Mercurial > mplayer.hg
annotate postproc/swscale.c @ 6532:9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
disabled broken (=new) converters (they cause random sig11)
various bugfixes
author | michael |
---|---|
date | Sun, 23 Jun 2002 18:51:43 +0000 |
parents | f9a1870bc4a5 |
children | 29f94210a67f |
rev | line source |
---|---|
4295 | 1 /* |
2 Copyright (C) 2001-2002 Michael Niedermayer <michaelni@gmx.at> | |
2216 | 3 |
4295 | 4 This program is free software; you can redistribute it and/or modify |
5 it under the terms of the GNU General Public License as published by | |
6 the Free Software Foundation; either version 2 of the License, or | |
7 (at your option) any later version. | |
2216 | 8 |
4295 | 9 This program is distributed in the hope that it will be useful, |
10 but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 GNU General Public License for more details. | |
13 | |
14 You should have received a copy of the GNU General Public License | |
15 along with this program; if not, write to the Free Software | |
16 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | |
17 */ | |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
18 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
19 /* |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
20 supported Input formats: YV12, I420, IYUV, YUY2, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8, Y800, YVU9 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
21 supported output formats: YV12, I420, IYUV, BGR15, BGR16, BGR24, BGR32, Y8, Y800, YVU9 |
4419 | 22 BGR15/16 support dithering |
4558 | 23 |
24 unscaled special converters | |
25 YV12/I420/IYUV -> BGR15/BGR16/BGR24/BGR32 | |
26 YV12/I420/IYUV -> YV12/I420/IYUV | |
27 YUY2/BGR15/BGR16/BGR24/BGR32/RGB24/RGB32 -> same format | |
4567 | 28 BGR24 -> BGR32 & RGB24 -> RGB32 |
29 BGR32 -> BGR24 & RGB32 -> RGB24 | |
4581 | 30 BGR15 -> BGR16 |
4567 | 31 */ |
32 | |
33 /* | |
34 tested special converters | |
35 YV12/I420 -> BGR16 | |
36 YV12 -> YV12 | |
4581 | 37 BGR15 -> BGR16 |
4791 | 38 BGR16 -> BGR16 |
4567 | 39 |
40 untested special converters | |
4791 | 41 YV12/I420 -> BGR15/BGR24/BGR32 (its the yuv2rgb stuff, so it should be ok) |
42 YV12/I420 -> YV12/I420 | |
43 YUY2/BGR15/BGR24/BGR32/RGB24/RGB32 -> same format | |
4567 | 44 BGR24 -> BGR32 & RGB24 -> RGB32 |
45 BGR32 -> BGR24 & RGB32 -> RGB24 | |
4623 | 46 BGR24 -> YV12 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
47 */ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
48 |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
49 #include <inttypes.h> |
2476 | 50 #include <string.h> |
3272 | 51 #include <math.h> |
3344 | 52 #include <stdio.h> |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
53 #include "../config.h" |
4248 | 54 #include "../mangle.h" |
4806 | 55 #include <assert.h> |
3344 | 56 #ifdef HAVE_MALLOC_H |
57 #include <malloc.h> | |
6057
31e465fda59c
various openbsd and general warning fixes - patch by Bj«Órn Sandell <biorn@dce.chalmers.se>
arpi
parents:
5937
diff
changeset
|
58 #else |
31e465fda59c
various openbsd and general warning fixes - patch by Bj«Órn Sandell <biorn@dce.chalmers.se>
arpi
parents:
5937
diff
changeset
|
59 #include <stdlib.h> |
3344 | 60 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
61 #include "swscale.h" |
3126 | 62 #include "../cpudetect.h" |
4558 | 63 #include "../bswap.h" |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
64 #include "../libvo/img_format.h" |
4554 | 65 #include "rgb2rgb.h" |
4586 | 66 #include "../libvo/fastmemcpy.h" |
5937 | 67 #include "../mp_msg.h" |
6492 | 68 |
69 #define MSG_WARN(args...) mp_msg(MSGT_SWS,MSGL_WARN, ##args ) | |
70 #define MSG_FATAL(args...) mp_msg(MSGT_SWS,MSGL_FATAL, ##args ) | |
71 #define MSG_ERR(args...) mp_msg(MSGT_SWS,MSGL_ERR, ##args ) | |
72 #define MSG_V(args...) mp_msg(MSGT_SWS,MSGL_V, ##args ) | |
73 #define MSG_DBG2(args...) mp_msg(MSGT_SWS,MSGL_DBG2, ##args ) | |
74 #define MSG_INFO(args...) mp_msg(MSGT_SWS,MSGL_INFO, ##args ) | |
75 | |
2540 | 76 #undef MOVNTQ |
2680 | 77 #undef PAVGB |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
78 |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
79 //#undef HAVE_MMX2 |
4281 | 80 //#define HAVE_3DNOW |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
81 //#undef HAVE_MMX |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
82 //#undef ARCH_X86 |
4794 | 83 //#define WORDS_BIGENDIAN |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
84 #define DITHER1XBPP |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
85 |
4612 | 86 #define FAST_BGR2YV12 // use 7 bit coeffs instead of 15bit |
87 | |
4467 | 88 #define RET 0xC3 //near return opcode for X86 |
3344 | 89 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
90 #ifdef MP_DEBUG |
4806 | 91 #define ASSERT(x) assert(x); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
92 #else |
3344 | 93 #define ASSERT(x) ; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
94 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
95 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
96 #ifdef M_PI |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
97 #define PI M_PI |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
98 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
99 #define PI 3.14159265358979323846 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
100 #endif |
3344 | 101 |
4419 | 102 //FIXME replace this with something faster |
6503 | 103 #define isPlanarYUV(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YVU9) |
104 #define isYUV(x) ((x)==IMGFMT_YUY2 || isPlanarYUV(x)) | |
105 #define isGray(x) ((x)==IMGFMT_Y800) | |
4481 | 106 #define isSupportedIn(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YUY2 \ |
4580 | 107 || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\ |
4558 | 108 || (x)==IMGFMT_RGB32|| (x)==IMGFMT_RGB24\ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
109 || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) |
4481 | 110 #define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 \ |
6492 | 111 || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
112 || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) |
6503 | 113 #define isRGB(x) (((x)&IMGFMT_RGB_MASK)==IMGFMT_RGB) |
114 #define isBGR(x) (((x)&IMGFMT_BGR_MASK)==IMGFMT_BGR) | |
115 #define isPacked(x) ((x)==IMGFMT_YUY2 || isRGB(x) || isBGR(x)) | |
4467 | 116 |
4481 | 117 #define RGB2YUV_SHIFT 16 |
4467 | 118 #define BY ((int)( 0.098*(1<<RGB2YUV_SHIFT)+0.5)) |
119 #define BV ((int)(-0.071*(1<<RGB2YUV_SHIFT)+0.5)) | |
120 #define BU ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5)) | |
121 #define GY ((int)( 0.504*(1<<RGB2YUV_SHIFT)+0.5)) | |
122 #define GV ((int)(-0.368*(1<<RGB2YUV_SHIFT)+0.5)) | |
123 #define GU ((int)(-0.291*(1<<RGB2YUV_SHIFT)+0.5)) | |
124 #define RY ((int)( 0.257*(1<<RGB2YUV_SHIFT)+0.5)) | |
125 #define RV ((int)( 0.439*(1<<RGB2YUV_SHIFT)+0.5)) | |
126 #define RU ((int)(-0.148*(1<<RGB2YUV_SHIFT)+0.5)) | |
4419 | 127 |
3352 | 128 extern int verbose; // defined in mplayer.c |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
129 /* |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
130 NOTES |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
131 Special versions: fast Y 1:1 scaling (no interpolation in y direction) |
2216 | 132 |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
133 TODO |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
134 more intelligent missalignment avoidance for the horizontal scaler |
3344 | 135 write special vertical cubic upscale version |
136 Optimize C code (yv12 / minmax) | |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
137 add support for packed pixel yuv input & output |
4481 | 138 add support for Y8 output |
139 optimize bgr24 & bgr32 | |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
140 add BGR4 output support |
4467 | 141 write special BGR->BGR scaler |
4554 | 142 deglobalize yuv2rgb*.c |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
143 */ |
2216 | 144 |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
145 #define ABS(a) ((a) > 0 ? (a) : (-(a))) |
2469 | 146 #define MIN(a,b) ((a) > (b) ? (b) : (a)) |
147 #define MAX(a,b) ((a) < (b) ? (b) : (a)) | |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
148 |
3126 | 149 #ifdef ARCH_X86 |
150 #define CAN_COMPILE_X86_ASM | |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
151 #endif |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
152 |
3126 | 153 #ifdef CAN_COMPILE_X86_ASM |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
154 static uint64_t __attribute__((aligned(8))) yCoeff= 0x2568256825682568LL; |
2503 | 155 static uint64_t __attribute__((aligned(8))) vrCoeff= 0x3343334333433343LL; |
156 static uint64_t __attribute__((aligned(8))) ubCoeff= 0x40cf40cf40cf40cfLL; | |
157 static uint64_t __attribute__((aligned(8))) vgCoeff= 0xE5E2E5E2E5E2E5E2LL; | |
158 static uint64_t __attribute__((aligned(8))) ugCoeff= 0xF36EF36EF36EF36ELL; | |
2669 | 159 static uint64_t __attribute__((aligned(8))) bF8= 0xF8F8F8F8F8F8F8F8LL; |
160 static uint64_t __attribute__((aligned(8))) bFC= 0xFCFCFCFCFCFCFCFCLL; | |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
161 static uint64_t __attribute__((aligned(8))) w400= 0x0400040004000400LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
162 static uint64_t __attribute__((aligned(8))) w80= 0x0080008000800080LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
163 static uint64_t __attribute__((aligned(8))) w10= 0x0010001000100010LL; |
3272 | 164 static uint64_t __attribute__((aligned(8))) w02= 0x0002000200020002LL; |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
165 static uint64_t __attribute__((aligned(8))) bm00001111=0x00000000FFFFFFFFLL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
166 static uint64_t __attribute__((aligned(8))) bm00000111=0x0000000000FFFFFFLL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
167 static uint64_t __attribute__((aligned(8))) bm11111000=0xFFFFFFFFFF000000LL; |
4481 | 168 static uint64_t __attribute__((aligned(8))) bm01010101=0x00FF00FF00FF00FFLL; |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
169 |
2750
9ef09e232505
gcc does optimize writes to non volatile variables away if it didnt know that they were read in between
michael
parents:
2748
diff
changeset
|
170 static volatile uint64_t __attribute__((aligned(8))) b5Dither; |
9ef09e232505
gcc does optimize writes to non volatile variables away if it didnt know that they were read in between
michael
parents:
2748
diff
changeset
|
171 static volatile uint64_t __attribute__((aligned(8))) g5Dither; |
9ef09e232505
gcc does optimize writes to non volatile variables away if it didnt know that they were read in between
michael
parents:
2748
diff
changeset
|
172 static volatile uint64_t __attribute__((aligned(8))) g6Dither; |
9ef09e232505
gcc does optimize writes to non volatile variables away if it didnt know that they were read in between
michael
parents:
2748
diff
changeset
|
173 static volatile uint64_t __attribute__((aligned(8))) r5Dither; |
2748 | 174 |
175 static uint64_t __attribute__((aligned(8))) dither4[2]={ | |
176 0x0103010301030103LL, | |
177 0x0200020002000200LL,}; | |
178 | |
179 static uint64_t __attribute__((aligned(8))) dither8[2]={ | |
180 0x0602060206020602LL, | |
181 0x0004000400040004LL,}; | |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
182 |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
183 static uint64_t __attribute__((aligned(8))) b16Mask= 0x001F001F001F001FLL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
184 static uint64_t __attribute__((aligned(8))) g16Mask= 0x07E007E007E007E0LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
185 static uint64_t __attribute__((aligned(8))) r16Mask= 0xF800F800F800F800LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
186 static uint64_t __attribute__((aligned(8))) b15Mask= 0x001F001F001F001FLL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
187 static uint64_t __attribute__((aligned(8))) g15Mask= 0x03E003E003E003E0LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
188 static uint64_t __attribute__((aligned(8))) r15Mask= 0x7C007C007C007C00LL; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
189 |
2730 | 190 static uint64_t __attribute__((aligned(8))) M24A= 0x00FF0000FF0000FFLL; |
191 static uint64_t __attribute__((aligned(8))) M24B= 0xFF0000FF0000FF00LL; | |
192 static uint64_t __attribute__((aligned(8))) M24C= 0x0000FF0000FF0000LL; | |
193 | |
4612 | 194 #ifdef FAST_BGR2YV12 |
195 static const uint64_t bgr2YCoeff __attribute__((aligned(8))) = 0x000000210041000DULL; | |
4619 | 196 static const uint64_t bgr2UCoeff __attribute__((aligned(8))) = 0x0000FFEEFFDC0038ULL; |
197 static const uint64_t bgr2VCoeff __attribute__((aligned(8))) = 0x00000038FFD2FFF8ULL; | |
4612 | 198 #else |
199 static const uint64_t bgr2YCoeff __attribute__((aligned(8))) = 0x000020E540830C8BULL; | |
4619 | 200 static const uint64_t bgr2UCoeff __attribute__((aligned(8))) = 0x0000ED0FDAC23831ULL; |
201 static const uint64_t bgr2VCoeff __attribute__((aligned(8))) = 0x00003831D0E6F6EAULL; | |
4612 | 202 #endif |
203 static const uint64_t bgr2YOffset __attribute__((aligned(8))) = 0x1010101010101010ULL; | |
4619 | 204 static const uint64_t bgr2UVOffset __attribute__((aligned(8)))= 0x8080808080808080ULL; |
4612 | 205 static const uint64_t w1111 __attribute__((aligned(8))) = 0x0001000100010001ULL; |
206 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
207 // FIXME remove |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
208 static uint64_t __attribute__((aligned(8))) asm_yalpha1; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
209 static uint64_t __attribute__((aligned(8))) asm_uvalpha1; |
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
210 #endif |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
211 |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
212 // clipping helper table for C implementations: |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
213 static unsigned char clip_table[768]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
214 |
2584 | 215 static unsigned short clip_table16b[768]; |
216 static unsigned short clip_table16g[768]; | |
217 static unsigned short clip_table16r[768]; | |
218 static unsigned short clip_table15b[768]; | |
219 static unsigned short clip_table15g[768]; | |
220 static unsigned short clip_table15r[768]; | |
221 | |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
222 // yuv->rgb conversion tables: |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
223 static int yuvtab_2568[256]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
224 static int yuvtab_3343[256]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
225 static int yuvtab_0c92[256]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
226 static int yuvtab_1a1e[256]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
227 static int yuvtab_40cf[256]; |
3344 | 228 // Needed for cubic scaler to catch overflows |
229 static int clip_yuvtab_2568[768]; | |
230 static int clip_yuvtab_3343[768]; | |
231 static int clip_yuvtab_0c92[768]; | |
232 static int clip_yuvtab_1a1e[768]; | |
233 static int clip_yuvtab_40cf[768]; | |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
234 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
235 //global sws_flags from the command line |
4421 | 236 int sws_flags=2; |
3272 | 237 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
238 //global srcFilter |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
239 SwsFilter src_filter= {NULL, NULL, NULL, NULL}; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
240 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
241 float sws_lum_gblur= 0.0; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
242 float sws_chr_gblur= 0.0; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
243 int sws_chr_vshift= 0; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
244 int sws_chr_hshift= 0; |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
245 float sws_chr_sharpen= 0.0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
246 float sws_lum_sharpen= 0.0; |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
247 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
248 /* cpuCaps combined from cpudetect and whats actually compiled in |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
249 (if there is no support for something compiled in it wont appear here) */ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
250 static CpuCaps cpuCaps; |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
251 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
252 void (*swScale)(SwsContext *context, uint8_t* src[], int srcStride[], int srcSliceY, |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
253 int srcSliceH, uint8_t* dst[], int dstStride[])=NULL; |
2469 | 254 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
255 static SwsVector *getConvVec(SwsVector *a, SwsVector *b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
256 |
3126 | 257 #ifdef CAN_COMPILE_X86_ASM |
2671 | 258 void in_asm_used_var_warning_killer() |
259 { | |
3272 | 260 volatile int i= yCoeff+vrCoeff+ubCoeff+vgCoeff+ugCoeff+bF8+bFC+w400+w80+w10+ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
261 bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+asm_yalpha1+ asm_uvalpha1+ |
4481 | 262 M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101; |
2671 | 263 if(i) i=0; |
264 } | |
265 #endif | |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
266 |
6517 | 267 static int testFormat[]={ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
268 IMGFMT_YVU9, |
6517 | 269 IMGFMT_YV12, |
270 //IMGFMT_IYUV, | |
271 IMGFMT_I420, | |
272 IMGFMT_BGR15, | |
273 IMGFMT_BGR16, | |
274 IMGFMT_BGR24, | |
275 IMGFMT_BGR32, | |
276 //IMGFMT_Y8, | |
277 IMGFMT_Y800, | |
278 //IMGFMT_YUY2, | |
279 0 | |
280 }; | |
281 | |
282 static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1, int stride2, int w, int h){ | |
283 int x,y; | |
284 uint64_t ssd=0; | |
285 | |
286 for(y=0; y<h; y++){ | |
287 for(x=0; x<w; x++){ | |
288 int d= src1[x + y*stride1] - src2[x + y*stride2]; | |
289 ssd+= d*d; | |
290 } | |
291 } | |
292 return ssd; | |
293 } | |
294 | |
295 // test by ref -> src -> dst -> out & compare out against ref | |
296 // ref & out are YV12 | |
297 static void doTest(uint8_t *ref[3], int refStride[3], int w, int h, int srcFormat, int dstFormat, | |
298 int srcW, int srcH, int dstW, int dstH, int flags){ | |
299 uint8_t *src[3]; | |
300 uint8_t *dst[3]; | |
301 uint8_t *out[3]; | |
302 int srcStride[3], dstStride[3]; | |
303 int i; | |
304 uint64_t ssdY, ssdU, ssdV; | |
305 SwsContext *srcContext, *dstContext, *outContext; | |
306 | |
307 for(i=0; i<3; i++){ | |
308 srcStride[i]= srcW*4; | |
309 dstStride[i]= dstW*4; | |
310 src[i]= malloc(srcStride[i]*srcH); | |
311 dst[i]= malloc(dstStride[i]*dstH); | |
312 out[i]= malloc(refStride[i]*h); | |
313 } | |
314 | |
315 srcContext= getSwsContext(w, h, IMGFMT_YV12, srcW, srcH, srcFormat, flags, NULL, NULL); | |
316 dstContext= getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, NULL, NULL); | |
317 outContext= getSwsContext(dstW, dstH, dstFormat, w, h, IMGFMT_YV12, flags, NULL, NULL); | |
318 if(srcContext==NULL ||dstContext==NULL ||outContext==NULL){ | |
319 printf("Failed allocating swsContext\n"); | |
320 goto end; | |
321 } | |
322 // printf("test %X %X %X -> %X %X %X\n", (int)ref[0], (int)ref[1], (int)ref[2], | |
323 // (int)src[0], (int)src[1], (int)src[2]); | |
324 | |
325 srcContext->swScale(srcContext, ref, refStride, 0, h , src, srcStride); | |
326 dstContext->swScale(dstContext, src, srcStride, 0, srcH, dst, dstStride); | |
327 outContext->swScale(outContext, dst, dstStride, 0, dstH, out, refStride); | |
328 | |
329 ssdY= getSSD(ref[0], out[0], refStride[0], refStride[0], w, h); | |
330 ssdU= getSSD(ref[1], out[1], refStride[1], refStride[1], (w+1)>>1, (h+1)>>1); | |
331 ssdV= getSSD(ref[2], out[2], refStride[2], refStride[2], (w+1)>>1, (h+1)>>1); | |
332 | |
333 if(isGray(srcFormat) || isGray(dstFormat)) ssdU=ssdV=0; //FIXME check that output is really gray | |
334 | |
335 ssdY/= w*h; | |
336 ssdU/= w*h/4; | |
337 ssdV/= w*h/4; | |
338 | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
339 if(ssdY>100 || ssdU>50 || ssdV>50){ |
6517 | 340 printf(" %s %dx%d -> %s %4dx%4d flags=%2d SSD=%5lld,%5lld,%5lld\n", |
341 vo_format_name(srcFormat), srcW, srcH, | |
342 vo_format_name(dstFormat), dstW, dstH, | |
343 flags, | |
344 ssdY, ssdU, ssdV); | |
345 } | |
346 | |
347 end: | |
348 | |
349 freeSwsContext(srcContext); | |
350 freeSwsContext(dstContext); | |
351 freeSwsContext(outContext); | |
352 | |
353 for(i=0; i<3; i++){ | |
354 free(src[i]); | |
355 free(dst[i]); | |
356 free(out[i]); | |
357 } | |
358 } | |
359 | |
360 static void selfTest(uint8_t *src[3], int stride[3], int w, int h){ | |
361 int srcFormat, dstFormat, srcFormatIndex, dstFormatIndex; | |
362 int srcW, srcH, dstW, dstH; | |
363 int flags; | |
364 | |
365 for(srcFormatIndex=0; ;srcFormatIndex++){ | |
366 srcFormat= testFormat[srcFormatIndex]; | |
367 if(!srcFormat) break; | |
368 for(dstFormatIndex=0; ;dstFormatIndex++){ | |
369 dstFormat= testFormat[dstFormatIndex]; | |
370 if(!dstFormat) break; | |
371 if(!isSupportedOut(dstFormat)) continue; | |
372 | |
373 srcW= w+w/3; | |
374 srcH= h+h/3; | |
375 for(dstW=w; dstW<w*2; dstW+= dstW/3){ | |
376 for(dstH=h; dstH<h*2; dstH+= dstH/3){ | |
377 for(flags=1; flags<33; flags*=2) | |
378 doTest(src, stride, w, h, srcFormat, dstFormat, | |
379 srcW, srcH, dstW, dstH, flags); | |
380 } | |
381 } | |
382 } | |
383 } | |
384 } | |
385 | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
386 static inline void yuv2yuvXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, |
3352 | 387 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
388 uint8_t *dest, uint8_t *uDest, uint8_t *vDest) |
3352 | 389 { |
390 //FIXME Optimize (just quickly writen not opti..) | |
391 int i; | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
392 for(i=0; i<c->dstW; i++) |
3352 | 393 { |
394 int val=0; | |
395 int j; | |
396 for(j=0; j<lumFilterSize; j++) | |
397 val += lumSrc[j][i] * lumFilter[j]; | |
398 | |
399 dest[i]= MIN(MAX(val>>19, 0), 255); | |
400 } | |
401 | |
402 if(uDest != NULL) | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
403 for(i=0; i<c->chrDstW; i++) |
3352 | 404 { |
405 int u=0; | |
406 int v=0; | |
407 int j; | |
3641 | 408 for(j=0; j<chrFilterSize; j++) |
3352 | 409 { |
410 u += chrSrc[j][i] * chrFilter[j]; | |
411 v += chrSrc[j][i + 2048] * chrFilter[j]; | |
412 } | |
413 | |
414 uDest[i]= MIN(MAX(u>>19, 0), 255); | |
415 vDest[i]= MIN(MAX(v>>19, 0), 255); | |
416 } | |
417 } | |
418 | |
419 static inline void yuv2rgbXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, | |
420 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
421 uint8_t *dest, int dstW, int dstFormat) |
3352 | 422 { |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
423 if(dstFormat==IMGFMT_BGR32) |
3352 | 424 { |
4794 | 425 int i; |
4793 | 426 #ifdef WORDS_BIGENDIAN |
427 dest++; | |
428 #endif | |
3352 | 429 for(i=0; i<(dstW>>1); i++){ |
430 int j; | |
431 int Y1=0; | |
432 int Y2=0; | |
433 int U=0; | |
434 int V=0; | |
435 int Cb, Cr, Cg; | |
436 for(j=0; j<lumFilterSize; j++) | |
437 { | |
438 Y1 += lumSrc[j][2*i] * lumFilter[j]; | |
439 Y2 += lumSrc[j][2*i+1] * lumFilter[j]; | |
440 } | |
441 for(j=0; j<chrFilterSize; j++) | |
442 { | |
443 U += chrSrc[j][i] * chrFilter[j]; | |
444 V += chrSrc[j][i+2048] * chrFilter[j]; | |
445 } | |
446 Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; | |
447 Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; | |
448 U >>= 19; | |
449 V >>= 19; | |
450 | |
451 Cb= clip_yuvtab_40cf[U+ 256]; | |
452 Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; | |
453 Cr= clip_yuvtab_3343[V+ 256]; | |
454 | |
455 dest[8*i+0]=clip_table[((Y1 + Cb) >>13)]; | |
456 dest[8*i+1]=clip_table[((Y1 + Cg) >>13)]; | |
457 dest[8*i+2]=clip_table[((Y1 + Cr) >>13)]; | |
458 | |
459 dest[8*i+4]=clip_table[((Y2 + Cb) >>13)]; | |
460 dest[8*i+5]=clip_table[((Y2 + Cg) >>13)]; | |
461 dest[8*i+6]=clip_table[((Y2 + Cr) >>13)]; | |
462 } | |
463 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
464 else if(dstFormat==IMGFMT_BGR24) |
3352 | 465 { |
466 int i; | |
467 for(i=0; i<(dstW>>1); i++){ | |
468 int j; | |
469 int Y1=0; | |
470 int Y2=0; | |
471 int U=0; | |
472 int V=0; | |
473 int Cb, Cr, Cg; | |
474 for(j=0; j<lumFilterSize; j++) | |
475 { | |
476 Y1 += lumSrc[j][2*i] * lumFilter[j]; | |
477 Y2 += lumSrc[j][2*i+1] * lumFilter[j]; | |
478 } | |
479 for(j=0; j<chrFilterSize; j++) | |
480 { | |
481 U += chrSrc[j][i] * chrFilter[j]; | |
482 V += chrSrc[j][i+2048] * chrFilter[j]; | |
483 } | |
484 Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; | |
485 Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; | |
486 U >>= 19; | |
487 V >>= 19; | |
488 | |
489 Cb= clip_yuvtab_40cf[U+ 256]; | |
490 Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; | |
491 Cr= clip_yuvtab_3343[V+ 256]; | |
492 | |
493 dest[0]=clip_table[((Y1 + Cb) >>13)]; | |
494 dest[1]=clip_table[((Y1 + Cg) >>13)]; | |
495 dest[2]=clip_table[((Y1 + Cr) >>13)]; | |
496 | |
497 dest[3]=clip_table[((Y2 + Cb) >>13)]; | |
498 dest[4]=clip_table[((Y2 + Cg) >>13)]; | |
499 dest[5]=clip_table[((Y2 + Cr) >>13)]; | |
500 dest+=6; | |
501 } | |
502 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
503 else if(dstFormat==IMGFMT_BGR16) |
3352 | 504 { |
505 int i; | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
506 #ifdef DITHER1XBPP |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
507 static int ditherb1=1<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
508 static int ditherg1=1<<13; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
509 static int ditherr1=2<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
510 static int ditherb2=3<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
511 static int ditherg2=3<<13; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
512 static int ditherr2=0<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
513 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
514 ditherb1 ^= (1^2)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
515 ditherg1 ^= (1^2)<<13; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
516 ditherr1 ^= (1^2)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
517 ditherb2 ^= (3^0)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
518 ditherg2 ^= (3^0)<<13; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
519 ditherr2 ^= (3^0)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
520 #else |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
521 const int ditherb1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
522 const int ditherg1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
523 const int ditherr1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
524 const int ditherb2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
525 const int ditherg2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
526 const int ditherr2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
527 #endif |
3352 | 528 for(i=0; i<(dstW>>1); i++){ |
529 int j; | |
530 int Y1=0; | |
531 int Y2=0; | |
532 int U=0; | |
533 int V=0; | |
534 int Cb, Cr, Cg; | |
535 for(j=0; j<lumFilterSize; j++) | |
536 { | |
537 Y1 += lumSrc[j][2*i] * lumFilter[j]; | |
538 Y2 += lumSrc[j][2*i+1] * lumFilter[j]; | |
539 } | |
540 for(j=0; j<chrFilterSize; j++) | |
541 { | |
542 U += chrSrc[j][i] * chrFilter[j]; | |
543 V += chrSrc[j][i+2048] * chrFilter[j]; | |
544 } | |
545 Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; | |
546 Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; | |
547 U >>= 19; | |
548 V >>= 19; | |
549 | |
550 Cb= clip_yuvtab_40cf[U+ 256]; | |
551 Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; | |
552 Cr= clip_yuvtab_3343[V+ 256]; | |
553 | |
554 ((uint16_t*)dest)[2*i] = | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
555 clip_table16b[(Y1 + Cb + ditherb1) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
556 clip_table16g[(Y1 + Cg + ditherg1) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
557 clip_table16r[(Y1 + Cr + ditherr1) >>13]; |
3352 | 558 |
559 ((uint16_t*)dest)[2*i+1] = | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
560 clip_table16b[(Y2 + Cb + ditherb2) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
561 clip_table16g[(Y2 + Cg + ditherg2) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
562 clip_table16r[(Y2 + Cr + ditherr2) >>13]; |
3352 | 563 } |
564 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
565 else if(dstFormat==IMGFMT_BGR15) |
3352 | 566 { |
567 int i; | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
568 #ifdef DITHER1XBPP |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
569 static int ditherb1=1<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
570 static int ditherg1=1<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
571 static int ditherr1=2<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
572 static int ditherb2=3<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
573 static int ditherg2=3<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
574 static int ditherr2=0<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
575 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
576 ditherb1 ^= (1^2)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
577 ditherg1 ^= (1^2)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
578 ditherr1 ^= (1^2)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
579 ditherb2 ^= (3^0)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
580 ditherg2 ^= (3^0)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
581 ditherr2 ^= (3^0)<<14; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
582 #else |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
583 const int ditherb1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
584 const int ditherg1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
585 const int ditherr1=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
586 const int ditherb2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
587 const int ditherg2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
588 const int ditherr2=0; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
589 #endif |
3352 | 590 for(i=0; i<(dstW>>1); i++){ |
591 int j; | |
592 int Y1=0; | |
593 int Y2=0; | |
594 int U=0; | |
595 int V=0; | |
596 int Cb, Cr, Cg; | |
597 for(j=0; j<lumFilterSize; j++) | |
598 { | |
599 Y1 += lumSrc[j][2*i] * lumFilter[j]; | |
600 Y2 += lumSrc[j][2*i+1] * lumFilter[j]; | |
601 } | |
602 for(j=0; j<chrFilterSize; j++) | |
603 { | |
604 U += chrSrc[j][i] * chrFilter[j]; | |
605 V += chrSrc[j][i+2048] * chrFilter[j]; | |
606 } | |
607 Y1= clip_yuvtab_2568[ (Y1>>19) + 256 ]; | |
608 Y2= clip_yuvtab_2568[ (Y2>>19) + 256 ]; | |
609 U >>= 19; | |
610 V >>= 19; | |
611 | |
612 Cb= clip_yuvtab_40cf[U+ 256]; | |
613 Cg= clip_yuvtab_1a1e[V+ 256] + yuvtab_0c92[U+ 256]; | |
614 Cr= clip_yuvtab_3343[V+ 256]; | |
615 | |
616 ((uint16_t*)dest)[2*i] = | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
617 clip_table15b[(Y1 + Cb + ditherb1) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
618 clip_table15g[(Y1 + Cg + ditherg1) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
619 clip_table15r[(Y1 + Cr + ditherr1) >>13]; |
3352 | 620 |
621 ((uint16_t*)dest)[2*i+1] = | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
622 clip_table15b[(Y2 + Cb + ditherb2) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
623 clip_table15g[(Y2 + Cg + ditherg2) >>13] | |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
624 clip_table15r[(Y2 + Cr + ditherr2) >>13]; |
3352 | 625 } |
626 } | |
627 } | |
628 | |
629 | |
3126 | 630 //Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one |
631 //Plain C versions | |
3152 | 632 #if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) |
633 #define COMPILE_C | |
634 #endif | |
635 | |
636 #ifdef CAN_COMPILE_X86_ASM | |
637 | |
638 #if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
639 #define COMPILE_MMX | |
640 #endif | |
641 | |
642 #if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT) | |
643 #define COMPILE_MMX2 | |
644 #endif | |
645 | |
646 #if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
647 #define COMPILE_3DNOW | |
648 #endif | |
649 #endif //CAN_COMPILE_X86_ASM | |
650 | |
651 #undef HAVE_MMX | |
652 #undef HAVE_MMX2 | |
653 #undef HAVE_3DNOW | |
654 | |
655 #ifdef COMPILE_C | |
3126 | 656 #undef HAVE_MMX |
657 #undef HAVE_MMX2 | |
658 #undef HAVE_3DNOW | |
659 #define RENAME(a) a ## _C | |
660 #include "swscale_template.c" | |
3152 | 661 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
662 |
3126 | 663 #ifdef CAN_COMPILE_X86_ASM |
2576 | 664 |
3126 | 665 //X86 versions |
666 /* | |
667 #undef RENAME | |
668 #undef HAVE_MMX | |
669 #undef HAVE_MMX2 | |
670 #undef HAVE_3DNOW | |
671 #define ARCH_X86 | |
672 #define RENAME(a) a ## _X86 | |
673 #include "swscale_template.c" | |
674 */ | |
675 //MMX versions | |
3152 | 676 #ifdef COMPILE_MMX |
3126 | 677 #undef RENAME |
678 #define HAVE_MMX | |
679 #undef HAVE_MMX2 | |
680 #undef HAVE_3DNOW | |
681 #define RENAME(a) a ## _MMX | |
682 #include "swscale_template.c" | |
3152 | 683 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
684 |
3126 | 685 //MMX2 versions |
3152 | 686 #ifdef COMPILE_MMX2 |
3126 | 687 #undef RENAME |
688 #define HAVE_MMX | |
689 #define HAVE_MMX2 | |
690 #undef HAVE_3DNOW | |
691 #define RENAME(a) a ## _MMX2 | |
692 #include "swscale_template.c" | |
3152 | 693 #endif |
2469 | 694 |
3126 | 695 //3DNOW versions |
3152 | 696 #ifdef COMPILE_3DNOW |
3126 | 697 #undef RENAME |
698 #define HAVE_MMX | |
699 #undef HAVE_MMX2 | |
700 #define HAVE_3DNOW | |
701 #define RENAME(a) a ## _3DNow | |
702 #include "swscale_template.c" | |
3152 | 703 #endif |
2469 | 704 |
3126 | 705 #endif //CAN_COMPILE_X86_ASM |
2469 | 706 |
3126 | 707 // minor note: the HAVE_xyz is messed up after that line so dont use it |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
708 |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
709 |
4419 | 710 // old global scaler, dont use for new code |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
711 // will use sws_flags from the command line |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
712 void SwScale_YV12slice(unsigned char* src[], int srcStride[], int srcSliceY , |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
713 int srcSliceH, uint8_t* dst[], int dstStride, int dstbpp, |
3209 | 714 int srcW, int srcH, int dstW, int dstH){ |
2216 | 715 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
716 static SwsContext *context=NULL; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
717 int dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
718 int dstStride3[3]= {dstStride, dstStride>>1, dstStride>>1}; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
719 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
720 switch(dstbpp) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
721 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
722 case 8 : dstFormat= IMGFMT_Y8; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
723 case 12: dstFormat= IMGFMT_YV12; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
724 case 15: dstFormat= IMGFMT_BGR15; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
725 case 16: dstFormat= IMGFMT_BGR16; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
726 case 24: dstFormat= IMGFMT_BGR24; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
727 case 32: dstFormat= IMGFMT_BGR32; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
728 default: return; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
729 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
730 |
4419 | 731 if(!context) context=getSwsContextFromCmdLine(srcW, srcH, IMGFMT_YV12, dstW, dstH, dstFormat); |
732 | |
4555 | 733 context->swScale(context, src, srcStride, srcSliceY, srcSliceH, dst, dstStride3); |
4419 | 734 } |
735 | |
736 // will use sws_flags & src_filter (from cmd line) | |
737 SwsContext *getSwsContextFromCmdLine(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat) | |
738 { | |
739 int flags=0; | |
740 static int firstTime=1; | |
741 | |
742 #ifdef ARCH_X86 | |
743 if(gCpuCaps.hasMMX) | |
744 asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions) | |
745 #endif | |
746 if(firstTime) | |
747 { | |
748 firstTime=0; | |
749 flags= SWS_PRINT_INFO; | |
750 } | |
751 else if(verbose>1) flags= SWS_PRINT_INFO; | |
752 | |
753 if(src_filter.lumH) freeVec(src_filter.lumH); | |
754 if(src_filter.lumV) freeVec(src_filter.lumV); | |
755 if(src_filter.chrH) freeVec(src_filter.chrH); | |
756 if(src_filter.chrV) freeVec(src_filter.chrV); | |
757 | |
758 if(sws_lum_gblur!=0.0){ | |
759 src_filter.lumH= getGaussianVec(sws_lum_gblur, 3.0); | |
760 src_filter.lumV= getGaussianVec(sws_lum_gblur, 3.0); | |
761 }else{ | |
762 src_filter.lumH= getIdentityVec(); | |
763 src_filter.lumV= getIdentityVec(); | |
764 } | |
765 | |
766 if(sws_chr_gblur!=0.0){ | |
767 src_filter.chrH= getGaussianVec(sws_chr_gblur, 3.0); | |
768 src_filter.chrV= getGaussianVec(sws_chr_gblur, 3.0); | |
769 }else{ | |
770 src_filter.chrH= getIdentityVec(); | |
771 src_filter.chrV= getIdentityVec(); | |
772 } | |
773 | |
774 if(sws_chr_sharpen!=0.0){ | |
775 SwsVector *g= getConstVec(-1.0, 3); | |
776 SwsVector *id= getConstVec(10.0/sws_chr_sharpen, 1); | |
777 g->coeff[1]=2.0; | |
778 addVec(id, g); | |
779 convVec(src_filter.chrH, id); | |
780 convVec(src_filter.chrV, id); | |
781 freeVec(g); | |
782 freeVec(id); | |
783 } | |
784 | |
785 if(sws_lum_sharpen!=0.0){ | |
786 SwsVector *g= getConstVec(-1.0, 3); | |
787 SwsVector *id= getConstVec(10.0/sws_lum_sharpen, 1); | |
788 g->coeff[1]=2.0; | |
789 addVec(id, g); | |
790 convVec(src_filter.lumH, id); | |
791 convVec(src_filter.lumV, id); | |
792 freeVec(g); | |
793 freeVec(id); | |
794 } | |
795 | |
796 if(sws_chr_hshift) | |
797 shiftVec(src_filter.chrH, sws_chr_hshift); | |
798 | |
799 if(sws_chr_vshift) | |
800 shiftVec(src_filter.chrV, sws_chr_vshift); | |
801 | |
802 normalizeVec(src_filter.chrH, 1.0); | |
803 normalizeVec(src_filter.chrV, 1.0); | |
804 normalizeVec(src_filter.lumH, 1.0); | |
805 normalizeVec(src_filter.lumV, 1.0); | |
806 | |
807 if(verbose > 1) printVec(src_filter.chrH); | |
808 if(verbose > 1) printVec(src_filter.lumH); | |
809 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
810 switch(sws_flags) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
811 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
812 case 0: flags|= SWS_FAST_BILINEAR; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
813 case 1: flags|= SWS_BILINEAR; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
814 case 2: flags|= SWS_BICUBIC; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
815 case 3: flags|= SWS_X; break; |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
816 case 4: flags|= SWS_POINT; break; |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
817 case 5: flags|= SWS_AREA; break; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
818 default:flags|= SWS_BILINEAR; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
819 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
820 |
4419 | 821 return getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, &src_filter, NULL); |
822 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
823 |
2216 | 824 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
825 static inline void initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
826 int srcW, int dstW, int filterAlign, int one, int flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
827 SwsVector *srcFilter, SwsVector *dstFilter) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
828 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
829 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
830 int filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
831 int filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
832 int minFilterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
833 double *filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
834 double *filter2=NULL; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
835 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
836 if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
837 asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
838 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
839 |
4563
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
840 // Note the +1 is for the MMXscaler which reads over the end |
4419 | 841 *filterPos = (int16_t*)memalign(8, (dstW+1)*sizeof(int16_t)); |
842 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
843 if(ABS(xInc - 0x10000) <10) // unscaled |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
844 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
845 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
846 filterSize= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
847 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
848 for(i=0; i<dstW*filterSize; i++) filter[i]=0; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
849 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
850 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
851 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
852 filter[i*filterSize]=1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
853 (*filterPos)[i]=i; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
854 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
855 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
856 } |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
857 else if(flags&SWS_POINT) // lame looking point sampling mode |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
858 { |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
859 int i; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
860 int xDstInSrc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
861 filterSize= 1; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
862 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
863 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
864 xDstInSrc= xInc/2 - 0x8000; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
865 for(i=0; i<dstW; i++) |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
866 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
867 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
868 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
869 (*filterPos)[i]= xx; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
870 filter[i]= 1.0; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
871 xDstInSrc+= xInc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
872 } |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
873 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
874 else if(xInc <= (1<<16) || (flags&SWS_FAST_BILINEAR)) // upscale |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
875 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
876 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
877 int xDstInSrc; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
878 if (flags&SWS_BICUBIC) filterSize= 4; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
879 else if(flags&SWS_X ) filterSize= 4; |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
880 else filterSize= 2; // SWS_BILINEAR / SWS_AREA |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
881 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
882 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
883 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
884 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
885 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
886 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
887 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
888 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
889 (*filterPos)[i]= xx; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
890 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
891 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
892 double d= ABS(((xx+1)<<16) - xDstInSrc)/(double)(1<<16); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
893 double y1,y2,y3,y4; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
894 double A= -0.6; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
895 if(flags & SWS_BICUBIC){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
896 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
897 y1 = ( + A*d - 2.0*A*d*d + A*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
898 y2 = (+ 1.0 - (A+3.0)*d*d + (A+2.0)*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
899 y3 = ( - A*d + (2.0*A+3.0)*d*d - (A+2.0)*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
900 y4 = ( + A*d*d - A*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
901 }else{ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
902 // cubic interpolation (derived it myself) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
903 y1 = ( -2.0*d + 3.0*d*d - 1.0*d*d*d)/6.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
904 y2 = (6.0 -3.0*d - 6.0*d*d + 3.0*d*d*d)/6.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
905 y3 = ( +6.0*d + 3.0*d*d - 3.0*d*d*d)/6.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
906 y4 = ( -1.0*d + 1.0*d*d*d)/6.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
907 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
908 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
909 filter[i*filterSize + 0]= y1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
910 filter[i*filterSize + 1]= y2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
911 filter[i*filterSize + 2]= y3; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
912 filter[i*filterSize + 3]= y4; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
913 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
914 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
915 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
916 //Bilinear upscale / linear interpolate / Area averaging |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
917 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
918 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
919 double d= ABS((xx<<16) - xDstInSrc)/(double)(1<<16); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
920 double coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
921 if(coeff<0) coeff=0; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
922 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
923 xx++; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
924 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
925 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
926 xDstInSrc+= xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
927 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
928 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
929 else // downscale |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
930 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
931 int xDstInSrc; |
4806 | 932 ASSERT(dstW <= srcW) |
933 | |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
934 if(flags&SWS_BICUBIC) filterSize= (int)ceil(1 + 4.0*srcW / (double)dstW); |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
935 else if(flags&SWS_X) filterSize= (int)ceil(1 + 4.0*srcW / (double)dstW); |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
936 else if(flags&SWS_AREA) filterSize= (int)ceil(1 + 1.0*srcW / (double)dstW); |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
937 else /* BILINEAR */ filterSize= (int)ceil(1 + 2.0*srcW / (double)dstW); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
938 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
939 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
940 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
941 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
942 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
943 int xx= (int)((double)xDstInSrc/(double)(1<<16) - (filterSize-1)*0.5 + 0.5); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
944 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
945 (*filterPos)[i]= xx; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
946 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
947 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
948 double d= ABS((xx<<16) - xDstInSrc)/(double)xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
949 double coeff; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
950 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
951 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
952 double A= -0.75; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
953 // d*=2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
954 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
955 if(d<1.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
956 coeff = (1.0 - (A+3.0)*d*d + (A+2.0)*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
957 else if(d<2.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
958 coeff = (-4.0*A + 8.0*A*d - 5.0*A*d*d + A*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
959 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
960 coeff=0.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
961 } |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
962 else if(flags & SWS_AREA) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
963 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
964 double srcPixelSize= (1<<16)/(double)xInc; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
965 if(d + srcPixelSize/2 < 0.5) coeff= 1.0; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
966 else if(d - srcPixelSize/2 < 0.5) coeff= (0.5-d)/srcPixelSize + 0.5; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
967 else coeff=0.0; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
968 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
969 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
970 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
971 coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
972 if(coeff<0) coeff=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
973 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
974 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
975 xx++; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
976 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
977 xDstInSrc+= xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
978 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
979 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
980 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
981 /* apply src & dst Filter to filter -> filter2 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
982 free(filter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
983 */ |
4806 | 984 ASSERT(filterSize>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
985 filter2Size= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
986 if(srcFilter) filter2Size+= srcFilter->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
987 if(dstFilter) filter2Size+= dstFilter->length - 1; |
4806 | 988 ASSERT(filter2Size>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
989 filter2= (double*)memalign(8, filter2Size*dstW*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
990 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
991 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
992 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
993 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
994 SwsVector scaleFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
995 SwsVector *outVec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
996 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
997 scaleFilter.coeff= filter + i*filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
998 scaleFilter.length= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
999 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1000 if(srcFilter) outVec= getConvVec(srcFilter, &scaleFilter); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1001 else outVec= &scaleFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1002 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1003 ASSERT(outVec->length == filter2Size) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1004 //FIXME dstFilter |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1005 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1006 for(j=0; j<outVec->length; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1007 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1008 filter2[i*filter2Size + j]= outVec->coeff[j]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1009 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1010 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1011 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1012 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1013 if(outVec != &scaleFilter) freeVec(outVec); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1014 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1015 free(filter); filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1016 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1017 /* try to reduce the filter-size (step1 find size and shift left) */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1018 // Assume its near normalized (*0.5 or *2.0 is ok but * 0.001 is not) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1019 minFilterSize= 0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1020 for(i=dstW-1; i>=0; i--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1021 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1022 int min= filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1023 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1024 double cutOff=0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1025 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1026 /* get rid off near zero elements on the left by shifting left */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1027 for(j=0; j<filter2Size; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1028 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1029 int k; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1030 cutOff += ABS(filter2[i*filter2Size]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1031 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1032 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1033 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1034 /* preserve Monotonicity because the core cant handle the filter otherwise */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1035 if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1036 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1037 // Move filter coeffs left |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1038 for(k=1; k<filter2Size; k++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1039 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1040 filter2[i*filter2Size + k - 1]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1041 (*filterPos)[i]++; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1042 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1043 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1044 cutOff=0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1045 /* count near zeros on the right */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1046 for(j=filter2Size-1; j>0; j--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1047 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1048 cutOff += ABS(filter2[i*filter2Size + j]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1049 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1050 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1051 min--; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1052 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1053 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1054 if(min>minFilterSize) minFilterSize= min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1055 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1056 |
4806 | 1057 ASSERT(minFilterSize > 0) |
4419 | 1058 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1)); |
4806 | 1059 ASSERT(filterSize > 0) |
4419 | 1060 filter= (double*)memalign(8, filterSize*dstW*sizeof(double)); |
1061 *outFilterSize= filterSize; | |
1062 | |
5937 | 1063 if(flags&SWS_PRINT_INFO) |
6492 | 1064 MSG_INFO("SwScaler: reducing / aligning filtersize %d -> %d\n", filter2Size, filterSize); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1065 /* try to reduce the filter-size (step2 reduce it) */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1066 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1067 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1068 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1069 |
4419 | 1070 for(j=0; j<filterSize; j++) |
1071 { | |
1072 if(j>=filter2Size) filter[i*filterSize + j]= 0.0; | |
1073 else filter[i*filterSize + j]= filter2[i*filter2Size + j]; | |
1074 } | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1075 } |
4419 | 1076 free(filter2); filter2=NULL; |
1077 | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1078 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1079 //FIXME try to align filterpos if possible |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1080 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1081 //fix borders |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1082 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1083 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1084 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1085 if((*filterPos)[i] < 0) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1086 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1087 // Move filter coeffs left to compensate for filterPos |
4419 | 1088 for(j=1; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1089 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1090 int left= MAX(j + (*filterPos)[i], 0); |
4419 | 1091 filter[i*filterSize + left] += filter[i*filterSize + j]; |
1092 filter[i*filterSize + j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1093 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1094 (*filterPos)[i]= 0; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1095 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1096 |
4419 | 1097 if((*filterPos)[i] + filterSize > srcW) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1098 { |
4419 | 1099 int shift= (*filterPos)[i] + filterSize - srcW; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1100 // Move filter coeffs right to compensate for filterPos |
4419 | 1101 for(j=filterSize-2; j>=0; j--) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1102 { |
4419 | 1103 int right= MIN(j + shift, filterSize-1); |
1104 filter[i*filterSize +right] += filter[i*filterSize +j]; | |
1105 filter[i*filterSize +j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1106 } |
4419 | 1107 (*filterPos)[i]= srcW - filterSize; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1108 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1109 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1110 |
4419 | 1111 // Note the +1 is for the MMXscaler which reads over the end |
1112 *outFilter= (int16_t*)memalign(8, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
1113 memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1114 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1115 /* Normalize & Store in outFilter */ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1116 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1117 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1118 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1119 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1120 double scale= one; |
4419 | 1121 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1122 { |
4419 | 1123 sum+= filter[i*filterSize + j]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1124 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1125 scale/= sum; |
4419 | 1126 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1127 { |
4419 | 1128 (*outFilter)[i*(*outFilterSize) + j]= (int)(filter[i*filterSize + j]*scale); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1129 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1130 } |
4563
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1131 |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1132 (*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1133 for(i=0; i<*outFilterSize; i++) |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1134 { |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1135 int j= dstW*(*outFilterSize); |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1136 (*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)]; |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1137 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1138 |
4419 | 1139 free(filter); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1140 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1141 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1142 #ifdef ARCH_X86 |
5452 | 1143 static void initMMX2HScaler(int dstW, int xInc, uint8_t *funnyCode, int16_t *filter, int32_t *filterPos, int numSplits) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1144 { |
5452 | 1145 uint8_t *fragmentA; |
1146 int imm8OfPShufW1A; | |
1147 int imm8OfPShufW2A; | |
1148 int fragmentLengthA; | |
1149 uint8_t *fragmentB; | |
1150 int imm8OfPShufW1B; | |
1151 int imm8OfPShufW2B; | |
1152 int fragmentLengthB; | |
1153 int fragmentPos; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1154 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1155 int xpos, i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1156 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1157 // create an optimized horizontal scaling routine |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1158 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1159 //code fragment |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1160 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1161 asm volatile( |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1162 "jmp 9f \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1163 // Begin |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1164 "0: \n\t" |
5452 | 1165 "movq (%%edx, %%eax), %%mm3 \n\t" |
1166 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1167 "movd 1(%%ecx, %%esi), %%mm1 \n\t" | |
1168 "punpcklbw %%mm7, %%mm1 \n\t" | |
1169 "punpcklbw %%mm7, %%mm0 \n\t" | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1170 "pshufw $0xFF, %%mm1, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1171 "1: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1172 "pshufw $0xFF, %%mm0, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1173 "2: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1174 "psubw %%mm1, %%mm0 \n\t" |
5452 | 1175 "movl 8(%%ebx, %%eax), %%esi \n\t" |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1176 "pmullw %%mm3, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1177 "psllw $7, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1178 "paddw %%mm1, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1179 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1180 "movq %%mm0, (%%edi, %%eax) \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1181 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1182 "addl $8, %%eax \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1183 // End |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1184 "9: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1185 // "int $3\n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1186 "leal 0b, %0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1187 "leal 1b, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1188 "leal 2b, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1189 "decl %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1190 "decl %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1191 "subl %0, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1192 "subl %0, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1193 "leal 9b, %3 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1194 "subl %0, %3 \n\t" |
5452 | 1195 |
1196 | |
1197 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), | |
1198 "=r" (fragmentLengthA) | |
1199 ); | |
1200 | |
1201 asm volatile( | |
1202 "jmp 9f \n\t" | |
1203 // Begin | |
1204 "0: \n\t" | |
1205 "movq (%%edx, %%eax), %%mm3 \n\t" | |
1206 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1207 "punpcklbw %%mm7, %%mm0 \n\t" | |
1208 "pshufw $0xFF, %%mm0, %%mm1 \n\t" | |
1209 "1: \n\t" | |
1210 "pshufw $0xFF, %%mm0, %%mm0 \n\t" | |
1211 "2: \n\t" | |
1212 "psubw %%mm1, %%mm0 \n\t" | |
1213 "movl 8(%%ebx, %%eax), %%esi \n\t" | |
1214 "pmullw %%mm3, %%mm0 \n\t" | |
1215 "psllw $7, %%mm1 \n\t" | |
1216 "paddw %%mm1, %%mm0 \n\t" | |
1217 | |
1218 "movq %%mm0, (%%edi, %%eax) \n\t" | |
1219 | |
1220 "addl $8, %%eax \n\t" | |
1221 // End | |
1222 "9: \n\t" | |
1223 // "int $3\n\t" | |
1224 "leal 0b, %0 \n\t" | |
1225 "leal 1b, %1 \n\t" | |
1226 "leal 2b, %2 \n\t" | |
1227 "decl %1 \n\t" | |
1228 "decl %2 \n\t" | |
1229 "subl %0, %1 \n\t" | |
1230 "subl %0, %2 \n\t" | |
1231 "leal 9b, %3 \n\t" | |
1232 "subl %0, %3 \n\t" | |
1233 | |
1234 | |
1235 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), | |
1236 "=r" (fragmentLengthB) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1237 ); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1238 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1239 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers |
5452 | 1240 fragmentPos=0; |
1241 | |
1242 for(i=0; i<dstW/numSplits; i++) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1243 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1244 int xx=xpos>>16; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1245 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1246 if((i&3) == 0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1247 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1248 int a=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1249 int b=((xpos+xInc)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1250 int c=((xpos+xInc*2)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1251 int d=((xpos+xInc*3)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1252 |
5452 | 1253 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9; |
1254 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9; | |
1255 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9; | |
1256 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9; | |
1257 filterPos[i/2]= xx; | |
1258 | |
1259 if(d+1<4) | |
1260 { | |
1261 int maxShift= 3-(d+1); | |
1262 int shift=0; | |
1263 | |
1264 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1265 |
5452 | 1266 funnyCode[fragmentPos + imm8OfPShufW1B]= |
1267 (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6); | |
1268 funnyCode[fragmentPos + imm8OfPShufW2B]= | |
1269 a | (b<<2) | (c<<4) | (d<<6); | |
1270 | |
1271 if(i+3>=dstW) shift=maxShift; //avoid overread | |
1272 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align | |
1273 | |
1274 if(shift && i>=shift) | |
1275 { | |
1276 funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift; | |
1277 funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift; | |
1278 filterPos[i/2]-=shift; | |
1279 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1280 |
5452 | 1281 fragmentPos+= fragmentLengthB; |
1282 } | |
1283 else | |
1284 { | |
1285 int maxShift= 3-d; | |
1286 int shift=0; | |
1287 | |
1288 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA); | |
1289 | |
1290 funnyCode[fragmentPos + imm8OfPShufW1A]= | |
1291 funnyCode[fragmentPos + imm8OfPShufW2A]= | |
1292 a | (b<<2) | (c<<4) | (d<<6); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1293 |
5452 | 1294 if(i+4>=dstW) shift=maxShift; //avoid overread |
1295 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align | |
1296 | |
1297 if(shift && i>=shift) | |
1298 { | |
1299 funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift; | |
1300 funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift; | |
1301 filterPos[i/2]-=shift; | |
1302 } | |
1303 | |
1304 fragmentPos+= fragmentLengthA; | |
1305 } | |
1306 | |
1307 funnyCode[fragmentPos]= RET; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1308 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1309 xpos+=xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1310 } |
5452 | 1311 filterPos[i/2]= xpos>>16; // needed to jump to the next part |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1312 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1313 #endif // ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1314 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1315 //FIXME remove |
2216 | 1316 void SwScale_Init(){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1317 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1318 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1319 static void globalInit(){ |
2216 | 1320 // generating tables: |
1321 int i; | |
3344 | 1322 for(i=0; i<768; i++){ |
1323 int c= MIN(MAX(i-256, 0), 255); | |
1324 clip_table[i]=c; | |
1325 yuvtab_2568[c]= clip_yuvtab_2568[i]=(0x2568*(c-16))+(256<<13); | |
1326 yuvtab_3343[c]= clip_yuvtab_3343[i]=0x3343*(c-128); | |
1327 yuvtab_0c92[c]= clip_yuvtab_0c92[i]=-0x0c92*(c-128); | |
1328 yuvtab_1a1e[c]= clip_yuvtab_1a1e[i]=-0x1a1e*(c-128); | |
1329 yuvtab_40cf[c]= clip_yuvtab_40cf[i]=0x40cf*(c-128); | |
2216 | 1330 } |
1331 | |
2584 | 1332 for(i=0; i<768; i++) |
1333 { | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1334 int v= clip_table[i]; |
5246 | 1335 clip_table16b[i]= v>>3; |
1336 clip_table16g[i]= (v<<3)&0x07E0; | |
1337 clip_table16r[i]= (v<<8)&0xF800; | |
1338 clip_table15b[i]= v>>3; | |
1339 clip_table15g[i]= (v<<2)&0x03E0; | |
1340 clip_table15r[i]= (v<<7)&0x7C00; | |
2584 | 1341 } |
3344 | 1342 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1343 cpuCaps= gCpuCaps; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1344 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1345 #ifdef RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1346 #ifdef CAN_COMPILE_X86_ASM |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1347 // ordered per speed fasterst first |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1348 if(gCpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1349 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1350 else if(gCpuCaps.has3DNow) |
4281 | 1351 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1352 else if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1353 swScale= swScale_MMX; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1354 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1355 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1356 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1357 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1358 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1359 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1360 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1361 #else //RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1362 #ifdef HAVE_MMX2 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1363 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1364 cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1365 #elif defined (HAVE_3DNOW) |
4281 | 1366 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1367 cpuCaps.hasMMX2 = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1368 #elif defined (HAVE_MMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1369 swScale= swScale_MMX; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1370 cpuCaps.hasMMX2 = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1371 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1372 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1373 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1374 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1375 #endif //!RUNTIME_CPUDETECT |
3126 | 1376 } |
2584 | 1377 |
6492 | 1378 static void PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1379 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ | |
1380 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1381 /* Copy Y plane */ | |
1382 if(dstStride[0]==srcStride[0]) | |
1383 memcpy(dst, src[0], srcSliceH*dstStride[0]); | |
1384 else | |
1385 { | |
1386 int i; | |
1387 uint8_t *srcPtr= src[0]; | |
1388 uint8_t *dstPtr= dst; | |
1389 for(i=0; i<srcSliceH; i++) | |
1390 { | |
1391 memcpy(dstPtr, srcPtr, srcStride[0]); | |
1392 srcPtr+= srcStride[0]; | |
1393 dstPtr+= dstStride[0]; | |
1394 } | |
1395 } | |
1396 dst = dstParam[1] + dstStride[1]*srcSliceY; | |
1397 if(c->srcFormat==IMGFMT_YV12) | |
1398 interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] ); | |
1399 else /* I420 & IYUV */ | |
1400 interleaveBytes( src[2],src[1],dst,c->srcW,srcSliceH,srcStride[2],srcStride[1],dstStride[0] ); | |
1401 } | |
1402 | |
1403 | |
4554 | 1404 /* Warper functions for yuv2bgr */ |
1405 static void planarYuvToBgr(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
4555 | 1406 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1407 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
4554 | 1408 |
1409 if(c->srcFormat==IMGFMT_YV12) | |
4555 | 1410 yuv2rgb( dst,src[0],src[1],src[2],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
4554 | 1411 else /* I420 & IYUV */ |
4555 | 1412 yuv2rgb( dst,src[0],src[2],src[1],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
1413 } | |
1414 | |
6503 | 1415 static void PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
6492 | 1416 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1417 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1418 | |
1419 if(c->srcFormat==IMGFMT_YV12) | |
1420 yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1421 else /* I420 & IYUV */ | |
1422 yv12toyuy2( src[0],src[2],src[1],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1423 } | |
1424 | |
4567 | 1425 static void bgr24to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1426 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1427 | |
1428 if(dstStride[0]*3==srcStride[0]*4) | |
4581 | 1429 rgb24to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 1430 else |
1431 { | |
1432 int i; | |
1433 uint8_t *srcPtr= src[0]; | |
1434 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1435 | |
1436 for(i=0; i<srcSliceH; i++) | |
1437 { | |
4581 | 1438 rgb24to32(srcPtr, dstPtr, c->srcW*3); |
4567 | 1439 srcPtr+= srcStride[0]; |
1440 dstPtr+= dstStride[0]; | |
1441 } | |
1442 } | |
1443 } | |
1444 | |
6492 | 1445 static void bgr24to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1446 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1447 | |
1448 if(dstStride[0]*3==srcStride[0]*2) | |
1449 rgb24to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1450 else | |
1451 { | |
1452 int i; | |
1453 uint8_t *srcPtr= src[0]; | |
1454 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1455 | |
1456 for(i=0; i<srcSliceH; i++) | |
1457 { | |
1458 rgb24to16(srcPtr, dstPtr, c->srcW*3); | |
1459 srcPtr+= srcStride[0]; | |
1460 dstPtr+= dstStride[0]; | |
1461 } | |
1462 } | |
1463 } | |
1464 | |
1465 static void bgr24to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1466 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1467 | |
1468 if(dstStride[0]*3==srcStride[0]*2) | |
1469 rgb24to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1470 else | |
1471 { | |
1472 int i; | |
1473 uint8_t *srcPtr= src[0]; | |
1474 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1475 | |
1476 for(i=0; i<srcSliceH; i++) | |
1477 { | |
1478 rgb24to15(srcPtr, dstPtr, c->srcW*3); | |
1479 srcPtr+= srcStride[0]; | |
1480 dstPtr+= dstStride[0]; | |
1481 } | |
1482 } | |
1483 } | |
1484 | |
4567 | 1485 static void bgr32to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1486 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1487 | |
1488 if(dstStride[0]*4==srcStride[0]*3) | |
4581 | 1489 rgb32to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 1490 else |
1491 { | |
1492 int i; | |
1493 uint8_t *srcPtr= src[0]; | |
1494 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1495 | |
1496 for(i=0; i<srcSliceH; i++) | |
1497 { | |
4581 | 1498 rgb32to24(srcPtr, dstPtr, c->srcW<<2); |
1499 srcPtr+= srcStride[0]; | |
1500 dstPtr+= dstStride[0]; | |
1501 } | |
1502 } | |
1503 } | |
1504 | |
6492 | 1505 static void bgr32to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1506 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1507 | |
1508 if(dstStride[0]*4==srcStride[0]*2) | |
1509 rgb32to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1510 else | |
1511 { | |
1512 int i; | |
1513 uint8_t *srcPtr= src[0]; | |
1514 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1515 | |
1516 for(i=0; i<srcSliceH; i++) | |
1517 { | |
1518 rgb32to16(srcPtr, dstPtr, c->srcW<<2); | |
1519 srcPtr+= srcStride[0]; | |
1520 dstPtr+= dstStride[0]; | |
1521 } | |
1522 } | |
1523 } | |
1524 | |
1525 static void bgr32to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1526 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1527 | |
1528 if(dstStride[0]*4==srcStride[0]*2) | |
1529 rgb32to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1530 else | |
1531 { | |
1532 int i; | |
1533 uint8_t *srcPtr= src[0]; | |
1534 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1535 | |
1536 for(i=0; i<srcSliceH; i++) | |
1537 { | |
1538 rgb32to15(srcPtr, dstPtr, c->srcW<<2); | |
1539 srcPtr+= srcStride[0]; | |
1540 dstPtr+= dstStride[0]; | |
1541 } | |
1542 } | |
1543 } | |
1544 | |
4581 | 1545 static void bgr15to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1546 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1547 | |
1548 if(dstStride[0]==srcStride[0]) | |
1549 rgb15to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1550 else | |
1551 { | |
1552 int i; | |
1553 uint8_t *srcPtr= src[0]; | |
1554 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1555 | |
1556 for(i=0; i<srcSliceH; i++) | |
1557 { | |
1558 rgb15to16(srcPtr, dstPtr, c->srcW<<1); | |
4567 | 1559 srcPtr+= srcStride[0]; |
1560 dstPtr+= dstStride[0]; | |
1561 } | |
1562 } | |
1563 } | |
1564 | |
6492 | 1565 static void bgr15to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1566 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1567 | |
1568 if(dstStride[0]*2==srcStride[0]*3) | |
1569 rgb15to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1570 else | |
1571 { | |
1572 int i; | |
1573 uint8_t *srcPtr= src[0]; | |
1574 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1575 | |
1576 for(i=0; i<srcSliceH; i++) | |
1577 { | |
1578 rgb15to24(srcPtr, dstPtr, c->srcW<<1); | |
1579 srcPtr+= srcStride[0]; | |
1580 dstPtr+= dstStride[0]; | |
1581 } | |
1582 } | |
1583 } | |
1584 | |
1585 static void bgr15to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1586 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1587 | |
1588 if(dstStride[0]*2==srcStride[0]*4) | |
1589 rgb15to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1590 else | |
1591 { | |
1592 int i; | |
1593 uint8_t *srcPtr= src[0]; | |
1594 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1595 | |
1596 for(i=0; i<srcSliceH; i++) | |
1597 { | |
1598 rgb15to32(srcPtr, dstPtr, c->srcW<<1); | |
1599 srcPtr+= srcStride[0]; | |
1600 dstPtr+= dstStride[0]; | |
1601 } | |
1602 } | |
1603 } | |
1604 | |
1605 static void bgr16to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1606 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1607 | |
1608 if(dstStride[0]*2==srcStride[0]*3) | |
1609 rgb16to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1610 else | |
1611 { | |
1612 int i; | |
1613 uint8_t *srcPtr= src[0]; | |
1614 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1615 | |
1616 for(i=0; i<srcSliceH; i++) | |
1617 { | |
1618 rgb16to24(srcPtr, dstPtr, c->srcW<<1); | |
1619 srcPtr+= srcStride[0]; | |
1620 dstPtr+= dstStride[0]; | |
1621 } | |
1622 } | |
1623 } | |
1624 | |
1625 static void bgr16to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1626 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1627 | |
1628 if(dstStride[0]*2==srcStride[0]*4) | |
1629 rgb16to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1630 else | |
1631 { | |
1632 int i; | |
1633 uint8_t *srcPtr= src[0]; | |
1634 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1635 | |
1636 for(i=0; i<srcSliceH; i++) | |
1637 { | |
1638 rgb16to32(srcPtr, dstPtr, c->srcW<<1); | |
1639 srcPtr+= srcStride[0]; | |
1640 dstPtr+= dstStride[0]; | |
1641 } | |
1642 } | |
1643 } | |
1644 | |
4623 | 1645 static void bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1646 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1647 | |
1648 rgb24toyv12( | |
1649 src[0], | |
1650 dst[0]+ srcSliceY *dstStride[0], | |
1651 dst[1]+(srcSliceY>>1)*dstStride[1], | |
1652 dst[2]+(srcSliceY>>1)*dstStride[2], | |
1653 c->srcW, srcSliceH, | |
1654 dstStride[0], dstStride[1], srcStride[0]); | |
1655 } | |
1656 | |
6503 | 1657 /** |
1658 * bring pointers in YUV order instead of YVU | |
1659 */ | |
6517 | 1660 static inline void orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]){ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1661 if(format == IMGFMT_YV12 || format == IMGFMT_YVU9){ |
6503 | 1662 sortedP[0]= p[0]; |
1663 sortedP[1]= p[1]; | |
1664 sortedP[2]= p[2]; | |
6517 | 1665 sortedStride[0]= stride[0]; |
1666 sortedStride[1]= stride[1]; | |
1667 sortedStride[2]= stride[2]; | |
6503 | 1668 } |
6517 | 1669 else if(isPacked(format) || isGray(format)) |
6503 | 1670 { |
1671 sortedP[0]= p[0]; | |
1672 sortedP[1]= | |
1673 sortedP[2]= NULL; | |
6517 | 1674 sortedStride[0]= stride[0]; |
6503 | 1675 sortedStride[1]= |
1676 sortedStride[2]= 0; | |
1677 } | |
1678 else /* I420 */ | |
1679 { | |
1680 sortedP[0]= p[0]; | |
1681 sortedP[1]= p[2]; | |
1682 sortedP[2]= p[1]; | |
6517 | 1683 sortedStride[0]= stride[0]; |
1684 sortedStride[1]= stride[2]; | |
1685 sortedStride[2]= stride[1]; | |
6503 | 1686 } |
1687 } | |
4567 | 1688 |
4555 | 1689 /* unscaled copy like stuff (assumes nearly identical formats) */ |
1690 static void simpleCopy(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY, | |
6503 | 1691 int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){ |
4555 | 1692 |
1693 int srcStride[3]; | |
6503 | 1694 int dstStride[3]; |
4555 | 1695 uint8_t *src[3]; |
1696 uint8_t *dst[3]; | |
1697 | |
6517 | 1698 orderYUV(c->srcFormat, src, srcStride, srcParam, srcStrideParam); |
1699 orderYUV(c->dstFormat, dst, dstStride, dstParam, dstStrideParam); | |
4555 | 1700 |
1701 if(isPacked(c->srcFormat)) | |
1702 { | |
1703 if(dstStride[0]==srcStride[0]) | |
1704 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]); | |
1705 else | |
1706 { | |
1707 int i; | |
1708 uint8_t *srcPtr= src[0]; | |
1709 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
4558 | 1710 int length=0; |
1711 | |
1712 /* universal length finder */ | |
4649 | 1713 while(length+c->srcW <= ABS(dstStride[0]) |
1714 && length+c->srcW <= ABS(srcStride[0])) length+= c->srcW; | |
4558 | 1715 ASSERT(length!=0); |
4555 | 1716 |
1717 for(i=0; i<srcSliceH; i++) | |
1718 { | |
1719 memcpy(dstPtr, srcPtr, length); | |
1720 srcPtr+= srcStride[0]; | |
1721 dstPtr+= dstStride[0]; | |
1722 } | |
1723 } | |
1724 } | |
1725 else | |
6503 | 1726 { /* Planar YUV or gray */ |
4555 | 1727 int plane; |
1728 for(plane=0; plane<3; plane++) | |
1729 { | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1730 int length= plane==0 ? c->srcW : -((-c->srcW )>>c->chrDstHSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1731 int y= plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1732 int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample); |
6503 | 1733 |
1734 if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0) | |
6492 | 1735 { |
6503 | 1736 if(!isGray(c->dstFormat)) |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1737 memset(dst[plane], 128, dstStride[plane]*height); |
6492 | 1738 } |
1739 else | |
1740 { | |
6503 | 1741 if(dstStride[plane]==srcStride[plane]) |
1742 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]); | |
1743 else | |
4555 | 1744 { |
6503 | 1745 int i; |
1746 uint8_t *srcPtr= src[plane]; | |
1747 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y; | |
1748 for(i=0; i<height; i++) | |
1749 { | |
1750 memcpy(dstPtr, srcPtr, length); | |
1751 srcPtr+= srcStride[plane]; | |
1752 dstPtr+= dstStride[plane]; | |
1753 } | |
4555 | 1754 } |
1755 } | |
1756 } | |
1757 } | |
4554 | 1758 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1759 |
6503 | 1760 static int remove_dup_fourcc(int fourcc) |
6492 | 1761 { |
1762 switch(fourcc) | |
1763 { | |
1764 case IMGFMT_IYUV: return IMGFMT_I420; | |
1765 case IMGFMT_Y8 : return IMGFMT_Y800; | |
1766 default: return fourcc; | |
1767 } | |
1768 } | |
1769 | |
6517 | 1770 static void getSubSampleFactors(int *h, int *v, int format){ |
1771 switch(format){ | |
1772 case IMGFMT_YUY2: | |
1773 *h=1; | |
1774 *v=0; | |
1775 break; | |
1776 case IMGFMT_YV12: | |
1777 case IMGFMT_I420: | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1778 case IMGFMT_Y800: //FIXME remove after different subsamplings are fully implemented |
6517 | 1779 *h=1; |
1780 *v=1; | |
1781 break; | |
1782 case IMGFMT_YVU9: | |
1783 *h=2; | |
1784 *v=2; | |
1785 break; | |
1786 default: | |
1787 *h=0; | |
1788 *v=0; | |
1789 break; | |
1790 } | |
1791 } | |
1792 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1793 SwsContext *getSwsContext(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat, int flags, |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1794 SwsFilter *srcFilter, SwsFilter *dstFilter){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1795 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1796 SwsContext *c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1797 int i; |
4554 | 1798 int usesFilter; |
6503 | 1799 int unscaled; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1800 SwsFilter dummyFilter= {NULL, NULL, NULL, NULL}; |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1801 #ifdef ARCH_X86 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1802 if(gCpuCaps.hasMMX) |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1803 asm volatile("emms\n\t"::: "memory"); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1804 #endif |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1805 if(swScale==NULL) globalInit(); |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1806 //srcFormat= IMGFMT_Y800; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1807 //srcFormat= IMGFMT_YVU9; |
4481 | 1808 /* avoid dupplicate Formats, so we dont need to check to much */ |
6492 | 1809 srcFormat = remove_dup_fourcc(srcFormat); |
1810 dstFormat = remove_dup_fourcc(dstFormat); | |
6503 | 1811 |
1812 unscaled = (srcW == dstW && srcH == dstH); | |
1813 | |
1814 if(!isSupportedIn(srcFormat)) | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1815 { |
6503 | 1816 MSG_ERR("swScaler: %s is not supported as input format\n", vo_format_name(srcFormat)); |
1817 return NULL; | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1818 } |
6503 | 1819 if(!isSupportedOut(dstFormat)) |
1820 { | |
1821 MSG_ERR("swScaler: %s is not supported as output format\n", vo_format_name(dstFormat)); | |
1822 return NULL; | |
1823 } | |
1824 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1825 /* sanity check */ |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1826 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 |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1827 { |
6492 | 1828 MSG_ERR("swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1829 srcW, srcH, dstW, dstH); |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1830 return NULL; |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1831 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1832 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1833 if(!dstFilter) dstFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1834 if(!srcFilter) srcFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1835 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1836 c= memalign(64, sizeof(SwsContext)); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1837 memset(c, 0, sizeof(SwsContext)); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1838 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1839 c->srcW= srcW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1840 c->srcH= srcH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1841 c->dstW= dstW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1842 c->dstH= dstH; |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
1843 c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
1844 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1845 c->flags= flags; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1846 c->dstFormat= dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1847 c->srcFormat= srcFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1848 |
4554 | 1849 usesFilter=0; |
1850 if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesFilter=1; | |
1851 if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesFilter=1; | |
1852 if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesFilter=1; | |
1853 if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesFilter=1; | |
1854 if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesFilter=1; | |
1855 if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesFilter=1; | |
1856 if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesFilter=1; | |
1857 if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesFilter=1; | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1858 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1859 getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1860 getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1861 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1862 // reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1863 if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1864 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1865 // drop eery 2. pixel for chroma calculation unless user wants full chroma |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1866 if((isBGR(srcFormat) || isRGB(srcFormat) || srcFormat==IMGFMT_YUY2) && !(flags&SWS_FULL_CHR_V)) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1867 c->chrSrcVSubSample=1; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1868 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1869 // drop eery 2. pixel for chroma calculation unless user wants full chroma |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1870 if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1871 c->chrSrcHSubSample=1; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1872 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1873 c->chrIntHSubSample= c->chrDstHSubSample; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1874 c->chrIntVSubSample= c->chrSrcVSubSample; |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1875 |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1876 // note the -((-x)>>y) is so that we allways round toward +inf |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1877 c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1878 c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1879 c->chrDstW= -((-dstW) >> c->chrDstHSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1880 c->chrDstH= -((-dstH) >> c->chrDstVSubSample); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1881 /* printf("%d %d %d %d / %d %d %d %d //\n", |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1882 c->chrSrcW, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1883 c->chrSrcH, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1884 c->chrDstW, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1885 c->chrDstH, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1886 srcW, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1887 srcH, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1888 dstW, |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1889 dstH);*/ |
4554 | 1890 |
4567 | 1891 /* unscaled special Cases */ |
6503 | 1892 if(unscaled && !usesFilter) |
4554 | 1893 { |
6492 | 1894 /* yv12_to_nv12 */ |
1895 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_NV12) | |
1896 { | |
1897 c->swScale= PlanarToNV12Wrapper; | |
1898 | |
1899 if(flags&SWS_PRINT_INFO) | |
1900 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
1901 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
1902 return c; | |
1903 } | |
1904 /* yv12_to_yuy2 */ | |
1905 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_YUY2) | |
1906 { | |
6503 | 1907 c->swScale= PlanarToYuy2Wrapper; |
6492 | 1908 |
1909 if(flags&SWS_PRINT_INFO) | |
1910 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
1911 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
1912 return c; | |
1913 } | |
4554 | 1914 /* yuv2bgr */ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1915 if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_I420) && isBGR(dstFormat)) |
4554 | 1916 { |
1917 // FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics | |
4791 | 1918 #ifdef WORDS_BIGENDIAN |
5246 | 1919 if(dstFormat==IMGFMT_BGR32) |
1920 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR); | |
1921 else | |
1922 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); | |
4791 | 1923 #else |
4555 | 1924 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); |
4791 | 1925 #endif |
4554 | 1926 c->swScale= planarYuvToBgr; |
4555 | 1927 |
1928 if(flags&SWS_PRINT_INFO) | |
6492 | 1929 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4555 | 1930 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
1931 return c; | |
1932 } | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1933 #if 1 |
4555 | 1934 /* simple copy */ |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1935 if( srcFormat == dstFormat |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1936 || (srcFormat==IMGFMT_YV12 && dstFormat==IMGFMT_I420) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1937 || (srcFormat==IMGFMT_I420 && dstFormat==IMGFMT_YV12) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1938 || (isPlanarYUV(srcFormat) && isGray(dstFormat)) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1939 || (isPlanarYUV(dstFormat) && isGray(srcFormat)) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1940 ) |
4555 | 1941 { |
1942 c->swScale= simpleCopy; | |
1943 | |
4554 | 1944 if(flags&SWS_PRINT_INFO) |
6492 | 1945 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4554 | 1946 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
1947 return c; | |
1948 } | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1949 #endif |
4567 | 1950 /* bgr32to24 & rgb32to24*/ |
1951 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR24) | |
1952 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB24)) | |
1953 { | |
1954 c->swScale= bgr32to24Wrapper; | |
1955 | |
1956 if(flags&SWS_PRINT_INFO) | |
6492 | 1957 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
1958 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
1959 return c; | |
1960 } | |
1961 | |
1962 /* bgr32to16 & rgb32to16*/ | |
1963 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR16) | |
1964 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB16)) | |
1965 { | |
1966 c->swScale= bgr32to16Wrapper; | |
1967 | |
1968 if(flags&SWS_PRINT_INFO) | |
1969 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
1970 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
1971 return c; | |
1972 } | |
1973 | |
1974 /* bgr32to15 & rgb32to15*/ | |
1975 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR15) | |
1976 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB15)) | |
1977 { | |
1978 c->swScale= bgr32to15Wrapper; | |
1979 | |
1980 if(flags&SWS_PRINT_INFO) | |
1981 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 1982 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
1983 return c; | |
1984 } | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
1985 |
4567 | 1986 /* bgr24to32 & rgb24to32*/ |
1987 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR32) | |
1988 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB32)) | |
1989 { | |
1990 c->swScale= bgr24to32Wrapper; | |
1991 | |
1992 if(flags&SWS_PRINT_INFO) | |
6492 | 1993 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
1994 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
1995 return c; | |
1996 } | |
1997 | |
1998 /* bgr24to16 & rgb24to16*/ | |
1999 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR16) | |
2000 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB16)) | |
2001 { | |
2002 c->swScale= bgr24to16Wrapper; | |
2003 | |
2004 if(flags&SWS_PRINT_INFO) | |
2005 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2006 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2007 return c; | |
2008 } | |
2009 | |
2010 /* bgr24to15 & rgb24to15*/ | |
2011 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR15) | |
2012 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB15)) | |
2013 { | |
2014 c->swScale= bgr24to15Wrapper; | |
2015 | |
2016 if(flags&SWS_PRINT_INFO) | |
2017 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 2018 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2019 return c; | |
2020 } | |
4581 | 2021 |
2022 /* bgr15to16 */ | |
2023 if(srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR16) | |
2024 { | |
2025 c->swScale= bgr15to16Wrapper; | |
2026 | |
2027 if(flags&SWS_PRINT_INFO) | |
6492 | 2028 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2029 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2030 return c; | |
2031 } | |
2032 | |
2033 /* bgr15to24 */ | |
2034 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR24) | |
2035 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB24)) | |
2036 { | |
2037 c->swScale= bgr15to24Wrapper; | |
2038 | |
2039 if(flags&SWS_PRINT_INFO) | |
2040 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2041 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2042 return c; | |
2043 } | |
2044 | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2045 #if 0 //segfaults |
6492 | 2046 /* bgr15to32 */ |
2047 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR32) | |
2048 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB32)) | |
2049 { | |
2050 c->swScale= bgr15to32Wrapper; | |
2051 | |
2052 if(flags&SWS_PRINT_INFO) | |
2053 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2054 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2055 return c; | |
2056 } | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2057 #endif |
6492 | 2058 /* bgr16to24 */ |
2059 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR24) | |
2060 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB24)) | |
2061 { | |
2062 c->swScale= bgr16to24Wrapper; | |
2063 | |
2064 if(flags&SWS_PRINT_INFO) | |
2065 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2066 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2067 return c; | |
2068 } | |
2069 | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2070 #if 0 //segfaults |
6492 | 2071 /* bgr16to32 */ |
2072 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR32) | |
2073 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB32)) | |
2074 { | |
2075 c->swScale= bgr16to32Wrapper; | |
2076 | |
2077 if(flags&SWS_PRINT_INFO) | |
2078 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4581 | 2079 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2080 return c; | |
2081 } | |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2082 #endif |
4623 | 2083 /* bgr24toYV12 */ |
2084 if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12) | |
2085 { | |
2086 c->swScale= bgr24toyv12Wrapper; | |
2087 | |
2088 if(flags&SWS_PRINT_INFO) | |
6492 | 2089 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4623 | 2090 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2091 return c; | |
2092 } | |
4554 | 2093 } |
2094 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2095 if(cpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2096 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2097 c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2098 if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2099 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2100 if(flags&SWS_PRINT_INFO) |
6492 | 2101 MSG_INFO("SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2102 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2103 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2104 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2105 c->canMMX2BeUsed=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2106 |
4467 | 2107 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW; |
2108 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH; | |
2109 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2110 // match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2111 // but only for the FAST_BILINEAR mode otherwise do correct scaling |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2112 // n-2 is the last chrominance sample available |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2113 // this is not perfect, but noone shuld notice the difference, the more correct variant |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2114 // would be like the vertical one, but that would require some special code for the |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2115 // first and last pixel |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2116 if(flags&SWS_FAST_BILINEAR) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2117 { |
4467 | 2118 if(c->canMMX2BeUsed) |
2119 { | |
2120 c->lumXInc+= 20; | |
2121 c->chrXInc+= 20; | |
2122 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2123 //we dont use the x86asm scaler if mmx is available |
4467 | 2124 else if(cpuCaps.hasMMX) |
2125 { | |
2126 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20; | |
2127 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20; | |
2128 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2129 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2130 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2131 /* precalculate horizontal scaler filter coefficients */ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2132 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2133 const int filterAlign= cpuCaps.hasMMX ? 4 : 1; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2134 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2135 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2136 srcW , dstW, filterAlign, 1<<14, flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2137 srcFilter->lumH, dstFilter->lumH); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2138 initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc, |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2139 c->chrSrcW, c->chrDstW, filterAlign, 1<<14, flags, |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2140 srcFilter->chrH, dstFilter->chrH); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2141 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2142 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2143 // cant downscale !!! |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2144 if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2145 { |
5452 | 2146 c->lumMmx2Filter = (int16_t*)memalign(8, (dstW /8+8)*sizeof(int16_t)); |
2147 c->chrMmx2Filter = (int16_t*)memalign(8, (c->chrDstW /4+8)*sizeof(int16_t)); | |
2148 c->lumMmx2FilterPos= (int32_t*)memalign(8, (dstW /2/8+8)*sizeof(int32_t)); | |
2149 c->chrMmx2FilterPos= (int32_t*)memalign(8, (c->chrDstW/2/4+8)*sizeof(int32_t)); | |
2150 | |
2151 initMMX2HScaler( dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8); | |
2152 initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2153 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2154 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2155 } // Init Horizontal stuff |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2156 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2157 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2158 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2159 /* precalculate vertical scaler filter coefficients */ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2160 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2161 srcH , dstH, 1, (1<<12)-4, flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2162 srcFilter->lumV, dstFilter->lumV); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2163 initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc, |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2164 c->chrSrcH, c->chrDstH, 1, (1<<12)-4, flags, |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2165 srcFilter->chrV, dstFilter->chrV); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2166 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2167 // Calculate Buffer Sizes so that they wont run out while handling these damn slices |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2168 c->vLumBufSize= c->vLumFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2169 c->vChrBufSize= c->vChrFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2170 for(i=0; i<dstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2171 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2172 int chrI= i*c->chrDstH / dstH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2173 int nextSlice= MAX(c->vLumFilterPos[i ] + c->vLumFilterSize - 1, |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2174 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample)); |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2175 nextSlice&= ~3; // Slices start at boundaries which are divisable through 4 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2176 if(c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2177 c->vLumBufSize= nextSlice - c->vLumFilterPos[i ]; |
6532
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2178 if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample)) |
9834d9980c45
yvu9 support (other planar yuv formats with other chroma subsamplings should be trivial to add, if they had a IMGFMT)
michael
parents:
6520
diff
changeset
|
2179 c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2180 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2181 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2182 // allocate pixbufs (we use dynamic allocation because otherwise we would need to |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2183 c->lumPixBuf= (int16_t**)memalign(4, c->vLumBufSize*2*sizeof(int16_t*)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2184 c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*)); |
4419 | 2185 //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) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2186 for(i=0; i<c->vLumBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2187 c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= (uint16_t*)memalign(8, 4000); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2188 for(i=0; i<c->vChrBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2189 c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= (uint16_t*)memalign(8, 8000); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2190 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2191 //try to avoid drawing green stuff between the right end and the stride end |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2192 for(i=0; i<c->vLumBufSize; i++) memset(c->lumPixBuf[i], 0, 4000); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2193 for(i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2194 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2195 ASSERT(c->chrDstH <= dstH) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2196 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2197 // pack filter data for mmx code |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2198 if(cpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2199 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2200 c->lumMmxFilter= (int16_t*)memalign(8, c->vLumFilterSize* dstH*4*sizeof(int16_t)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2201 c->chrMmxFilter= (int16_t*)memalign(8, c->vChrFilterSize*c->chrDstH*4*sizeof(int16_t)); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2202 for(i=0; i<c->vLumFilterSize*dstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2203 c->lumMmxFilter[4*i]=c->lumMmxFilter[4*i+1]=c->lumMmxFilter[4*i+2]=c->lumMmxFilter[4*i+3]= |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2204 c->vLumFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2205 for(i=0; i<c->vChrFilterSize*c->chrDstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2206 c->chrMmxFilter[4*i]=c->chrMmxFilter[4*i+1]=c->chrMmxFilter[4*i+2]=c->chrMmxFilter[4*i+3]= |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2207 c->vChrFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2208 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2209 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2210 if(flags&SWS_PRINT_INFO) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2211 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2212 #ifdef DITHER1XBPP |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2213 char *dither= " dithered"; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2214 #else |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2215 char *dither= ""; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2216 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2217 if(flags&SWS_FAST_BILINEAR) |
6492 | 2218 MSG_INFO("\nSwScaler: FAST_BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2219 else if(flags&SWS_BILINEAR) |
6492 | 2220 MSG_INFO("\nSwScaler: BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2221 else if(flags&SWS_BICUBIC) |
6492 | 2222 MSG_INFO("\nSwScaler: BICUBIC scaler, "); |
4467 | 2223 else if(flags&SWS_X) |
6492 | 2224 MSG_INFO("\nSwScaler: Experimental scaler, "); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2225 else if(flags&SWS_POINT) |
6492 | 2226 MSG_INFO("\nSwScaler: Nearest Neighbor / POINT scaler, "); |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
2227 else if(flags&SWS_AREA) |
6492 | 2228 MSG_INFO("\nSwScaler: Area Averageing scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2229 else |
6492 | 2230 MSG_INFO("\nSwScaler: ehh flags invalid?! "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2231 |
6492 | 2232 if(dstFormat==IMGFMT_BGR15 || dstFormat==IMGFMT_BGR16) |
2233 MSG_INFO("from %s to%s %s ", | |
2234 vo_format_name(srcFormat), dither, vo_format_name(dstFormat)); | |
2235 else | |
2236 MSG_INFO("from %s to %s ", | |
2237 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2238 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2239 if(cpuCaps.hasMMX2) |
6492 | 2240 MSG_INFO("using MMX2\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2241 else if(cpuCaps.has3DNow) |
6492 | 2242 MSG_INFO("using 3DNOW\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2243 else if(cpuCaps.hasMMX) |
6492 | 2244 MSG_INFO("using MMX\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2245 else |
6492 | 2246 MSG_INFO("using C\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2247 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2248 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2249 if((flags & SWS_PRINT_INFO) && verbose) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2250 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2251 if(cpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2252 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2253 if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR)) |
6492 | 2254 MSG_V("SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2255 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2256 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2257 if(c->hLumFilterSize==4) |
6492 | 2258 MSG_V("SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2259 else if(c->hLumFilterSize==8) |
6492 | 2260 MSG_V("SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2261 else |
6492 | 2262 MSG_V("SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2263 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2264 if(c->hChrFilterSize==4) |
6492 | 2265 MSG_V("SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2266 else if(c->hChrFilterSize==8) |
6492 | 2267 MSG_V("SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2268 else |
6492 | 2269 MSG_V("SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2270 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2271 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2272 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2273 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2274 #ifdef ARCH_X86 |
6492 | 2275 MSG_V("SwScaler: using X86-Asm scaler for horizontal scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2276 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2277 if(flags & SWS_FAST_BILINEAR) |
6492 | 2278 MSG_V("SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2279 else |
6492 | 2280 MSG_V("SwScaler: using C scaler for horizontal scaling\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2281 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2282 } |
4419 | 2283 if(isPlanarYUV(dstFormat)) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2284 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2285 if(c->vLumFilterSize==1) |
6492 | 2286 MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2287 else |
6492 | 2288 MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2289 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2290 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2291 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2292 if(c->vLumFilterSize==1 && c->vChrFilterSize==2) |
6492 | 2293 MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n" |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2294 "SwScaler: 2-tap scaler for vertical chrominance scaling (BGR)\n",cpuCaps.hasMMX ? "MMX" : "C"); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2295 else if(c->vLumFilterSize==2 && c->vChrFilterSize==2) |
6492 | 2296 MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2297 else |
6492 | 2298 MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2299 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2300 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2301 if(dstFormat==IMGFMT_BGR24) |
6492 | 2302 MSG_V("SwScaler: using %s YV12->BGR24 Converter\n", |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2303 cpuCaps.hasMMX2 ? "MMX2" : (cpuCaps.hasMMX ? "MMX" : "C")); |
4304 | 2304 else if(dstFormat==IMGFMT_BGR32) |
6492 | 2305 MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2306 else if(dstFormat==IMGFMT_BGR16) |
6492 | 2307 MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2308 else if(dstFormat==IMGFMT_BGR15) |
6492 | 2309 MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2310 |
6492 | 2311 MSG_V("SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2312 } |
4467 | 2313 if((flags & SWS_PRINT_INFO) && verbose>1) |
2314 { | |
6492 | 2315 MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2316 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); |
6492 | 2317 MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2318 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc); |
2319 } | |
4554 | 2320 |
2321 c->swScale= swScale; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2322 return c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2323 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2324 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2325 /** |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2326 * returns a normalized gaussian curve used to filter stuff |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2327 * quality=3 is high quality, lowwer is lowwer quality |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2328 */ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2329 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2330 SwsVector *getGaussianVec(double variance, double quality){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2331 const int length= (int)(variance*quality + 0.5) | 1; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2332 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2333 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2334 double middle= (length-1)*0.5; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2335 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2336 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2337 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2338 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2339 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2340 for(i=0; i<length; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2341 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2342 double dist= i-middle; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2343 coeff[i]= exp( -dist*dist/(2*variance*variance) ) / sqrt(2*variance*PI); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2344 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2345 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2346 normalizeVec(vec, 1.0); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2347 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2348 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2349 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2350 |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2351 SwsVector *getConstVec(double c, int length){ |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2352 int i; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2353 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2354 SwsVector *vec= malloc(sizeof(SwsVector)); |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2355 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2356 vec->coeff= coeff; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2357 vec->length= length; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2358 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2359 for(i=0; i<length; i++) |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2360 coeff[i]= c; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2361 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2362 return vec; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2363 } |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2364 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2365 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2366 SwsVector *getIdentityVec(void){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2367 double *coeff= memalign(sizeof(double), sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2368 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2369 coeff[0]= 1.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2370 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2371 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2372 vec->length= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2373 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2374 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2375 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2376 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2377 void normalizeVec(SwsVector *a, double height){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2378 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2379 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2380 double inv; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2381 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2382 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2383 sum+= a->coeff[i]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2384 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2385 inv= height/sum; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2386 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2387 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2388 a->coeff[i]*= height; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2389 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2390 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2391 void scaleVec(SwsVector *a, double scalar){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2392 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2393 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2394 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2395 a->coeff[i]*= scalar; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2396 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2397 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2398 static SwsVector *getConvVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2399 int length= a->length + b->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2400 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2401 int i, j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2402 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2403 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2404 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2405 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2406 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2407 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2408 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2409 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2410 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2411 for(j=0; j<b->length; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2412 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2413 coeff[i+j]+= a->coeff[i]*b->coeff[j]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2414 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2415 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2416 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2417 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2418 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2419 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2420 static SwsVector *sumVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2421 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2422 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2423 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2424 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2425 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2426 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2427 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2428 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2429 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2430 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2431 for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2432 for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]+= b->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2433 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2434 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2435 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2436 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2437 static SwsVector *diffVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2438 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2439 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2440 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2441 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2442 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2443 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2444 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2445 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2446 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2447 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2448 for(i=0; i<a->length; i++) coeff[i + (length-1)/2 - (a->length-1)/2]+= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2449 for(i=0; i<b->length; i++) coeff[i + (length-1)/2 - (b->length-1)/2]-= b->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2450 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2451 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2452 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2453 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2454 /* shift left / or right if "shift" is negative */ |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2455 static SwsVector *getShiftedVec(SwsVector *a, int shift){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2456 int length= a->length + ABS(shift)*2; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2457 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2458 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2459 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2460 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2461 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2462 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2463 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2464 for(i=0; i<length; i++) coeff[i]= 0.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2465 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2466 for(i=0; i<a->length; i++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2467 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2468 coeff[i + (length-1)/2 - (a->length-1)/2 - shift]= a->coeff[i]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2469 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2470 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2471 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2472 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2473 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2474 void shiftVec(SwsVector *a, int shift){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2475 SwsVector *shifted= getShiftedVec(a, shift); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2476 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2477 a->coeff= shifted->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2478 a->length= shifted->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2479 free(shifted); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2480 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2481 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2482 void addVec(SwsVector *a, SwsVector *b){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2483 SwsVector *sum= sumVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2484 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2485 a->coeff= sum->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2486 a->length= sum->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2487 free(sum); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2488 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2489 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2490 void subVec(SwsVector *a, SwsVector *b){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2491 SwsVector *diff= diffVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2492 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2493 a->coeff= diff->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2494 a->length= diff->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2495 free(diff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2496 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2497 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2498 void convVec(SwsVector *a, SwsVector *b){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2499 SwsVector *conv= getConvVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2500 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2501 a->coeff= conv->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2502 a->length= conv->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2503 free(conv); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2504 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2505 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2506 SwsVector *cloneVec(SwsVector *a){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2507 double *coeff= memalign(sizeof(double), a->length*sizeof(double)); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2508 int i; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2509 SwsVector *vec= malloc(sizeof(SwsVector)); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2510 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2511 vec->coeff= coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2512 vec->length= a->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2513 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2514 for(i=0; i<a->length; i++) coeff[i]= a->coeff[i]; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2515 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2516 return vec; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2517 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2518 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2519 void printVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2520 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2521 double max=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2522 double min=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2523 double range; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2524 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2525 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2526 if(a->coeff[i]>max) max= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2527 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2528 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2529 if(a->coeff[i]<min) min= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2530 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2531 range= max - min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2532 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2533 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2534 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2535 int x= (int)((a->coeff[i]-min)*60.0/range +0.5); |
6492 | 2536 MSG_DBG2("%1.3f ", a->coeff[i]); |
2537 for(;x>0; x--) MSG_DBG2(" "); | |
2538 MSG_DBG2("|\n"); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2539 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2540 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2541 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2542 void freeVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2543 if(!a) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2544 if(a->coeff) free(a->coeff); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2545 a->coeff=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2546 a->length=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2547 free(a); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2548 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2549 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2550 void freeSwsContext(SwsContext *c){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2551 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2552 if(!c) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2553 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2554 if(c->lumPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2555 { |
4419 | 2556 for(i=0; i<c->vLumBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2557 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2558 if(c->lumPixBuf[i]) free(c->lumPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2559 c->lumPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2560 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2561 free(c->lumPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2562 c->lumPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2563 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2564 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2565 if(c->chrPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2566 { |
4419 | 2567 for(i=0; i<c->vChrBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2568 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2569 if(c->chrPixBuf[i]) free(c->chrPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2570 c->chrPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2571 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2572 free(c->chrPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2573 c->chrPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2574 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2575 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2576 if(c->vLumFilter) free(c->vLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2577 c->vLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2578 if(c->vChrFilter) free(c->vChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2579 c->vChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2580 if(c->hLumFilter) free(c->hLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2581 c->hLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2582 if(c->hChrFilter) free(c->hChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2583 c->hChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2584 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2585 if(c->vLumFilterPos) free(c->vLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2586 c->vLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2587 if(c->vChrFilterPos) free(c->vChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2588 c->vChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2589 if(c->hLumFilterPos) free(c->hLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2590 c->hLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2591 if(c->hChrFilterPos) free(c->hChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2592 c->hChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2593 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2594 if(c->lumMmxFilter) free(c->lumMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2595 c->lumMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2596 if(c->chrMmxFilter) free(c->chrMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2597 c->chrMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2598 |
5452 | 2599 if(c->lumMmx2Filter) free(c->lumMmx2Filter); |
2600 c->lumMmx2Filter=NULL; | |
2601 if(c->chrMmx2Filter) free(c->chrMmx2Filter); | |
2602 c->chrMmx2Filter=NULL; | |
2603 if(c->lumMmx2FilterPos) free(c->lumMmx2FilterPos); | |
2604 c->lumMmx2FilterPos=NULL; | |
2605 if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos); | |
2606 c->chrMmx2FilterPos=NULL; | |
2607 | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2608 free(c); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2609 } |
4281 | 2610 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2611 |