride[0]);
221b804f
DB
2013
2014 return srcSliceH;
0d9f3d85
A
2015}
2016
d4e24275 2017static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2018 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2019{
221b804f 2020 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
0d9f3d85 2021
30c48a0a 2022 yv12toyuy2(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
fccb9b2b 2023
221b804f 2024 return srcSliceH;
0d9f3d85
A
2025}
2026
caeaabe7 2027static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2028 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2029{
221b804f 2030 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
caeaabe7 2031
30c48a0a 2032 yv12touyvy(src[0], src[1], src[2], dst, c->srcW, srcSliceH, srcStride[0], srcStride[1], dstStride[0]);
caeaabe7 2033
221b804f 2034 return srcSliceH;
caeaabe7
AB
2035}
2036
a6100f39 2037static int YUV422PToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2038 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2039{
a6100f39
BC
2040 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2041
2042 yuv422ptoyuy2(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2043
2044 return srcSliceH;
2045}
2046
2047static int YUV422PToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2048 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2049{
a6100f39
BC
2050 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
2051
2052 yuv422ptouyvy(src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0]);
2053
2054 return srcSliceH;
2055}
2056
0411072e 2057static int YUYV2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2058 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2059{
0411072e
MN
2060 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2061 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2062 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2063
2064 yuyvtoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2065
4626ee1a
CS
2066 if (dstParam[3])
2067 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2068
0411072e
MN
2069 return srcSliceH;
2070}
2071
2072static int YUYV2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2073 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2074{
0411072e 2075 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
72ef3dd7
MN
2076 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2077 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
0411072e
MN
2078
2079 yuyvtoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2080
2081 return srcSliceH;
2082}
2083
2084static int UYVY2YUV420Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2085 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2086{
0411072e
MN
2087 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
2088 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY/2;
2089 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY/2;
2090
2091 uyvytoyuv420(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2092
4626ee1a
CS
2093 if (dstParam[3])
2094 fillPlane(dstParam[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
2095
0411072e
MN
2096 return srcSliceH;
2097}
2098
2099static int UYVY2YUV422Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2100 int srcSliceH, uint8_t* dstParam[], int dstStride[])
2101{
0411072e 2102 uint8_t *ydst=dstParam[0] + dstStride[0]*srcSliceY;
72ef3dd7
MN
2103 uint8_t *udst=dstParam[1] + dstStride[1]*srcSliceY;
2104 uint8_t *vdst=dstParam[2] + dstStride[2]*srcSliceY;
0411072e
MN
2105
2106 uyvytoyuv422(ydst, udst, vdst, src[0], c->srcW, srcSliceH, dstStride[0], dstStride[1], srcStride[0]);
2107
2108 return srcSliceH;
2109}
2110
49004617 2111static int pal2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2112 int srcSliceH, uint8_t* dst[], int dstStride[])
2113{
f5a2c981
CEH
2114 const enum PixelFormat srcFormat= c->srcFormat;
2115 const enum PixelFormat dstFormat= c->dstFormat;
49004617
VS
2116 void (*conv)(const uint8_t *src, uint8_t *dst, long num_pixels,
2117 const uint8_t *palette)=NULL;
2118 int i;
2119 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2120 uint8_t *srcPtr= src[0];
2121
2122 if (!usePal(srcFormat))
2123 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2124 sws_format_name(srcFormat), sws_format_name(dstFormat));
2125
dd68318c 2126 switch(dstFormat) {
522ce957
VS
2127 case PIX_FMT_RGB32 : conv = palette8topacked32; break;
2128 case PIX_FMT_BGR32 : conv = palette8topacked32; break;
2129 case PIX_FMT_BGR32_1: conv = palette8topacked32; break;
2130 case PIX_FMT_RGB32_1: conv = palette8topacked32; break;
2131 case PIX_FMT_RGB24 : conv = palette8topacked24; break;
2132 case PIX_FMT_BGR24 : conv = palette8topacked24; break;
49004617
VS
2133 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
2134 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2135 }
2136
2137
2138 for (i=0; i<srcSliceH; i++) {
65f65c30 2139 conv(srcPtr, dstPtr, c->srcW, (uint8_t *) c->pal_rgb);
49004617
VS
2140 srcPtr+= srcStride[0];
2141 dstPtr+= dstStride[0];
2142 }
2143
2144 return srcSliceH;
2145}
2146
9990e426 2147/* {RGB,BGR}{15,16,24,32,32_1} -> {RGB,BGR}{15,16,24,32} */
d4e24275 2148static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2149 int srcSliceH, uint8_t* dst[], int dstStride[])
2150{
58e4b706
CEH
2151 const enum PixelFormat srcFormat= c->srcFormat;
2152 const enum PixelFormat dstFormat= c->dstFormat;
221b804f
DB
2153 const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
2154 const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
2155 const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
2156 const int dstId= fmt_depth(dstFormat) >> 2;
2157 void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;
2158
2159 /* BGR -> BGR */
2160 if ( (isBGR(srcFormat) && isBGR(dstFormat))
dd68318c
RP
2161 || (isRGB(srcFormat) && isRGB(dstFormat))) {
2162 switch(srcId | (dstId<<4)) {
221b804f
DB
2163 case 0x34: conv= rgb16to15; break;
2164 case 0x36: conv= rgb24to15; break;
2165 case 0x38: conv= rgb32to15; break;
2166 case 0x43: conv= rgb15to16; break;
2167 case 0x46: conv= rgb24to16; break;
2168 case 0x48: conv= rgb32to16; break;
2169 case 0x63: conv= rgb15to24; break;
2170 case 0x64: conv= rgb16to24; break;
2171 case 0x68: conv= rgb32to24; break;
2172 case 0x83: conv= rgb15to32; break;
2173 case 0x84: conv= rgb16to32; break;
2174 case 0x86: conv= rgb24to32; break;
3f0bc115 2175 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
221b804f
DB
2176 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2177 }
dd68318c
RP
2178 } else if ( (isBGR(srcFormat) && isRGB(dstFormat))
2179 || (isRGB(srcFormat) && isBGR(dstFormat))) {
2180 switch(srcId | (dstId<<4)) {
221b804f
DB
2181 case 0x33: conv= rgb15tobgr15; break;
2182 case 0x34: conv= rgb16tobgr15; break;
2183 case 0x36: conv= rgb24tobgr15; break;
2184 case 0x38: conv= rgb32tobgr15; break;
2185 case 0x43: conv= rgb15tobgr16; break;
2186 case 0x44: conv= rgb16tobgr16; break;
2187 case 0x46: conv= rgb24tobgr16; break;
2188 case 0x48: conv= rgb32tobgr16; break;
2189 case 0x63: conv= rgb15tobgr24; break;
2190 case 0x64: conv= rgb16tobgr24; break;
2191 case 0x66: conv= rgb24tobgr24; break;
2192 case 0x68: conv= rgb32tobgr24; break;
2193 case 0x83: conv= rgb15tobgr32; break;
2194 case 0x84: conv= rgb16tobgr32; break;
2195 case 0x86: conv= rgb24tobgr32; break;
2196 case 0x88: conv= rgb32tobgr32; break;
3f0bc115 2197 default: av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
221b804f
DB
2198 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
2199 }
dd68318c 2200 } else {
3f0bc115 2201 av_log(c, AV_LOG_ERROR, "internal error %s -> %s converter\n",
221b804f
DB
2202 sws_format_name(srcFormat), sws_format_name(dstFormat));
2203 }
2204
dd68318c 2205 if(conv) {
9990e426
MN
2206 uint8_t *srcPtr= src[0];
2207 if(srcFormat == PIX_FMT_RGB32_1 || srcFormat == PIX_FMT_BGR32_1)
2208 srcPtr += ALT32_CORR;
2209
5efaf000 2210 if (dstStride[0]*srcBpp == srcStride[0]*dstBpp && srcStride[0] > 0)
9990e426 2211 conv(srcPtr, dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
dd68318c 2212 else {
c4ca31d0 2213 int i;
c4ca31d0
BF
2214 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2215
dd68318c 2216 for (i=0; i<srcSliceH; i++) {
c4ca31d0
BF
2217 conv(srcPtr, dstPtr, c->srcW*srcBpp);
2218 srcPtr+= srcStride[0];
2219 dstPtr+= dstStride[0];
2220 }
221b804f
DB
2221 }
2222 }
2223 return srcSliceH;
0d9f3d85
A
2224}
2225
d4e24275 2226static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2227 int srcSliceH, uint8_t* dst[], int dstStride[])
2228{
221b804f
DB
2229
2230 rgb24toyv12(
2231 src[0],
2232 dst[0]+ srcSliceY *dstStride[0],
2233 dst[1]+(srcSliceY>>1)*dstStride[1],
2234 dst[2]+(srcSliceY>>1)*dstStride[2],
2235 c->srcW, srcSliceH,
2236 dstStride[0], dstStride[1], srcStride[0]);
08218e6d
CS
2237 if (dst[3])
2238 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
221b804f 2239 return srcSliceH;
ec22603f
MN
2240}
2241
d4e24275 2242static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
2243 int srcSliceH, uint8_t* dst[], int dstStride[])
2244{
221b804f
DB
2245 int i;
2246
2247 /* copy Y */
2248 if (srcStride[0]==dstStride[0] && srcStride[0] > 0)
2249 memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
dd68318c 2250 else {
221b804f
DB
2251 uint8_t *srcPtr= src[0];
2252 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2253
dd68318c 2254 for (i=0; i<srcSliceH; i++) {
221b804f
DB
2255 memcpy(dstPtr, srcPtr, c->srcW);
2256 srcPtr+= srcStride[0];
2257 dstPtr+= dstStride[0];
2258 }
2259 }
2260
dd68318c 2261 if (c->dstFormat==PIX_FMT_YUV420P || c->dstFormat==PIX_FMT_YUVA420P) {
5f9ae198
VS
2262 planar2x(src[1], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2263 srcSliceH >> 2, srcStride[1], dstStride[1]);
2264 planar2x(src[2], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2265 srcSliceH >> 2, srcStride[2], dstStride[2]);
dd68318c 2266 } else {
5f9ae198
VS
2267 planar2x(src[1], dst[2] + dstStride[2]*(srcSliceY >> 1), c->chrSrcW,
2268 srcSliceH >> 2, srcStride[1], dstStride[2]);
2269 planar2x(src[2], dst[1] + dstStride[1]*(srcSliceY >> 1), c->chrSrcW,
2270 srcSliceH >> 2, srcStride[2], dstStride[1]);
221b804f 2271 }
08218e6d
CS
2272 if (dst[3])
2273 fillPlane(dst[3], dstStride[3], c->srcW, srcSliceH, srcSliceY, 255);
221b804f 2274 return srcSliceH;
b241cbf2
MN
2275}
2276
b6654a54 2277/* unscaled copy like stuff (assumes nearly identical formats) */
2d35ae56 2278static int packedCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
bc5a0444
LB
2279 int srcSliceH, uint8_t* dst[], int dstStride[])
2280{
2281 if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
2282 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
dd68318c 2283 else {
bc5a0444
LB
2284 int i;
2285 uint8_t *srcPtr= src[0];
2286 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
2287 int length=0;
221b804f 2288
bc5a0444
LB
2289 /* universal length finder */
2290 while(length+c->srcW <= FFABS(dstStride[0])
2291 && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
fcc402b1 2292 assert(length!=0);
2d35ae56 2293
dd68318c 2294 for (i=0; i<srcSliceH; i++) {
bc5a0444
LB
2295 memcpy(dstPtr, srcPtr, length);
2296 srcPtr+= srcStride[0];
2297 dstPtr+= dstStride[0];
221b804f 2298 }
bc5a0444 2299 }
2d35ae56
LB
2300 return srcSliceH;
2301}
bc5a0444 2302
2d35ae56
LB
2303static int planarCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
2304 int srcSliceH, uint8_t* dst[], int dstStride[])
2305{
61a3f379 2306 int plane, i, j;
dd68318c 2307 for (plane=0; plane<4; plane++) {
6268f55b
CS
2308 int length= (plane==0 || plane==3) ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample);
2309 int y= (plane==0 || plane==3) ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
2310 int height= (plane==0 || plane==3) ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);
61a3f379
MN
2311 uint8_t *srcPtr= src[plane];
2312 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
2d35ae56 2313
20ddf5a8 2314 if (!dst[plane]) continue;
6f348086
RD
2315 // ignore palette for GRAY8
2316 if (plane == 1 && !dst[2]) continue;
dd68318c 2317 if (!src[plane] || (plane == 1 && !src[2])) {
61a3f379
MN
2318 if(is16BPS(c->dstFormat))
2319 length*=2;
6268f55b 2320 fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
dd68318c
RP
2321 } else {
2322 if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
61a3f379 2323 if (!isBE(c->srcFormat)) srcPtr++;
dd68318c 2324 for (i=0; i<height; i++) {
61a3f379
MN
2325 for (j=0; j<length; j++) dstPtr[j] = srcPtr[j<<1];
2326 srcPtr+= srcStride[plane];
2327 dstPtr+= dstStride[plane];
2328 }
dd68318c
RP
2329 } else if(!is16BPS(c->srcFormat) && is16BPS(c->dstFormat)) {
2330 for (i=0; i<height; i++) {
2331 for (j=0; j<length; j++) {
61a3f379
MN
2332 dstPtr[ j<<1 ] = srcPtr[j];
2333 dstPtr[(j<<1)+1] = srcPtr[j];
2334 }
2335 srcPtr+= srcStride[plane];
2336 dstPtr+= dstStride[plane];
2337 }
dd68318c
RP
2338 } else if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat)
2339 && isBE(c->srcFormat) != isBE(c->dstFormat)) {
61a3f379 2340
dd68318c 2341 for (i=0; i<height; i++) {
61a3f379
MN
2342 for (j=0; j<length; j++)
2343 ((uint16_t*)dstPtr)[j] = bswap_16(((uint16_t*)srcPtr)[j]);
2344 srcPtr+= srcStride[plane];
2345 dstPtr+= dstStride[plane];
2346 }
2347 } else if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
bc5a0444 2348 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
dd68318c 2349 else {
61a3f379
MN
2350 if(is16BPS(c->srcFormat) && is16BPS(c->dstFormat))
2351 length*=2;
dd68318c 2352 for (i=0; i<height; i++) {
bc5a0444
LB
2353 memcpy(dstPtr, srcPtr, length);
2354 srcPtr+= srcStride[plane];
2355 dstPtr+= dstStride[plane];
221b804f
DB
2356 }
2357 }
2358 }
bc5a0444 2359 }
221b804f 2360 return srcSliceH;
37079906 2361}
28bf81c9 2362
4884b9e5 2363
dd68318c
RP
2364static void getSubSampleFactors(int *h, int *v, int format)
2365{
2366 switch(format) {
221b804f
DB
2367 case PIX_FMT_UYVY422:
2368 case PIX_FMT_YUYV422:
2369 *h=1;
2370 *v=0;
2371 break;
2372 case PIX_FMT_YUV420P:
de1275d5
MN
2373 case PIX_FMT_YUV420PLE:
2374 case PIX_FMT_YUV420PBE:
79973335 2375 case PIX_FMT_YUVA420P:
221b804f
DB
2376 case PIX_FMT_GRAY16BE:
2377 case PIX_FMT_GRAY16LE:
2378 case PIX_FMT_GRAY8: //FIXME remove after different subsamplings are fully implemented
2379 case PIX_FMT_NV12:
2380 case PIX_FMT_NV21:
2381 *h=1;
2382 *v=1;
2383 break;
9ba7fe6d
2384 case PIX_FMT_YUV440P:
2385 *h=0;
2386 *v=1;
2387 break;
221b804f
DB
2388 case PIX_FMT_YUV410P:
2389 *h=2;
2390 *v=2;
2391 break;
2392 case PIX_FMT_YUV444P:
de1275d5
MN
2393 case PIX_FMT_YUV444PLE:
2394 case PIX_FMT_YUV444PBE:
221b804f
DB
2395 *h=0;
2396 *v=0;
2397 break;
2398 case PIX_FMT_YUV422P:
de1275d5
MN
2399 case PIX_FMT_YUV422PLE:
2400 case PIX_FMT_YUV422PBE:
221b804f
DB
2401 *h=1;
2402 *v=0;
2403 break;
2404 case PIX_FMT_YUV411P:
2405 *h=2;
2406 *v=0;
2407 break;
2408 default:
2409 *h=0;
2410 *v=0;
2411 break;
2412 }
c7a810cc
MN
2413}
2414
dd68318c
RP
2415static uint16_t roundToInt16(int64_t f)
2416{
221b804f
DB
2417 int r= (f + (1<<15))>>16;
2418 if (r<-0x7FFF) return 0x8000;
2419 else if (r> 0x7FFF) return 0x7FFF;
2420 else return r;
0481412a
MN
2421}
2422
dd68318c
RP
2423int sws_setColorspaceDetails(SwsContext *c, const int inv_table[4], int srcRange, const int table[4], int dstRange, int brightness, int contrast, int saturation)
2424{
221b804f
DB
2425 int64_t crv = inv_table[0];
2426 int64_t cbu = inv_table[1];
2427 int64_t cgu = -inv_table[2];
2428 int64_t cgv = -inv_table[3];
2429 int64_t cy = 1<<16;
2430 int64_t oy = 0;
2431
221b804f
DB
2432 memcpy(c->srcColorspaceTable, inv_table, sizeof(int)*4);
2433 memcpy(c->dstColorspaceTable, table, sizeof(int)*4);
2434
2435 c->brightness= brightness;
2436 c->contrast = contrast;
2437 c->saturation= saturation;
2438 c->srcRange = srcRange;
2439 c->dstRange = dstRange;
0c067dc3 2440 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
221b804f
DB
2441
2442 c->uOffset= 0x0400040004000400LL;
2443 c->vOffset= 0x0400040004000400LL;
2444
dd68318c 2445 if (!srcRange) {
221b804f
DB
2446 cy= (cy*255) / 219;
2447 oy= 16<<16;
dd68318c 2448 } else {
221b804f
DB
2449 crv= (crv*224) / 255;
2450 cbu= (cbu*224) / 255;
2451 cgu= (cgu*224) / 255;
2452 cgv= (cgv*224) / 255;
2453 }
0481412a 2454
221b804f
DB
2455 cy = (cy *contrast )>>16;
2456 crv= (crv*contrast * saturation)>>32;
2457 cbu= (cbu*contrast * saturation)>>32;
2458 cgu= (cgu*contrast * saturation)>>32;
2459 cgv= (cgv*contrast * saturation)>>32;
0481412a 2460
221b804f 2461 oy -= 256*brightness;
0481412a 2462
221b804f
DB
2463 c->yCoeff= roundToInt16(cy *8192) * 0x0001000100010001ULL;
2464 c->vrCoeff= roundToInt16(crv*8192) * 0x0001000100010001ULL;
2465 c->ubCoeff= roundToInt16(cbu*8192) * 0x0001000100010001ULL;
2466 c->vgCoeff= roundToInt16(cgv*8192) * 0x0001000100010001ULL;
2467 c->ugCoeff= roundToInt16(cgu*8192) * 0x0001000100010001ULL;
2468 c->yOffset= roundToInt16(oy * 8) * 0x0001000100010001ULL;
5427e242 2469
43c16478
MN
2470 c->yuv2rgb_y_coeff = (int16_t)roundToInt16(cy <<13);
2471 c->yuv2rgb_y_offset = (int16_t)roundToInt16(oy << 9);
2472 c->yuv2rgb_v2r_coeff= (int16_t)roundToInt16(crv<<13);
2473 c->yuv2rgb_v2g_coeff= (int16_t)roundToInt16(cgv<<13);
2474 c->yuv2rgb_u2g_coeff= (int16_t)roundToInt16(cgu<<13);
2475 c->yuv2rgb_u2b_coeff= (int16_t)roundToInt16(cbu<<13);
f0faee4c 2476
780daf2b 2477 ff_yuv2rgb_c_init_tables(c, inv_table, srcRange, brightness, contrast, saturation);
221b804f 2478 //FIXME factorize
a31de956 2479
6634d0ef 2480#ifdef COMPILE_ALTIVEC
221b804f 2481 if (c->flags & SWS_CPU_CAPS_ALTIVEC)
780daf2b 2482 ff_yuv2rgb_init_tables_altivec(c, inv_table, brightness, contrast, saturation);
6a4970ab 2483#endif
221b804f 2484 return 0;
5427e242
MN
2485}
2486
dd68318c
RP
2487int sws_getColorspaceDetails(SwsContext *c, int **inv_table, int *srcRange, int **table, int *dstRange, int *brightness, int *contrast, int *saturation)
2488{
221b804f 2489 if (isYUV(c->dstFormat) || isGray(c->dstFormat)) return -1;
5427e242 2490
221b804f
DB
2491 *inv_table = c->srcColorspaceTable;
2492 *table = c->dstColorspaceTable;
2493 *srcRange = c->srcRange;
2494 *dstRange = c->dstRange;
2495 *brightness= c->brightness;
2496 *contrast = c->contrast;
2497 *saturation= c->saturation;
6a4970ab 2498
221b804f 2499 return 0;
0481412a
MN
2500}
2501
13394e8c 2502static int handle_jpeg(enum PixelFormat *format)
44cdb423 2503{
221b804f 2504 switch (*format) {
9b734d44
RP
2505 case PIX_FMT_YUVJ420P:
2506 *format = PIX_FMT_YUV420P;
2507 return 1;
2508 case PIX_FMT_YUVJ422P:
2509 *format = PIX_FMT_YUV422P;
2510 return 1;
2511 case PIX_FMT_YUVJ444P:
2512 *format = PIX_FMT_YUV444P;
2513 return 1;
2514 case PIX_FMT_YUVJ440P:
2515 *format = PIX_FMT_YUV440P;
2516 return 1;
2517 default:
2518 return 0;
221b804f 2519 }
44cdb423
LA
2520}
2521
58e4b706 2522SwsContext *sws_getContext(int srcW, int srcH, enum PixelFormat srcFormat, int dstW, int dstH, enum PixelFormat dstFormat, int flags,
0020c54c
DB
2523 SwsFilter *srcFilter, SwsFilter *dstFilter, const double *param)
2524{
221b804f
DB
2525
2526 SwsContext *c;
2527 int i;
2528 int usesVFilter, usesHFilter;
2529 int unscaled, needsDither;
2530 int srcRange, dstRange;
2531 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL};
b63f641e 2532#if ARCH_X86
221b804f 2533 if (flags & SWS_CPU_CAPS_MMX)
7ad6469e 2534 __asm__ volatile("emms\n\t"::: "memory");
5cebb24b 2535#endif
516b1f82 2536
8b1a6441 2537#if !CONFIG_RUNTIME_CPUDETECT //ensure that the flags match the compiled variant if cpudetect is off
d3f3eea9 2538 flags &= ~(SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2|SWS_CPU_CAPS_3DNOW|SWS_CPU_CAPS_ALTIVEC|SWS_CPU_CAPS_BFIN);
94daf2e9 2539#if COMPILE_TEMPLATE_MMX2
221b804f 2540 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
94daf2e9 2541#elif COMPILE_TEMPLATE_AMD3DNOW
221b804f 2542 flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
94daf2e9 2543#elif COMPILE_TEMPLATE_MMX
221b804f 2544 flags |= SWS_CPU_CAPS_MMX;
94daf2e9 2545#elif COMPILE_TEMPLATE_ALTIVEC
221b804f 2546 flags |= SWS_CPU_CAPS_ALTIVEC;
b63f641e 2547#elif ARCH_BFIN
d3f3eea9 2548 flags |= SWS_CPU_CAPS_BFIN;
516b1f82 2549#endif
10a7d216 2550#endif /* CONFIG_RUNTIME_CPUDETECT */
221b804f 2551 if (clip_table[512] != 255) globalInit();
1b0a4572 2552 if (!rgb15to16) sws_rgb2rgb_init(flags);
221b804f
DB
2553
2554 unscaled = (srcW == dstW && srcH == dstH);
2555 needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
2556 && (fmt_depth(dstFormat))<24
2557 && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));
2558
2559 srcRange = handle_jpeg(&srcFormat);
2560 dstRange = handle_jpeg(&dstFormat);
2561
dd68318c 2562 if (!isSupportedIn(srcFormat)) {
0d6fd5ec 2563 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input pixel format\n", sws_format_name(srcFormat));
221b804f
DB
2564 return NULL;
2565 }
dd68318c 2566 if (!isSupportedOut(dstFormat)) {
0d6fd5ec 2567 av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as output pixel format\n", sws_format_name(dstFormat));
221b804f
DB
2568 return NULL;
2569 }
2570
010c00bc
MN
2571 i= flags & ( SWS_POINT
2572 |SWS_AREA
6afc7c19 2573 |SWS_BILINEAR
010c00bc
MN
2574 |SWS_FAST_BILINEAR
2575 |SWS_BICUBIC
2576 |SWS_X
2577 |SWS_GAUSS
2578 |SWS_LANCZOS
2579 |SWS_SINC
2580 |SWS_SPLINE
2581 |SWS_BICUBLIN);
dd68318c 2582 if(!i || (i & (i-1))) {
f40c7dbb 2583 av_log(NULL, AV_LOG_ERROR, "swScaler: Exactly one scaler algorithm must be chosen\n");
010c00bc
MN
2584 return NULL;
2585 }
010c00bc 2586
221b804f 2587 /* sanity check */
dd68318c 2588 if (srcW<4 || srcH<1 || dstW<8 || dstH<1) { //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
221b804f
DB
2589 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
2590 srcW, srcH, dstW, dstH);
2591 return NULL;
2592 }
dd68318c 2593 if(srcW > VOFW || dstW > VOFW) {
f40c7dbb 2594 av_log(NULL, AV_LOG_ERROR, "swScaler: Compile-time maximum width is "AV_STRINGIFY(VOFW)" change VOF/VOFW and recompile\n");
8b2fce0d
MN
2595 return NULL;
2596 }
221b804f
DB
2597
2598 if (!dstFilter) dstFilter= &dummyFilter;
2599 if (!srcFilter) srcFilter= &dummyFilter;
2600
2601 c= av_mallocz(sizeof(SwsContext));
2602
2603 c->av_class = &sws_context_class;
2604 c->srcW= srcW;
2605 c->srcH= srcH;
2606 c->dstW= dstW;
2607 c->dstH= dstH;
2608 c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
2609 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
2610 c->flags= flags;
2611 c->dstFormat= dstFormat;
2612 c->srcFormat= srcFormat;
2613 c->vRounder= 4* 0x0001000100010001ULL;
2614
2615 usesHFilter= usesVFilter= 0;
1b0a4572
BF
2616 if (dstFilter->lumV && dstFilter->lumV->length>1) usesVFilter=1;
2617 if (dstFilter->lumH && dstFilter->lumH->length>1) usesHFilter=1;
2618 if (dstFilter->chrV && dstFilter->chrV->length>1) usesVFilter=1;
2619 if (dstFilter->chrH && dstFilter->chrH->length>1) usesHFilter=1;
2620 if (srcFilter->lumV && srcFilter->lumV->length>1) usesVFilter=1;
2621 if (srcFilter->lumH && srcFilter->lumH->length>1) usesHFilter=1;
2622 if (srcFilter->chrV && srcFilter->chrV->length>1) usesVFilter=1;
2623 if (srcFilter->chrH && srcFilter->chrH->length>1) usesHFilter=1;
221b804f
DB
2624
2625 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
2626 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);
2627
f40c7dbb 2628 // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma interpolation
221b804f
DB
2629 if ((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;
2630
2631 // drop some chroma lines if the user wants it
2632 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
2633 c->chrSrcVSubSample+= c->vChrDrop;
2634
f40c7dbb 2635 // drop every other pixel for chroma calculation unless user wants full chroma
221b804f
DB
2636 if ((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)
2637 && srcFormat!=PIX_FMT_RGB8 && srcFormat!=PIX_FMT_BGR8
2638 && srcFormat!=PIX_FMT_RGB4 && srcFormat!=PIX_FMT_BGR4
dfb09bd1 2639 && srcFormat!=PIX_FMT_RGB4_BYTE && srcFormat!=PIX_FMT_BGR4_BYTE
2f60f629 2640 && ((dstW>>c->chrDstHSubSample) <= (srcW>>1) || (flags&(SWS_FAST_BILINEAR|SWS_POINT))))
221b804f
DB
2641 c->chrSrcHSubSample=1;
2642
dd68318c 2643 if (param) {
221b804f
DB
2644 c->param[0] = param[0];
2645 c->param[1] = param[1];
dd68318c 2646 } else {
221b804f
DB
2647 c->param[0] =
2648 c->param[1] = SWS_PARAM_DEFAULT;
2649 }
2650
221b804f
DB
2651 // Note the -((-x)>>y) is so that we always round toward +inf.
2652 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
2653 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
2654 c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
2655 c->chrDstH= -((-dstH) >> c->chrDstVSubSample);
2656
fa58ba15 2657 sws_setColorspaceDetails(c, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT], srcRange, ff_yuv2rgb_coeffs[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);
221b804f 2658
f40c7dbb 2659 /* unscaled special cases */
dd68318c 2660 if (unscaled && !usesHFilter && !usesVFilter && (srcRange == dstRange || isBGR(dstFormat) || isRGB(dstFormat))) {
221b804f 2661 /* yv12_to_nv12 */
dd68318c 2662 if ((srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) && (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21)) {
221b804f
DB
2663 c->swScale= PlanarToNV12Wrapper;
2664 }
221b804f 2665 /* yuv2bgr */
2c897342 2666 if ((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P || srcFormat==PIX_FMT_YUVA420P) && (isBGR(dstFormat) || isRGB(dstFormat))
dd68318c 2667 && !(flags & SWS_ACCURATE_RND) && !(dstH&1)) {
780daf2b 2668 c->swScale= ff_yuv2rgb_get_func_ptr(c);
221b804f 2669 }
6a4970ab 2670
dd68318c 2671 if (srcFormat==PIX_FMT_YUV410P && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_BITEXACT)) {
221b804f
DB
2672 c->swScale= yvu9toyv12Wrapper;
2673 }
2674
2675 /* bgr24toYV12 */
6268f55b 2676 if (srcFormat==PIX_FMT_BGR24 && (dstFormat==PIX_FMT_YUV420P || dstFormat==PIX_FMT_YUVA420P) && !(flags & SWS_ACCURATE_RND))
221b804f
DB
2677 c->swScale= bgr24toyv12Wrapper;
2678
f40c7dbb 2679 /* RGB/BGR -> RGB/BGR (no dither needed forms) */
221b804f
DB
2680 if ( (isBGR(srcFormat) || isRGB(srcFormat))
2681 && (isBGR(dstFormat) || isRGB(dstFormat))
2682 && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
2683 && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
2684 && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
2685 && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
2686 && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE
2687 && srcFormat != PIX_FMT_RGB4_BYTE && dstFormat != PIX_FMT_RGB4_BYTE
2688 && srcFormat != PIX_FMT_MONOBLACK && dstFormat != PIX_FMT_MONOBLACK
ec1bca2a 2689 && srcFormat != PIX_FMT_MONOWHITE && dstFormat != PIX_FMT_MONOWHITE
9990e426
MN
2690 && dstFormat != PIX_FMT_RGB32_1
2691 && dstFormat != PIX_FMT_BGR32_1
fa8c408b
VS
2692 && srcFormat != PIX_FMT_RGB48LE && dstFormat != PIX_FMT_RGB48LE
2693 && srcFormat != PIX_FMT_RGB48BE && dstFormat != PIX_FMT_RGB48BE
736143c8 2694 && (!needsDither || (c->flags&(SWS_FAST_BILINEAR|SWS_POINT))))
221b804f
DB
2695 c->swScale= rgb2rgbWrapper;
2696
49004617 2697 if ((usePal(srcFormat) && (
76e9fd01 2698 dstFormat == PIX_FMT_RGB32 ||
522ce957 2699 dstFormat == PIX_FMT_RGB32_1 ||
76e9fd01
VS
2700 dstFormat == PIX_FMT_RGB24 ||
2701 dstFormat == PIX_FMT_BGR32 ||
522ce957 2702 dstFormat == PIX_FMT_BGR32_1 ||
49004617
VS
2703 dstFormat == PIX_FMT_BGR24)))
2704 c->swScale= pal2rgbWrapper;
2705
dd68318c 2706 if (srcFormat == PIX_FMT_YUV422P) {
a6100f39
BC
2707 if (dstFormat == PIX_FMT_YUYV422)
2708 c->swScale= YUV422PToYuy2Wrapper;
2709 else if (dstFormat == PIX_FMT_UYVY422)
2710 c->swScale= YUV422PToUyvyWrapper;
2711 }
2712
221b804f 2713 /* LQ converters if -sws 0 or -sws 4*/
dd68318c 2714 if (c->flags&(SWS_FAST_BILINEAR|SWS_POINT)) {
221b804f 2715 /* yv12_to_yuy2 */
dd68318c 2716 if (srcFormat == PIX_FMT_YUV420P || srcFormat == PIX_FMT_YUVA420P) {
221b804f
DB
2717 if (dstFormat == PIX_FMT_YUYV422)
2718 c->swScale= PlanarToYuy2Wrapper;
3b38f1c6 2719 else if (dstFormat == PIX_FMT_UYVY422)
221b804f
DB
2720 c->swScale= PlanarToUyvyWrapper;
2721 }
2722 }
4626ee1a 2723 if(srcFormat == PIX_FMT_YUYV422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
ff8084cf 2724 c->swScale= YUYV2YUV420Wrapper;
4626ee1a 2725 if(srcFormat == PIX_FMT_UYVY422 && (dstFormat == PIX_FMT_YUV420P || dstFormat == PIX_FMT_YUVA420P))
ff8084cf 2726 c->swScale= UYVY2YUV420Wrapper;
0411072e
MN
2727 if(srcFormat == PIX_FMT_YUYV422 && dstFormat == PIX_FMT_YUV422P)
2728 c->swScale= YUYV2YUV422Wrapper;
2729 if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
2730 c->swScale= UYVY2YUV422Wrapper;
ec22603f 2731
6634d0ef 2732#ifdef COMPILE_ALTIVEC
221b804f 2733 if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
12794f73 2734 !(c->flags & SWS_BITEXACT) &&
3b38f1c6 2735 srcFormat == PIX_FMT_YUV420P) {
221b804f
DB
2736 // unscaled YV12 -> packed YUV, we want speed
2737 if (dstFormat == PIX_FMT_YUYV422)
2738 c->swScale= yv12toyuy2_unscaled_altivec;
3b38f1c6 2739 else if (dstFormat == PIX_FMT_UYVY422)
221b804f
DB
2740 c->swScale= yv12touyvy_unscaled_altivec;
2741 }
b71cf33c
RD
2742#endif
2743
221b804f
DB
2744 /* simple copy */
2745 if ( srcFormat == dstFormat
2c897342 2746 || (srcFormat == PIX_FMT_YUVA420P && dstFormat == PIX_FMT_YUV420P)
6268f55b 2747 || (srcFormat == PIX_FMT_YUV420P && dstFormat == PIX_FMT_YUVA420P)
221b804f 2748 || (isPlanarYUV(srcFormat) && isGray(dstFormat))
61a3f379 2749 || (isPlanarYUV(dstFormat) && isGray(srcFormat))
de1275d5
MN
2750 || (isGray(dstFormat) && isGray(srcFormat))
2751 || (isPlanarYUV(srcFormat) && isPlanarYUV(dstFormat)
2752 && c->chrDstHSubSample == c->chrSrcHSubSample
062394e0
GL
2753 && c->chrDstVSubSample == c->chrSrcVSubSample
2754 && dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21
2755 && srcFormat != PIX_FMT_NV12 && srcFormat != PIX_FMT_NV21))
221b804f 2756 {
2d35ae56
LB
2757 if (isPacked(c->srcFormat))
2758 c->swScale= packedCopy;
2759 else /* Planar YUV or gray */
2760 c->swScale= planarCopy;
221b804f 2761 }
b63f641e 2762#if ARCH_BFIN
1ebbfe15
MH
2763 if (flags & SWS_CPU_CAPS_BFIN)
2764 ff_bfin_get_unscaled_swscale (c);
2765#endif
2766
dd68318c 2767 if (c->swScale) {
221b804f 2768 if (flags&SWS_PRINT_INFO)
4b0c30b7 2769 av_log(c, AV_LOG_INFO, "using unscaled %s -> %s special converter\n",
9b734d44 2770 sws_format_name(srcFormat), sws_format_name(dstFormat));
221b804f
DB
2771 return c;
2772 }
2773 }
2774
dd68318c 2775 if (flags & SWS_CPU_CAPS_MMX2) {
221b804f 2776 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
dd68318c 2777 if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
221b804f 2778 if (flags&SWS_PRINT_INFO)
f40c7dbb 2779 av_log(c, AV_LOG_INFO, "output width is not a multiple of 32 -> no MMX2 scaler\n");
221b804f
DB
2780 }
2781 if (usesHFilter) c->canMMX2BeUsed=0;
2782 }
2783 else
2784 c->canMMX2BeUsed=0;
2785
2786 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
2787 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;
2788
2789 // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
2790 // but only for the FAST_BILINEAR mode otherwise do correct scaling
2791 // n-2 is the last chrominance sample available
86bdf3fd 2792 // this is not perfect, but no one should notice the difference, the more correct variant
221b804f
DB
2793 // would be like the vertical one, but that would require some special code for the
2794 // first and last pixel
dd68318c
RP
2795 if (flags&SWS_FAST_BILINEAR) {
2796 if (c->canMMX2BeUsed) {
221b804f
DB
2797 c->lumXInc+= 20;
2798 c->chrXInc+= 20;
2799 }
f40c7dbb 2800 //we don't use the x86 asm scaler if MMX is available
dd68318c 2801 else if (flags & SWS_CPU_CAPS_MMX) {
221b804f
DB
2802 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
2803 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
2804 }
2805 }
2806
2807 /* precalculate horizontal scaler filter coefficients */
2808 {
2809 const int filterAlign=
2810 (flags & SWS_CPU_CAPS_MMX) ? 4 :
2811 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2812 1;
2813
2814 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
2815 srcW , dstW, filterAlign, 1<<14,
2816 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
2817 srcFilter->lumH, dstFilter->lumH, c->param);
2818 initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
2819 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
2820 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2821 srcFilter->chrH, dstFilter->chrH, c->param);
28bf81c9 2822
17c613ef 2823#if defined(COMPILE_MMX2)
77a416e8 2824// can't downscale !!!
dd68318c 2825 if (c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) {
48f4c612
RP
2826 c->lumMmx2FilterCodeSize = initMMX2HScaler( dstW, c->lumXInc, NULL, NULL, NULL, 8);
2827 c->chrMmx2FilterCodeSize = initMMX2HScaler(c->chrDstW, c->chrXInc, NULL, NULL, NULL, 4);
2828
113ef149 2829#ifdef MAP_ANONYMOUS
48f4c612
RP
2830 c->lumMmx2FilterCode = mmap(NULL, c->lumMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
2831 c->chrMmx2FilterCode = mmap(NULL, c->chrMmx2FilterCodeSize, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANONYMOUS, 0, 0);
dd35beb2 2832#elif HAVE_VIRTUALALLOC
48f4c612
RP
2833 c->lumMmx2FilterCode = VirtualAlloc(NULL, c->lumMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
2834 c->chrMmx2FilterCode = VirtualAlloc(NULL, c->chrMmx2FilterCodeSize, MEM_COMMIT, PAGE_EXECUTE_READWRITE);
38d5c282 2835#else
48f4c612
RP
2836 c->lumMmx2FilterCode = av_malloc(c->lumMmx2FilterCodeSize);
2837 c->chrMmx2FilterCode = av_malloc(c->chrMmx2FilterCodeSize);
38d5c282
AJ
2838#endif
2839
221b804f
DB
2840 c->lumMmx2Filter = av_malloc((dstW /8+8)*sizeof(int16_t));
2841 c->chrMmx2Filter = av_malloc((c->chrDstW /4+8)*sizeof(int16_t));
2842 c->lumMmx2FilterPos= av_malloc((dstW /2/8+8)*sizeof(int32_t));
2843 c->chrMmx2FilterPos= av_malloc((c->chrDstW/2/4+8)*sizeof(int32_t));
b7dc6f66 2844
bcdedf67
RP
2845 initMMX2HScaler( dstW, c->lumXInc, c->lumMmx2FilterCode, c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
2846 initMMX2HScaler(c->chrDstW, c->chrXInc, c->chrMmx2FilterCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
07287788
RP
2847
2848#ifdef MAP_ANONYMOUS
48f4c612
RP
2849 mprotect(c->lumMmx2FilterCode, c->lumMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
2850 mprotect(c->chrMmx2FilterCode, c->chrMmx2FilterCodeSize, PROT_EXEC | PROT_READ);
07287788 2851#endif
221b804f 2852 }
17c613ef 2853#endif /* defined(COMPILE_MMX2) */
f40c7dbb 2854 } // initialize horizontal stuff
28bf81c9
MN
2855
2856
2857
221b804f
DB
2858 /* precalculate vertical scaler filter coefficients */
2859 {
2860 const int filterAlign=
2861 (flags & SWS_CPU_CAPS_MMX) && (flags & SWS_ACCURATE_RND) ? 2 :
2862 (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
2863 1;
8c266f0c 2864
221b804f 2865 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
5fe4aad7 2866 srcH , dstH, filterAlign, (1<<12),
221b804f
DB
2867 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags,
2868 srcFilter->lumV, dstFilter->lumV, c->param);
2869 initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
5fe4aad7 2870 c->chrSrcH, c->chrDstH, filterAlign, (1<<12),
221b804f
DB
2871 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
2872 srcFilter->chrV, dstFilter->chrV, c->param);
d33d485e 2873
184410e6 2874#ifdef COMPILE_ALTIVEC
221b804f
DB
2875 c->vYCoeffsBank = av_malloc(sizeof (vector signed short)*c->vLumFilterSize*c->dstH);
2876 c->vCCoeffsBank = av_malloc(sizeof (vector signed short)*c->vChrFilterSize*c->chrDstH);
2877
2878 for (i=0;i<c->vLumFilterSize*c->dstH;i++) {
2879 int j;
2880 short *p = (short *)&c->vYCoeffsBank[i];
2881 for (j=0;j<8;j++)
2882 p[j] = c->vLumFilter[i];
2883 }
2884
2885 for (i=0;i<c->vChrFilterSize*c->chrDstH;i++) {
2886 int j;
2887 short *p = (short *)&c->vCCoeffsBank[i];
2888 for (j=0;j<8;j++)
2889 p[j] = c->vChrFilter[i];
2890 }
d33d485e 2891#endif
221b804f
DB
2892 }
2893
f40c7dbb 2894 // calculate buffer sizes so that they won't run out while handling these damn slices
221b804f
DB
2895 c->vLumBufSize= c->vLumFilterSize;
2896 c->vChrBufSize= c->vChrFilterSize;
dd68318c 2897 for (i=0; i<dstH; i++) {
221b804f
DB
2898 int chrI= i*c->chrDstH / dstH;
2899 int nextSlice= FFMAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1,
2900 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));
2901
2902 nextSlice>>= c->chrSrcVSubSample;
2903 nextSlice<<= c->chrSrcVSubSample;
2904 if (c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice)
e5091488 2905 c->vLumBufSize= nextSlice - c->vLumFilterPos[i];
221b804f
DB
2906 if (c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
2907 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
2908 }
2909
2910 // allocate pixbufs (we use dynamic allocation because otherwise we would need to
0607b090 2911 // allocate several megabytes to handle all possible cases)
221b804f
DB
2912 c->lumPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
2913 c->chrPixBuf= av_malloc(c->vChrBufSize*2*sizeof(int16_t*));
6858492e
CS
2914 if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat) && isALPHA(c->dstFormat))
2915 c->alpPixBuf= av_malloc(c->vLumBufSize*2*sizeof(int16_t*));
f40c7dbb 2916 //Note we need at least one pixel more at the end because of the MMX code (just in case someone wanna replace the 4000/8000)
221b804f
DB
2917 /* align at 16 bytes for AltiVec */
2918 for (i=0; i<c->vLumBufSize; i++)
8b2fce0d 2919 c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
221b804f 2920 for (i=0; i<c->vChrBufSize; i++)
8b2fce0d 2921 c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= av_malloc((VOF+1)*2);
6858492e
CS
2922 if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf)
2923 for (i=0; i<c->vLumBufSize; i++)
2924 c->alpPixBuf[i]= c->alpPixBuf[i+c->vLumBufSize]= av_mallocz(VOF+1);
221b804f
DB
2925
2926 //try to avoid drawing green stuff between the right end and the stride end
8b2fce0d
MN
2927 for (i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, (VOF+1)*2);
2928
5a352b14 2929 assert(2*VOFW == VOF);
221b804f 2930
fcc402b1 2931 assert(c->chrDstH <= dstH);
221b804f 2932
dd68318c 2933 if (flags&SWS_PRINT_INFO) {
28bf81c9 2934#ifdef DITHER1XBPP
3e62b7e3 2935 const char *dither= " dithered";
5521b193 2936#else
3e62b7e3 2937 const char *dither= "";
28bf81c9 2938#endif
221b804f 2939 if (flags&SWS_FAST_BILINEAR)
4b0c30b7 2940 av_log(c, AV_LOG_INFO, "FAST_BILINEAR scaler, ");
221b804f 2941 else if (flags&SWS_BILINEAR)
4b0c30b7 2942 av_log(c, AV_LOG_INFO, "BILINEAR scaler, ");
221b804f 2943 else if (flags&SWS_BICUBIC)
4b0c30b7 2944 av_log(c, AV_LOG_INFO, "BICUBIC scaler, ");
221b804f 2945 else if (flags&SWS_X)
4b0c30b7 2946 av_log(c, AV_LOG_INFO, "Experimental scaler, ");
221b804f 2947 else if (flags&SWS_POINT)
4b0c30b7 2948 av_log(c, AV_LOG_INFO, "Nearest Neighbor / POINT scaler, ");
221b804f 2949 else if (flags&SWS_AREA)
4b0c30b7 2950 av_log(c, AV_LOG_INFO, "Area Averageing scaler, ");
221b804f 2951 else if (flags&SWS_BICUBLIN)
4b0c30b7 2952 av_log(c, AV_LOG_INFO, "luma BICUBIC / chroma BILINEAR scaler, ");
221b804f 2953 else if (flags&SWS_GAUSS)
4b0c30b7 2954 av_log(c, AV_LOG_INFO, "Gaussian scaler, ");
221b804f 2955 else if (flags&SWS_SINC)
4b0c30b7 2956 av_log(c, AV_LOG_INFO, "Sinc scaler, ");
221b804f 2957 else if (flags&SWS_LANCZOS)
4b0c30b7 2958 av_log(c, AV_LOG_INFO, "Lanczos scaler, ");
221b804f 2959 else if (flags&SWS_SPLINE)
4b0c30b7 2960 av_log(c, AV_LOG_INFO, "Bicubic spline scaler, ");
221b804f 2961 else
4b0c30b7 2962 av_log(c, AV_LOG_INFO, "ehh flags invalid?! ");
221b804f
DB
2963
2964 if (dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
2965 av_log(c, AV_LOG_INFO, "from %s to%s %s ",
2966 sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
2967 else
2968 av_log(c, AV_LOG_INFO, "from %s to %s ",
2969 sws_format_name(srcFormat), sws_format_name(dstFormat));
2970
2971 if (flags & SWS_CPU_CAPS_MMX2)
2972 av_log(c, AV_LOG_INFO, "using MMX2\n");
2973 else if (flags & SWS_CPU_CAPS_3DNOW)
2974 av_log(c, AV_LOG_INFO, "using 3DNOW\n");
2975 else if (flags & SWS_CPU_CAPS_MMX)
2976 av_log(c, AV_LOG_INFO, "using MMX\n");
2977 else if (flags & SWS_CPU_CAPS_ALTIVEC)
2978 av_log(c, AV_LOG_INFO, "using AltiVec\n");
2979 else
2980 av_log(c, AV_LOG_INFO, "using C\n");
2981 }
2982
dd68318c
RP
2983 if (flags & SWS_PRINT_INFO) {
2984 if (flags & SWS_CPU_CAPS_MMX) {
221b804f 2985 if (c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
4b0c30b7 2986 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
dd68318c 2987 else {
221b804f 2988 if (c->hLumFilterSize==4)
4b0c30b7 2989 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal luminance scaling\n");
221b804f 2990 else if (c->hLumFilterSize==8)
4b0c30b7 2991 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal luminance scaling\n");
221b804f 2992 else
4b0c30b7 2993 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal luminance scaling\n");
221b804f
DB
2994
2995 if (c->hChrFilterSize==4)
4b0c30b7 2996 av_log(c, AV_LOG_VERBOSE, "using 4-tap MMX scaler for horizontal chrominance scaling\n");
221b804f 2997 else if (c->hChrFilterSize==8)
4b0c30b7 2998 av_log(c, AV_LOG_VERBOSE, "using 8-tap MMX scaler for horizontal chrominance scaling\n");
221b804f 2999 else
4b0c30b7 3000 av_log(c, AV_LOG_VERBOSE, "using n-tap MMX scaler for horizontal chrominance scaling\n");
221b804f 3001 }
dd68318c 3002 } else {
b63f641e 3003#if ARCH_X86
f40c7dbb 3004 av_log(c, AV_LOG_VERBOSE, "using x86 asm scaler for horizontal scaling\n");
28bf81c9 3005#else
221b804f 3006 if (flags & SWS_FAST_BILINEAR)
4b0c30b7 3007 av_log(c, AV_LOG_VERBOSE, "using FAST_BILINEAR C scaler for horizontal scaling\n");
221b804f 3008 else
4b0c30b7 3009 av_log(c, AV_LOG_VERBOSE, "using C scaler for horizontal scaling\n");
28bf81c9 3010#endif
221b804f 3011 }
dd68318c 3012 if (isPlanarYUV(dstFormat)) {
221b804f 3013 if (c->vLumFilterSize==1)
4b0c30b7 3014 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3015 else
4b0c30b7 3016 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
dd68318c 3017 } else {
221b804f 3018 if (c->vLumFilterSize==1 && c->vChrFilterSize==2)
4b0c30b7
BC
3019 av_log(c, AV_LOG_VERBOSE, "using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
3020 " 2-tap scaler for vertical chrominance scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3021 else if (c->vLumFilterSize==2 && c->vChrFilterSize==2)
4b0c30b7 3022 av_log(c, AV_LOG_VERBOSE, "using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3023 else
4b0c30b7 3024 av_log(c, AV_LOG_VERBOSE, "using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f
DB
3025 }
3026
3027 if (dstFormat==PIX_FMT_BGR24)
f40c7dbb 3028 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR24 converter\n",
221b804f
DB
3029 (flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
3030 else if (dstFormat==PIX_FMT_RGB32)
f40c7dbb 3031 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR32 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3032 else if (dstFormat==PIX_FMT_BGR565)
f40c7dbb 3033 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR16 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3034 else if (dstFormat==PIX_FMT_BGR555)
f40c7dbb 3035 av_log(c, AV_LOG_VERBOSE, "using %s YV12->BGR15 converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
221b804f 3036
4b0c30b7 3037 av_log(c, AV_LOG_VERBOSE, "%dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
221b804f 3038 }
dd68318c 3039 if (flags & SWS_PRINT_INFO) {
f40c7dbb 3040 av_log(c, AV_LOG_DEBUG, "lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
221b804f 3041 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
f40c7dbb 3042 av_log(c, AV_LOG_DEBUG, "chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
221b804f
DB
3043 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
3044 }
3045
40fa5140 3046 c->swScale= getSwsFunc(c);
221b804f 3047 return c;
28bf81c9
MN
3048}
3049
dd68318c
RP
3050static void reset_ptr(uint8_t* src[], int format)
3051{
b87fae9f
MN
3052 if(!isALPHA(format))
3053 src[3]=NULL;
dd68318c 3054 if(!isPlanarYUV(format)) {
b87fae9f
MN
3055 src[3]=src[2]=NULL;
3056 if( format != PIX_FMT_PAL8
3057 && format != PIX_FMT_RGB8
3058 && format != PIX_FMT_BGR8
3059 && format != PIX_FMT_RGB4_BYTE
3060 && format != PIX_FMT_BGR4_BYTE
3061 )
3062 src[1]= NULL;
3063 }
3064}
3065
28bf81c9 3066/**
56b69633 3067 * swscale wrapper, so we don't need to export the SwsContext.
f40c7dbb 3068 * Assumes planar YUV to be in YUV order instead of YVU.
fccb9b2b 3069 */
703b56fb 3070int sws_scale(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
3071 int srcSliceH, uint8_t* dst[], int dstStride[])
3072{
221b804f 3073 int i;
6208d676 3074 uint8_t* src2[4]= {src[0], src[1], src[2], src[3]};
b87fae9f 3075 uint8_t* dst2[4]= {dst[0], dst[1], dst[2], dst[3]};
14623020 3076
221b804f 3077 if (c->sliceDir == 0 && srcSliceY != 0 && srcSliceY + srcSliceH != c->srcH) {
3f0bc115 3078 av_log(c, AV_LOG_ERROR, "Slices start in the middle!\n");
221b804f
DB
3079 return 0;
3080 }
3081 if (c->sliceDir == 0) {
3082 if (srcSliceY == 0) c->sliceDir = 1; else c->sliceDir = -1;
3083 }
3084
dd68318c
RP
3085 if (usePal(c->srcFormat)) {
3086 for (i=0; i<256; i++) {
14623020 3087 int p, r, g, b,y,u,v;
dd68318c 3088 if(c->srcFormat == PIX_FMT_PAL8) {
14623020
MN
3089 p=((uint32_t*)(src[1]))[i];
3090 r= (p>>16)&0xFF;
3091 g= (p>> 8)&0xFF;
3092 b= p &0xFF;
dd68318c 3093 } else if(c->srcFormat == PIX_FMT_RGB8) {
14623020
MN
3094 r= (i>>5 )*36;
3095 g= ((i>>2)&7)*36;
3096 b= (i&3 )*85;
dd68318c 3097 } else if(c->srcFormat == PIX_FMT_BGR8) {
14623020
MN
3098 b= (i>>6 )*85;
3099 g= ((i>>3)&7)*36;
3100 r= (i&7 )*36;
dd68318c 3101 } else if(c->srcFormat == PIX_FMT_RGB4_BYTE) {
14623020
MN
3102 r= (i>>3 )*255;
3103 g= ((i>>1)&3)*85;
3104 b= (i&1 )*255;
dd68318c 3105 } else {
75add6fb 3106 assert(c->srcFormat == PIX_FMT_BGR4_BYTE);
14623020
MN
3107 b= (i>>3 )*255;
3108 g= ((i>>1)&3)*85;
3109 r= (i&1 )*255;
3110 }
cbcb408f
MN
3111 y= av_clip_uint8((RY*r + GY*g + BY*b + ( 33<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3112 u= av_clip_uint8((RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
3113 v= av_clip_uint8((RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1)))>>RGB2YUV_SHIFT);
49004617 3114 c->pal_yuv[i]= y + (u<<8) + (v<<16);
522ce957
VS
3115
3116
3117 switch(c->dstFormat) {
3118 case PIX_FMT_BGR32:
a898cdc9 3119#if !HAVE_BIGENDIAN
522ce957
VS
3120 case PIX_FMT_RGB24:
3121#endif
3122 c->pal_rgb[i]= r + (g<<8) + (b<<16);
3123 break;
3124 case PIX_FMT_BGR32_1:
a898cdc9 3125#if HAVE_BIGENDIAN
522ce957
VS
3126 case PIX_FMT_BGR24:
3127#endif
3128 c->pal_rgb[i]= (r + (g<<8) + (b<<16)) << 8;
3129 break;
3130 case PIX_FMT_RGB32_1:
a898cdc9 3131#if HAVE_BIGENDIAN
522ce957
VS
3132 case PIX_FMT_RGB24:
3133#endif
3134 c->pal_rgb[i]= (b + (g<<8) + (r<<16)) << 8;
3135 break;
3136 case PIX_FMT_RGB32:
a898cdc9 3137#if !HAVE_BIGENDIAN
522ce957
VS
3138 case PIX_FMT_BGR24:
3139#endif
3140 default:
3141 c->pal_rgb[i]= b + (g<<8) + (r<<16);
3142 }
21c08a3f 3143 }
221b804f 3144 }
21c08a3f 3145
221b804f
DB
3146 // copy strides, so they can safely be modified
3147 if (c->sliceDir == 1) {
3148 // slices go from top to bottom
6208d676
CS
3149 int srcStride2[4]= {srcStride[0], srcStride[1], srcStride[2], srcStride[3]};
3150 int dstStride2[4]= {dstStride[0], dstStride[1], dstStride[2], dstStride[3]};
b87fae9f
MN
3151
3152 reset_ptr(src2, c->srcFormat);
3153 reset_ptr(dst2, c->dstFormat);
3154
3155 return c->swScale(c, src2, srcStride2, srcSliceY, srcSliceH, dst2, dstStride2);
221b804f
DB
3156 } else {
3157 // slices go from bottom to top => we flip the image internally
6208d676
CS
3158 int srcStride2[4]= {-srcStride[0], -srcStride[1], -srcStride[2], -srcStride[3]};
3159 int dstStride2[4]= {-dstStride[0], -dstStride[1], -dstStride[2], -dstStride[3]};
221b804f
DB
3160
3161 src2[0] += (srcSliceH-1)*srcStride[0];
49004617 3162 if (!usePal(c->srcFormat))
221b804f
DB
3163 src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
3164 src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
6208d676 3165 src2[3] += (srcSliceH-1)*srcStride[3];
b87fae9f
MN
3166 dst2[0] += ( c->dstH -1)*dstStride[0];
3167 dst2[1] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[1];
3168 dst2[2] += ((c->dstH>>c->chrDstVSubSample)-1)*dstStride[2];
3169 dst2[3] += ( c->dstH -1)*dstStride[3];
3170
3171 reset_ptr(src2, c->srcFormat);
3172 reset_ptr(dst2, c->dstFormat);
221b804f
DB
3173
3174 return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
3175 }
fccb9b2b
MN
3176}
3177
c099766f 3178#if LIBSWSCALE_VERSION_MAJOR < 1
703b56fb 3179int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
dd68318c
RP
3180 int srcSliceH, uint8_t* dst[], int dstStride[])
3181{
221b804f 3182 return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
d4e24275 3183}
c099766f 3184#endif
d4e24275 3185
6a4970ab 3186SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
221b804f
DB
3187 float lumaSharpen, float chromaSharpen,
3188 float chromaHShift, float chromaVShift,
3189 int verbose)
e21206a8 3190{
221b804f
DB
3191 SwsFilter *filter= av_malloc(sizeof(SwsFilter));
3192
dd68318c 3193 if (lumaGBlur!=0.0) {
221b804f
DB
3194 filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
3195 filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);