Mercurial > mplayer.hg
annotate postproc/swscale.c @ 6578:3727eb94a783
use unified yuv2rgb init
{RGB,BGR}{1,4,8,15,16,24,32} output supported
author | michael |
---|---|
date | Thu, 27 Jun 2002 18:00:47 +0000 |
parents | 5aa2dbf4d300 |
children | 5059143aa020 |
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 /* |
6535
29f94210a67f
IF09 is alias for YVU9 (actually it has extra 4th plane containing MC change
arpi
parents:
6532
diff
changeset
|
20 supported Input formats: YV12, I420/IYUV, YUY2, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09 |
6578 | 21 supported output formats: YV12, I420/IYUV, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09 |
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) | |
6578 | 106 #define isRGB(x) (((x)&IMGFMT_RGB_MASK)==IMGFMT_RGB) |
107 #define isBGR(x) (((x)&IMGFMT_BGR_MASK)==IMGFMT_BGR) | |
4481 | 108 #define isSupportedIn(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 || (x)==IMGFMT_YUY2 \ |
4580 | 109 || (x)==IMGFMT_BGR32|| (x)==IMGFMT_BGR24|| (x)==IMGFMT_BGR16|| (x)==IMGFMT_BGR15\ |
4558 | 110 || (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
|
111 || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) |
4481 | 112 #define isSupportedOut(x) ((x)==IMGFMT_YV12 || (x)==IMGFMT_I420 \ |
6578 | 113 || isRGB(x) || isBGR(x)\ |
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
|
114 || (x)==IMGFMT_Y800 || (x)==IMGFMT_YVU9) |
6503 | 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; |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
206 #endif |
2264
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
207 |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
208 // clipping helper table for C implementations: |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
209 static unsigned char clip_table[768]; |
7851375ea156
increased precission of s_xinc s_xinc2 (needed for the mmx2 bugfix)
michael
parents:
2237
diff
changeset
|
210 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
211 //global sws_flags from the command line |
4421 | 212 int sws_flags=2; |
3272 | 213 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
214 //global srcFilter |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
215 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
|
216 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
217 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
|
218 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
|
219 int sws_chr_vshift= 0; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
220 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
|
221 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
|
222 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
|
223 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
224 /* 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
|
225 (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
|
226 static CpuCaps cpuCaps; |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
227 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
228 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
|
229 int srcSliceH, uint8_t* dst[], int dstStride[])=NULL; |
2469 | 230 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
231 static SwsVector *getConvVec(SwsVector *a, SwsVector *b); |
6540 | 232 static inline void orderYUV(int format, uint8_t * sortedP[], int sortedStride[], uint8_t * p[], int stride[]); |
6578 | 233 void *yuv2rgb_c_init (unsigned bpp, int mode, void *table_rV[256], void *table_gU[256], int table_gV[256], void *table_bU[256]); |
6540 | 234 |
6578 | 235 extern const uint8_t dither_2x2_4[2][8]; |
236 extern const uint8_t dither_2x2_8[2][8]; | |
237 extern const uint8_t dither_8x8_32[8][8]; | |
238 extern const uint8_t dither_8x8_73[8][8]; | |
239 extern const uint8_t dither_8x8_220[8][8]; | |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
240 |
3126 | 241 #ifdef CAN_COMPILE_X86_ASM |
2671 | 242 void in_asm_used_var_warning_killer() |
243 { | |
3272 | 244 volatile int i= yCoeff+vrCoeff+ubCoeff+vgCoeff+ugCoeff+bF8+bFC+w400+w80+w10+ |
6554 | 245 bm00001111+bm00000111+bm11111000+b16Mask+g16Mask+r16Mask+b15Mask+g15Mask+r15Mask+ |
4481 | 246 M24A+M24B+M24C+w02 + b5Dither+g5Dither+r5Dither+g6Dither+dither4[0]+dither8[0]+bm01010101; |
2671 | 247 if(i) i=0; |
248 } | |
249 #endif | |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
250 |
6517 | 251 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
|
252 IMGFMT_YVU9, |
6517 | 253 IMGFMT_YV12, |
254 //IMGFMT_IYUV, | |
255 IMGFMT_I420, | |
256 IMGFMT_BGR15, | |
257 IMGFMT_BGR16, | |
258 IMGFMT_BGR24, | |
259 IMGFMT_BGR32, | |
260 //IMGFMT_Y8, | |
261 IMGFMT_Y800, | |
262 //IMGFMT_YUY2, | |
263 0 | |
264 }; | |
265 | |
266 static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1, int stride2, int w, int h){ | |
267 int x,y; | |
268 uint64_t ssd=0; | |
269 | |
270 for(y=0; y<h; y++){ | |
271 for(x=0; x<w; x++){ | |
272 int d= src1[x + y*stride1] - src2[x + y*stride2]; | |
273 ssd+= d*d; | |
274 } | |
275 } | |
276 return ssd; | |
277 } | |
278 | |
279 // test by ref -> src -> dst -> out & compare out against ref | |
280 // ref & out are YV12 | |
281 static void doTest(uint8_t *ref[3], int refStride[3], int w, int h, int srcFormat, int dstFormat, | |
282 int srcW, int srcH, int dstW, int dstH, int flags){ | |
283 uint8_t *src[3]; | |
284 uint8_t *dst[3]; | |
285 uint8_t *out[3]; | |
286 int srcStride[3], dstStride[3]; | |
287 int i; | |
288 uint64_t ssdY, ssdU, ssdV; | |
289 SwsContext *srcContext, *dstContext, *outContext; | |
290 | |
291 for(i=0; i<3; i++){ | |
292 srcStride[i]= srcW*4; | |
293 dstStride[i]= dstW*4; | |
294 src[i]= malloc(srcStride[i]*srcH); | |
295 dst[i]= malloc(dstStride[i]*dstH); | |
296 out[i]= malloc(refStride[i]*h); | |
297 } | |
298 | |
299 srcContext= getSwsContext(w, h, IMGFMT_YV12, srcW, srcH, srcFormat, flags, NULL, NULL); | |
300 dstContext= getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, NULL, NULL); | |
301 outContext= getSwsContext(dstW, dstH, dstFormat, w, h, IMGFMT_YV12, flags, NULL, NULL); | |
302 if(srcContext==NULL ||dstContext==NULL ||outContext==NULL){ | |
303 printf("Failed allocating swsContext\n"); | |
304 goto end; | |
305 } | |
306 // printf("test %X %X %X -> %X %X %X\n", (int)ref[0], (int)ref[1], (int)ref[2], | |
307 // (int)src[0], (int)src[1], (int)src[2]); | |
308 | |
309 srcContext->swScale(srcContext, ref, refStride, 0, h , src, srcStride); | |
310 dstContext->swScale(dstContext, src, srcStride, 0, srcH, dst, dstStride); | |
311 outContext->swScale(outContext, dst, dstStride, 0, dstH, out, refStride); | |
312 | |
313 ssdY= getSSD(ref[0], out[0], refStride[0], refStride[0], w, h); | |
314 ssdU= getSSD(ref[1], out[1], refStride[1], refStride[1], (w+1)>>1, (h+1)>>1); | |
315 ssdV= getSSD(ref[2], out[2], refStride[2], refStride[2], (w+1)>>1, (h+1)>>1); | |
316 | |
317 if(isGray(srcFormat) || isGray(dstFormat)) ssdU=ssdV=0; //FIXME check that output is really gray | |
318 | |
319 ssdY/= w*h; | |
320 ssdU/= w*h/4; | |
321 ssdV/= w*h/4; | |
322 | |
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
|
323 if(ssdY>100 || ssdU>50 || ssdV>50){ |
6517 | 324 printf(" %s %dx%d -> %s %4dx%4d flags=%2d SSD=%5lld,%5lld,%5lld\n", |
325 vo_format_name(srcFormat), srcW, srcH, | |
326 vo_format_name(dstFormat), dstW, dstH, | |
327 flags, | |
328 ssdY, ssdU, ssdV); | |
329 } | |
330 | |
331 end: | |
332 | |
333 freeSwsContext(srcContext); | |
334 freeSwsContext(dstContext); | |
335 freeSwsContext(outContext); | |
336 | |
337 for(i=0; i<3; i++){ | |
338 free(src[i]); | |
339 free(dst[i]); | |
340 free(out[i]); | |
341 } | |
342 } | |
343 | |
344 static void selfTest(uint8_t *src[3], int stride[3], int w, int h){ | |
345 int srcFormat, dstFormat, srcFormatIndex, dstFormatIndex; | |
346 int srcW, srcH, dstW, dstH; | |
347 int flags; | |
348 | |
349 for(srcFormatIndex=0; ;srcFormatIndex++){ | |
350 srcFormat= testFormat[srcFormatIndex]; | |
351 if(!srcFormat) break; | |
352 for(dstFormatIndex=0; ;dstFormatIndex++){ | |
353 dstFormat= testFormat[dstFormatIndex]; | |
354 if(!dstFormat) break; | |
355 if(!isSupportedOut(dstFormat)) continue; | |
356 | |
357 srcW= w+w/3; | |
358 srcH= h+h/3; | |
359 for(dstW=w; dstW<w*2; dstW+= dstW/3){ | |
360 for(dstH=h; dstH<h*2; dstH+= dstH/3){ | |
361 for(flags=1; flags<33; flags*=2) | |
362 doTest(src, stride, w, h, srcFormat, dstFormat, | |
363 srcW, srcH, dstW, dstH, flags); | |
364 } | |
365 } | |
366 } | |
367 } | |
368 } | |
369 | |
6540 | 370 static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, |
3352 | 371 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, |
6540 | 372 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW) |
3352 | 373 { |
374 //FIXME Optimize (just quickly writen not opti..) | |
375 int i; | |
6540 | 376 for(i=0; i<dstW; i++) |
3352 | 377 { |
378 int val=0; | |
379 int j; | |
380 for(j=0; j<lumFilterSize; j++) | |
381 val += lumSrc[j][i] * lumFilter[j]; | |
382 | |
383 dest[i]= MIN(MAX(val>>19, 0), 255); | |
384 } | |
385 | |
386 if(uDest != NULL) | |
6540 | 387 for(i=0; i<chrDstW; i++) |
3352 | 388 { |
389 int u=0; | |
390 int v=0; | |
391 int j; | |
3641 | 392 for(j=0; j<chrFilterSize; j++) |
3352 | 393 { |
394 u += chrSrc[j][i] * chrFilter[j]; | |
395 v += chrSrc[j][i + 2048] * chrFilter[j]; | |
396 } | |
397 | |
398 uDest[i]= MIN(MAX(u>>19, 0), 255); | |
399 vDest[i]= MIN(MAX(v>>19, 0), 255); | |
400 } | |
401 } | |
402 | |
6578 | 403 #define YSCALE_YUV_2_RGBX_C(type) \ |
404 for(i=0; i<(dstW>>1); i++){\ | |
405 int j;\ | |
406 int Y1=0;\ | |
407 int Y2=0;\ | |
408 int U=0;\ | |
409 int V=0;\ | |
410 type *r, *b, *g;\ | |
411 const int i2= 2*i;\ | |
412 \ | |
413 for(j=0; j<lumFilterSize; j++)\ | |
414 {\ | |
415 Y1 += lumSrc[j][i2] * lumFilter[j];\ | |
416 Y2 += lumSrc[j][i2+1] * lumFilter[j];\ | |
417 }\ | |
418 for(j=0; j<chrFilterSize; j++)\ | |
419 {\ | |
420 U += chrSrc[j][i] * chrFilter[j];\ | |
421 V += chrSrc[j][i+2048] * chrFilter[j];\ | |
422 }\ | |
423 Y1>>=19;\ | |
424 Y2>>=19;\ | |
425 U >>=19;\ | |
426 V >>=19;\ | |
427 if((Y1|Y2|U|V)&256)\ | |
428 {\ | |
429 if(Y1>255) Y1=255;\ | |
430 else if(Y1<0)Y1=0;\ | |
431 if(Y2>255) Y2=255;\ | |
432 else if(Y2<0)Y2=0;\ | |
433 if(U>255) U=255;\ | |
434 else if(U<0) U=0;\ | |
435 if(V>255) V=255;\ | |
436 else if(V<0) V=0;\ | |
437 }\ | |
438 r = c->table_rV[V];\ | |
439 g = c->table_gU[U] + c->table_gV[V];\ | |
440 b = c->table_bU[U];\ | |
3352 | 441 |
6578 | 442 #define YSCALE_YUV_2_RGB2_C(type) \ |
443 for(i=0; i<(dstW>>1); i++){\ | |
444 const int i2= 2*i;\ | |
445 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19;\ | |
446 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;\ | |
447 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19;\ | |
448 int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\ | |
449 type *r, *b, *g;\ | |
450 r = c->table_rV[V];\ | |
451 g = c->table_gU[U] + c->table_gV[V];\ | |
452 b = c->table_bU[U];\ | |
3352 | 453 |
6578 | 454 #define YSCALE_YUV_2_RGB1_C(type) \ |
455 for(i=0; i<(dstW>>1); i++){\ | |
456 const int i2= 2*i;\ | |
457 int Y1= buf0[i2 ]>>7;\ | |
458 int Y2= buf0[i2+1]>>7;\ | |
459 int U= (uvbuf1[i ])>>7;\ | |
460 int V= (uvbuf1[i+2048])>>7;\ | |
461 type *r, *b, *g;\ | |
462 r = c->table_rV[V];\ | |
463 g = c->table_gU[U] + c->table_gV[V];\ | |
464 b = c->table_bU[U];\ | |
3352 | 465 |
6578 | 466 #define YSCALE_YUV_2_RGB1B_C(type) \ |
467 for(i=0; i<(dstW>>1); i++){\ | |
468 const int i2= 2*i;\ | |
469 int Y1= buf0[i2 ]>>7;\ | |
470 int Y2= buf0[i2+1]>>7;\ | |
471 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\ | |
472 int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\ | |
473 type *r, *b, *g;\ | |
474 r = c->table_rV[V];\ | |
475 g = c->table_gU[U] + c->table_gV[V];\ | |
476 b = c->table_bU[U];\ | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
477 |
6578 | 478 #define YSCALE_YUV_2_ANYRGB_C(func)\ |
479 switch(c->dstFormat)\ | |
480 {\ | |
481 case IMGFMT_BGR32:\ | |
482 case IMGFMT_RGB32:\ | |
483 func(uint32_t)\ | |
484 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\ | |
485 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\ | |
486 } \ | |
487 break;\ | |
488 case IMGFMT_RGB24:\ | |
489 func(uint8_t)\ | |
490 ((uint8_t*)dest)[0]= r[Y1];\ | |
491 ((uint8_t*)dest)[1]= g[Y1];\ | |
492 ((uint8_t*)dest)[2]= b[Y1];\ | |
493 ((uint8_t*)dest)[3]= r[Y2];\ | |
494 ((uint8_t*)dest)[4]= g[Y2];\ | |
495 ((uint8_t*)dest)[5]= b[Y2];\ | |
496 ((uint8_t*)dest)+=6;\ | |
497 }\ | |
498 break;\ | |
499 case IMGFMT_BGR24:\ | |
500 func(uint8_t)\ | |
501 ((uint8_t*)dest)[0]= b[Y1];\ | |
502 ((uint8_t*)dest)[1]= g[Y1];\ | |
503 ((uint8_t*)dest)[2]= r[Y1];\ | |
504 ((uint8_t*)dest)[3]= b[Y2];\ | |
505 ((uint8_t*)dest)[4]= g[Y2];\ | |
506 ((uint8_t*)dest)[5]= r[Y2];\ | |
507 ((uint8_t*)dest)+=6;\ | |
508 }\ | |
509 break;\ | |
510 case IMGFMT_RGB16:\ | |
511 case IMGFMT_BGR16:\ | |
512 {\ | |
513 const int dr1= dither_2x2_8[y&1 ][0];\ | |
514 const int dg1= dither_2x2_4[y&1 ][0];\ | |
515 const int db1= dither_2x2_8[(y&1)^1][0];\ | |
516 const int dr2= dither_2x2_8[y&1 ][1];\ | |
517 const int dg2= dither_2x2_4[y&1 ][1];\ | |
518 const int db2= dither_2x2_8[(y&1)^1][1];\ | |
519 func(uint16_t)\ | |
520 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\ | |
521 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\ | |
522 }\ | |
523 }\ | |
524 break;\ | |
525 case IMGFMT_RGB15:\ | |
526 case IMGFMT_BGR15:\ | |
527 {\ | |
528 const int dr1= dither_2x2_8[y&1 ][0];\ | |
529 const int dg1= dither_2x2_8[y&1 ][1];\ | |
530 const int db1= dither_2x2_8[(y&1)^1][0];\ | |
531 const int dr2= dither_2x2_8[y&1 ][1];\ | |
532 const int dg2= dither_2x2_8[y&1 ][0];\ | |
533 const int db2= dither_2x2_8[(y&1)^1][1];\ | |
534 func(uint16_t)\ | |
535 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\ | |
536 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\ | |
537 }\ | |
538 }\ | |
539 break;\ | |
540 case IMGFMT_RGB8:\ | |
541 case IMGFMT_BGR8:\ | |
542 {\ | |
543 const uint8_t * const d64= dither_8x8_73[y&7];\ | |
544 const uint8_t * const d32= dither_8x8_32[y&7];\ | |
545 func(uint8_t)\ | |
546 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\ | |
547 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\ | |
548 }\ | |
549 }\ | |
550 break;\ | |
551 case IMGFMT_RGB4:\ | |
552 case IMGFMT_BGR4:\ | |
553 {\ | |
554 const uint8_t * const d64= dither_8x8_73 [y&7];\ | |
555 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
556 func(uint8_t)\ | |
557 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\ | |
558 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\ | |
559 }\ | |
560 }\ | |
561 break;\ | |
562 case IMGFMT_RGB1:\ | |
563 case IMGFMT_BGR1:\ | |
564 {\ | |
565 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
566 uint8_t *g= c->table_gU[128] + c->table_gV[128];\ | |
567 for(i=0; i<dstW-7; i+=8){\ | |
568 int acc;\ | |
569 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\ | |
570 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\ | |
571 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\ | |
572 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\ | |
573 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\ | |
574 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\ | |
575 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\ | |
576 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\ | |
577 ((uint8_t*)dest)[0]= acc;\ | |
578 ((uint8_t*)dest)++;\ | |
579 }\ | |
580 \ | |
581 /*\ | |
582 ((uint8_t*)dest)-= dstW>>4;\ | |
583 {\ | |
584 int acc=0;\ | |
585 int left=0;\ | |
586 static int top[1024];\ | |
587 static int last_new[1024][1024];\ | |
588 static int last_in3[1024][1024];\ | |
589 static int drift[1024][1024];\ | |
590 int topLeft=0;\ | |
591 int shift=0;\ | |
592 int count=0;\ | |
593 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
594 int error_new=0;\ | |
595 int error_in3=0;\ | |
596 int f=0;\ | |
597 \ | |
598 for(i=dstW>>1; i<dstW; i++){\ | |
599 int in= ((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19);\ | |
600 int in2 = (76309 * (in - 16) + 32768) >> 16;\ | |
601 int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\ | |
602 int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\ | |
603 + (last_new[y][i] - in3)*f/256;\ | |
604 int new= old> 128 ? 255 : 0;\ | |
605 \ | |
606 error_new+= ABS(last_new[y][i] - new);\ | |
607 error_in3+= ABS(last_in3[y][i] - in3);\ | |
608 f= error_new - error_in3*4;\ | |
609 if(f<0) f=0;\ | |
610 if(f>256) f=256;\ | |
611 \ | |
612 topLeft= top[i];\ | |
613 left= top[i]= old - new;\ | |
614 last_new[y][i]= new;\ | |
615 last_in3[y][i]= in3;\ | |
616 \ | |
617 acc+= acc + (new&1);\ | |
618 if((i&7)==6){\ | |
619 ((uint8_t*)dest)[0]= acc;\ | |
620 ((uint8_t*)dest)++;\ | |
621 }\ | |
622 }\ | |
623 }\ | |
624 */\ | |
625 }\ | |
626 break;\ | |
627 }\ | |
3352 | 628 |
629 | |
6578 | 630 static inline void yuv2rgbXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, |
631 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
632 uint8_t *dest, int dstW, int y) | |
633 { | |
634 int i; | |
635 switch(c->dstFormat) | |
3352 | 636 { |
6578 | 637 case IMGFMT_RGB32: |
638 case IMGFMT_BGR32: | |
639 YSCALE_YUV_2_RGBX_C(uint32_t) | |
640 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1]; | |
641 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2]; | |
642 } | |
643 break; | |
644 case IMGFMT_RGB24: | |
645 YSCALE_YUV_2_RGBX_C(uint8_t) | |
646 ((uint8_t*)dest)[0]= r[Y1]; | |
647 ((uint8_t*)dest)[1]= g[Y1]; | |
648 ((uint8_t*)dest)[2]= b[Y1]; | |
649 ((uint8_t*)dest)[3]= r[Y2]; | |
650 ((uint8_t*)dest)[4]= g[Y2]; | |
651 ((uint8_t*)dest)[5]= b[Y2]; | |
652 ((uint8_t*)dest)+=6; | |
653 } | |
654 break; | |
655 case IMGFMT_BGR24: | |
656 YSCALE_YUV_2_RGBX_C(uint8_t) | |
657 ((uint8_t*)dest)[0]= b[Y1]; | |
658 ((uint8_t*)dest)[1]= g[Y1]; | |
659 ((uint8_t*)dest)[2]= r[Y1]; | |
660 ((uint8_t*)dest)[3]= b[Y2]; | |
661 ((uint8_t*)dest)[4]= g[Y2]; | |
662 ((uint8_t*)dest)[5]= r[Y2]; | |
663 ((uint8_t*)dest)+=6; | |
664 } | |
665 break; | |
666 case IMGFMT_RGB16: | |
667 case IMGFMT_BGR16: | |
668 { | |
669 const int dr1= dither_2x2_8[y&1 ][0]; | |
670 const int dg1= dither_2x2_4[y&1 ][0]; | |
671 const int db1= dither_2x2_8[(y&1)^1][0]; | |
672 const int dr2= dither_2x2_8[y&1 ][1]; | |
673 const int dg2= dither_2x2_4[y&1 ][1]; | |
674 const int db2= dither_2x2_8[(y&1)^1][1]; | |
675 YSCALE_YUV_2_RGBX_C(uint16_t) | |
676 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1]; | |
677 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2]; | |
678 } | |
679 } | |
680 break; | |
681 case IMGFMT_RGB15: | |
682 case IMGFMT_BGR15: | |
683 { | |
684 const int dr1= dither_2x2_8[y&1 ][0]; | |
685 const int dg1= dither_2x2_8[y&1 ][1]; | |
686 const int db1= dither_2x2_8[(y&1)^1][0]; | |
687 const int dr2= dither_2x2_8[y&1 ][1]; | |
688 const int dg2= dither_2x2_8[y&1 ][0]; | |
689 const int db2= dither_2x2_8[(y&1)^1][1]; | |
690 YSCALE_YUV_2_RGBX_C(uint16_t) | |
691 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1]; | |
692 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2]; | |
693 } | |
694 } | |
695 break; | |
696 case IMGFMT_RGB8: | |
697 case IMGFMT_BGR8: | |
698 { | |
699 const uint8_t * const d64= dither_8x8_73[y&7]; | |
700 const uint8_t * const d32= dither_8x8_32[y&7]; | |
701 YSCALE_YUV_2_RGBX_C(uint8_t) | |
702 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]]; | |
703 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]]; | |
704 } | |
705 } | |
706 break; | |
707 case IMGFMT_RGB4: | |
708 case IMGFMT_BGR4: | |
709 { | |
710 const uint8_t * const d64= dither_8x8_73 [y&7]; | |
711 const uint8_t * const d128=dither_8x8_220[y&7]; | |
712 YSCALE_YUV_2_RGBX_C(uint8_t) | |
713 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]; | |
714 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]]; | |
715 } | |
716 } | |
717 break; | |
718 case IMGFMT_RGB1: | |
719 case IMGFMT_BGR1: | |
720 { | |
721 const uint8_t * const d128=dither_8x8_220[y&7]; | |
722 uint8_t *g= c->table_gU[128] + c->table_gV[128]; | |
723 int acc=0; | |
724 for(i=0; i<dstW-1; i+=2){ | |
725 int j; | |
726 int Y1=0; | |
727 int Y2=0; | |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
728 |
6578 | 729 for(j=0; j<lumFilterSize; j++) |
730 { | |
731 Y1 += lumSrc[j][i] * lumFilter[j]; | |
732 Y2 += lumSrc[j][i+1] * lumFilter[j]; | |
733 } | |
734 Y1>>=19; | |
735 Y2>>=19; | |
736 if((Y1|Y2)&256) | |
737 { | |
738 if(Y1>255) Y1=255; | |
739 else if(Y1<0)Y1=0; | |
740 if(Y2>255) Y2=255; | |
741 else if(Y2<0)Y2=0; | |
742 } | |
743 acc+= acc + g[Y1+d128[(i+0)&7]]; | |
744 acc+= acc + g[Y2+d128[(i+1)&7]]; | |
745 if((i&7)==6){ | |
746 ((uint8_t*)dest)[0]= acc; | |
747 ((uint8_t*)dest)++; | |
748 } | |
3352 | 749 } |
750 } | |
6578 | 751 break; |
3352 | 752 } |
753 } | |
754 | |
755 | |
3126 | 756 //Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one |
757 //Plain C versions | |
3152 | 758 #if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) |
759 #define COMPILE_C | |
760 #endif | |
761 | |
762 #ifdef CAN_COMPILE_X86_ASM | |
763 | |
764 #if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
765 #define COMPILE_MMX | |
766 #endif | |
767 | |
768 #if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT) | |
769 #define COMPILE_MMX2 | |
770 #endif | |
771 | |
772 #if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
773 #define COMPILE_3DNOW | |
774 #endif | |
775 #endif //CAN_COMPILE_X86_ASM | |
776 | |
777 #undef HAVE_MMX | |
778 #undef HAVE_MMX2 | |
779 #undef HAVE_3DNOW | |
780 | |
781 #ifdef COMPILE_C | |
3126 | 782 #undef HAVE_MMX |
783 #undef HAVE_MMX2 | |
784 #undef HAVE_3DNOW | |
785 #define RENAME(a) a ## _C | |
786 #include "swscale_template.c" | |
3152 | 787 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
788 |
3126 | 789 #ifdef CAN_COMPILE_X86_ASM |
2576 | 790 |
3126 | 791 //X86 versions |
792 /* | |
793 #undef RENAME | |
794 #undef HAVE_MMX | |
795 #undef HAVE_MMX2 | |
796 #undef HAVE_3DNOW | |
797 #define ARCH_X86 | |
798 #define RENAME(a) a ## _X86 | |
799 #include "swscale_template.c" | |
800 */ | |
801 //MMX versions | |
3152 | 802 #ifdef COMPILE_MMX |
3126 | 803 #undef RENAME |
804 #define HAVE_MMX | |
805 #undef HAVE_MMX2 | |
806 #undef HAVE_3DNOW | |
807 #define RENAME(a) a ## _MMX | |
808 #include "swscale_template.c" | |
3152 | 809 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
810 |
3126 | 811 //MMX2 versions |
3152 | 812 #ifdef COMPILE_MMX2 |
3126 | 813 #undef RENAME |
814 #define HAVE_MMX | |
815 #define HAVE_MMX2 | |
816 #undef HAVE_3DNOW | |
817 #define RENAME(a) a ## _MMX2 | |
818 #include "swscale_template.c" | |
3152 | 819 #endif |
2469 | 820 |
3126 | 821 //3DNOW versions |
3152 | 822 #ifdef COMPILE_3DNOW |
3126 | 823 #undef RENAME |
824 #define HAVE_MMX | |
825 #undef HAVE_MMX2 | |
826 #define HAVE_3DNOW | |
827 #define RENAME(a) a ## _3DNow | |
828 #include "swscale_template.c" | |
3152 | 829 #endif |
2469 | 830 |
3126 | 831 #endif //CAN_COMPILE_X86_ASM |
2469 | 832 |
3126 | 833 // 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
|
834 |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
835 |
4419 | 836 // 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
|
837 // will use sws_flags from the command line |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
838 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
|
839 int srcSliceH, uint8_t* dst[], int dstStride, int dstbpp, |
3209 | 840 int srcW, int srcH, int dstW, int dstH){ |
2216 | 841 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
842 static SwsContext *context=NULL; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
843 int dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
844 int dstStride3[3]= {dstStride, dstStride>>1, dstStride>>1}; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
845 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
846 switch(dstbpp) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
847 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
848 case 8 : dstFormat= IMGFMT_Y8; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
849 case 12: dstFormat= IMGFMT_YV12; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
850 case 15: dstFormat= IMGFMT_BGR15; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
851 case 16: dstFormat= IMGFMT_BGR16; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
852 case 24: dstFormat= IMGFMT_BGR24; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
853 case 32: dstFormat= IMGFMT_BGR32; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
854 default: return; |
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 |
4419 | 857 if(!context) context=getSwsContextFromCmdLine(srcW, srcH, IMGFMT_YV12, dstW, dstH, dstFormat); |
858 | |
4555 | 859 context->swScale(context, src, srcStride, srcSliceY, srcSliceH, dst, dstStride3); |
4419 | 860 } |
861 | |
6540 | 862 void swsGetFlagsAndFilterFromCmdLine(int *flags, SwsFilter **srcFilterParam, SwsFilter **dstFilterParam) |
4419 | 863 { |
864 static int firstTime=1; | |
6540 | 865 *flags=0; |
4419 | 866 |
867 #ifdef ARCH_X86 | |
868 if(gCpuCaps.hasMMX) | |
869 asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions) | |
870 #endif | |
871 if(firstTime) | |
872 { | |
873 firstTime=0; | |
6540 | 874 *flags= SWS_PRINT_INFO; |
4419 | 875 } |
6540 | 876 else if(verbose>1) *flags= SWS_PRINT_INFO; |
4419 | 877 |
878 if(src_filter.lumH) freeVec(src_filter.lumH); | |
879 if(src_filter.lumV) freeVec(src_filter.lumV); | |
880 if(src_filter.chrH) freeVec(src_filter.chrH); | |
881 if(src_filter.chrV) freeVec(src_filter.chrV); | |
882 | |
883 if(sws_lum_gblur!=0.0){ | |
884 src_filter.lumH= getGaussianVec(sws_lum_gblur, 3.0); | |
885 src_filter.lumV= getGaussianVec(sws_lum_gblur, 3.0); | |
886 }else{ | |
887 src_filter.lumH= getIdentityVec(); | |
888 src_filter.lumV= getIdentityVec(); | |
889 } | |
890 | |
891 if(sws_chr_gblur!=0.0){ | |
892 src_filter.chrH= getGaussianVec(sws_chr_gblur, 3.0); | |
893 src_filter.chrV= getGaussianVec(sws_chr_gblur, 3.0); | |
894 }else{ | |
895 src_filter.chrH= getIdentityVec(); | |
896 src_filter.chrV= getIdentityVec(); | |
897 } | |
898 | |
899 if(sws_chr_sharpen!=0.0){ | |
900 SwsVector *g= getConstVec(-1.0, 3); | |
901 SwsVector *id= getConstVec(10.0/sws_chr_sharpen, 1); | |
902 g->coeff[1]=2.0; | |
903 addVec(id, g); | |
904 convVec(src_filter.chrH, id); | |
905 convVec(src_filter.chrV, id); | |
906 freeVec(g); | |
907 freeVec(id); | |
908 } | |
909 | |
910 if(sws_lum_sharpen!=0.0){ | |
911 SwsVector *g= getConstVec(-1.0, 3); | |
912 SwsVector *id= getConstVec(10.0/sws_lum_sharpen, 1); | |
913 g->coeff[1]=2.0; | |
914 addVec(id, g); | |
915 convVec(src_filter.lumH, id); | |
916 convVec(src_filter.lumV, id); | |
917 freeVec(g); | |
918 freeVec(id); | |
919 } | |
920 | |
921 if(sws_chr_hshift) | |
922 shiftVec(src_filter.chrH, sws_chr_hshift); | |
923 | |
924 if(sws_chr_vshift) | |
925 shiftVec(src_filter.chrV, sws_chr_vshift); | |
926 | |
927 normalizeVec(src_filter.chrH, 1.0); | |
928 normalizeVec(src_filter.chrV, 1.0); | |
929 normalizeVec(src_filter.lumH, 1.0); | |
930 normalizeVec(src_filter.lumV, 1.0); | |
931 | |
932 if(verbose > 1) printVec(src_filter.chrH); | |
933 if(verbose > 1) printVec(src_filter.lumH); | |
934 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
935 switch(sws_flags) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
936 { |
6540 | 937 case 0: *flags|= SWS_FAST_BILINEAR; break; |
938 case 1: *flags|= SWS_BILINEAR; break; | |
939 case 2: *flags|= SWS_BICUBIC; break; | |
940 case 3: *flags|= SWS_X; break; | |
941 case 4: *flags|= SWS_POINT; break; | |
942 case 5: *flags|= SWS_AREA; break; | |
943 default:*flags|= SWS_BILINEAR; break; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
944 } |
6540 | 945 |
946 *srcFilterParam= &src_filter; | |
947 *dstFilterParam= NULL; | |
948 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
949 |
6540 | 950 // will use sws_flags & src_filter (from cmd line) |
951 SwsContext *getSwsContextFromCmdLine(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat) | |
952 { | |
953 int flags; | |
954 SwsFilter *dstFilterParam, *srcFilterParam; | |
955 swsGetFlagsAndFilterFromCmdLine(&flags, &srcFilterParam, &dstFilterParam); | |
956 | |
957 return getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, srcFilterParam, dstFilterParam); | |
4419 | 958 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
959 |
2216 | 960 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
961 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
|
962 int srcW, int dstW, int filterAlign, int one, int flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
963 SwsVector *srcFilter, SwsVector *dstFilter) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
964 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
965 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
966 int filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
967 int filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
968 int minFilterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
969 double *filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
970 double *filter2=NULL; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
971 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
972 if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
973 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
|
974 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
975 |
4563
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
976 // Note the +1 is for the MMXscaler which reads over the end |
4419 | 977 *filterPos = (int16_t*)memalign(8, (dstW+1)*sizeof(int16_t)); |
978 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
979 if(ABS(xInc - 0x10000) <10) // unscaled |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
980 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
981 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
982 filterSize= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
983 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
984 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
|
985 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
986 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
987 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
988 filter[i*filterSize]=1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
989 (*filterPos)[i]=i; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
990 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
991 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
992 } |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
993 else if(flags&SWS_POINT) // lame looking point sampling mode |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
994 { |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
995 int i; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
996 int xDstInSrc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
997 filterSize= 1; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
998 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
999 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1000 xDstInSrc= xInc/2 - 0x8000; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1001 for(i=0; i<dstW; i++) |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1002 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
1003 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1004 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1005 (*filterPos)[i]= xx; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1006 filter[i]= 1.0; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1007 xDstInSrc+= xInc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1008 } |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1009 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1010 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
|
1011 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1012 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1013 int xDstInSrc; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1014 if (flags&SWS_BICUBIC) filterSize= 4; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1015 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
|
1016 else filterSize= 2; // SWS_BILINEAR / SWS_AREA |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1017 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
|
1018 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1019 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1020 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1021 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
1022 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
|
1023 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1024 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1025 (*filterPos)[i]= xx; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1026 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1027 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1028 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
|
1029 double y1,y2,y3,y4; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1030 double A= -0.6; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1031 if(flags & SWS_BICUBIC){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1032 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1033 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
|
1034 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
|
1035 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
|
1036 y4 = ( + A*d*d - A*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1037 }else{ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1038 // cubic interpolation (derived it myself) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1039 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
|
1040 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
|
1041 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
|
1042 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
|
1043 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1044 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1045 filter[i*filterSize + 0]= y1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1046 filter[i*filterSize + 1]= y2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1047 filter[i*filterSize + 2]= y3; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1048 filter[i*filterSize + 3]= y4; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1049 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1050 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1051 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1052 //Bilinear upscale / linear interpolate / Area averaging |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1053 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1054 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1055 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
|
1056 double coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1057 if(coeff<0) coeff=0; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1058 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1059 xx++; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1060 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1061 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1062 xDstInSrc+= xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1063 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1064 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1065 else // downscale |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1066 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1067 int xDstInSrc; |
4806 | 1068 ASSERT(dstW <= srcW) |
1069 | |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1070 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
|
1071 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
|
1072 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
|
1073 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
|
1074 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
|
1075 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1076 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1077 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1078 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1079 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
|
1080 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1081 (*filterPos)[i]= xx; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1082 for(j=0; j<filterSize; j++) |
4276
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 double d= ABS((xx<<16) - xDstInSrc)/(double)xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1085 double coeff; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1086 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1087 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1088 double A= -0.75; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1089 // d*=2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1090 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1091 if(d<1.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1092 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
|
1093 else if(d<2.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1094 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
|
1095 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1096 coeff=0.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1097 } |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1098 else if(flags & SWS_AREA) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1099 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1100 double srcPixelSize= (1<<16)/(double)xInc; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1101 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
|
1102 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
|
1103 else coeff=0.0; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1104 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1105 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1106 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1107 coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1108 if(coeff<0) coeff=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1109 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1110 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1111 xx++; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1112 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1113 xDstInSrc+= xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1114 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1115 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1116 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1117 /* apply src & dst Filter to filter -> filter2 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1118 free(filter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1119 */ |
4806 | 1120 ASSERT(filterSize>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1121 filter2Size= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1122 if(srcFilter) filter2Size+= srcFilter->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1123 if(dstFilter) filter2Size+= dstFilter->length - 1; |
4806 | 1124 ASSERT(filter2Size>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1125 filter2= (double*)memalign(8, filter2Size*dstW*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1126 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1127 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1128 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1129 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1130 SwsVector scaleFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1131 SwsVector *outVec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1132 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1133 scaleFilter.coeff= filter + i*filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1134 scaleFilter.length= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1135 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1136 if(srcFilter) outVec= getConvVec(srcFilter, &scaleFilter); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1137 else outVec= &scaleFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1138 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1139 ASSERT(outVec->length == filter2Size) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1140 //FIXME dstFilter |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1141 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1142 for(j=0; j<outVec->length; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1143 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1144 filter2[i*filter2Size + j]= outVec->coeff[j]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1145 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1146 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1147 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1148 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1149 if(outVec != &scaleFilter) freeVec(outVec); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1150 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1151 free(filter); filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1152 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1153 /* 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
|
1154 // 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
|
1155 minFilterSize= 0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1156 for(i=dstW-1; i>=0; i--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1157 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1158 int min= filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1159 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1160 double cutOff=0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1161 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1162 /* 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
|
1163 for(j=0; j<filter2Size; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1164 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1165 int k; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1166 cutOff += ABS(filter2[i*filter2Size]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1167 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1168 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1169 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1170 /* preserve Monotonicity because the core cant handle the filter otherwise */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1171 if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1172 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1173 // Move filter coeffs left |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1174 for(k=1; k<filter2Size; k++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1175 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1176 filter2[i*filter2Size + k - 1]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1177 (*filterPos)[i]++; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1178 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1179 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1180 cutOff=0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1181 /* count near zeros on the right */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1182 for(j=filter2Size-1; j>0; j--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1183 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1184 cutOff += ABS(filter2[i*filter2Size + j]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1185 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1186 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1187 min--; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1188 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1189 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1190 if(min>minFilterSize) minFilterSize= min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1191 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1192 |
4806 | 1193 ASSERT(minFilterSize > 0) |
4419 | 1194 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1)); |
4806 | 1195 ASSERT(filterSize > 0) |
4419 | 1196 filter= (double*)memalign(8, filterSize*dstW*sizeof(double)); |
1197 *outFilterSize= filterSize; | |
1198 | |
5937 | 1199 if(flags&SWS_PRINT_INFO) |
6492 | 1200 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
|
1201 /* try to reduce the filter-size (step2 reduce it) */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1202 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1203 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1204 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1205 |
4419 | 1206 for(j=0; j<filterSize; j++) |
1207 { | |
1208 if(j>=filter2Size) filter[i*filterSize + j]= 0.0; | |
1209 else filter[i*filterSize + j]= filter2[i*filter2Size + j]; | |
1210 } | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1211 } |
4419 | 1212 free(filter2); filter2=NULL; |
1213 | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1214 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1215 //FIXME try to align filterpos if possible |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1216 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1217 //fix borders |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1218 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1219 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1220 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1221 if((*filterPos)[i] < 0) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1222 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1223 // Move filter coeffs left to compensate for filterPos |
4419 | 1224 for(j=1; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1225 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1226 int left= MAX(j + (*filterPos)[i], 0); |
4419 | 1227 filter[i*filterSize + left] += filter[i*filterSize + j]; |
1228 filter[i*filterSize + j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1229 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1230 (*filterPos)[i]= 0; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1231 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1232 |
4419 | 1233 if((*filterPos)[i] + filterSize > srcW) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1234 { |
4419 | 1235 int shift= (*filterPos)[i] + filterSize - srcW; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1236 // Move filter coeffs right to compensate for filterPos |
4419 | 1237 for(j=filterSize-2; j>=0; j--) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1238 { |
4419 | 1239 int right= MIN(j + shift, filterSize-1); |
1240 filter[i*filterSize +right] += filter[i*filterSize +j]; | |
1241 filter[i*filterSize +j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1242 } |
4419 | 1243 (*filterPos)[i]= srcW - filterSize; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1244 } |
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 |
4419 | 1247 // Note the +1 is for the MMXscaler which reads over the end |
1248 *outFilter= (int16_t*)memalign(8, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
1249 memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1250 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1251 /* Normalize & Store in outFilter */ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1252 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1253 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1254 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1255 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1256 double scale= one; |
4419 | 1257 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1258 { |
4419 | 1259 sum+= filter[i*filterSize + j]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1260 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1261 scale/= sum; |
4419 | 1262 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1263 { |
4419 | 1264 (*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
|
1265 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1266 } |
4563
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1267 |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1268 (*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
|
1269 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
|
1270 { |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1271 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
|
1272 (*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
|
1273 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1274 |
4419 | 1275 free(filter); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1276 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1277 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1278 #ifdef ARCH_X86 |
5452 | 1279 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
|
1280 { |
5452 | 1281 uint8_t *fragmentA; |
1282 int imm8OfPShufW1A; | |
1283 int imm8OfPShufW2A; | |
1284 int fragmentLengthA; | |
1285 uint8_t *fragmentB; | |
1286 int imm8OfPShufW1B; | |
1287 int imm8OfPShufW2B; | |
1288 int fragmentLengthB; | |
1289 int fragmentPos; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1290 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1291 int xpos, i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1292 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1293 // create an optimized horizontal scaling routine |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1294 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1295 //code fragment |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1296 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1297 asm volatile( |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1298 "jmp 9f \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1299 // Begin |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1300 "0: \n\t" |
5452 | 1301 "movq (%%edx, %%eax), %%mm3 \n\t" |
1302 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1303 "movd 1(%%ecx, %%esi), %%mm1 \n\t" | |
1304 "punpcklbw %%mm7, %%mm1 \n\t" | |
1305 "punpcklbw %%mm7, %%mm0 \n\t" | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1306 "pshufw $0xFF, %%mm1, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1307 "1: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1308 "pshufw $0xFF, %%mm0, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1309 "2: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1310 "psubw %%mm1, %%mm0 \n\t" |
5452 | 1311 "movl 8(%%ebx, %%eax), %%esi \n\t" |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1312 "pmullw %%mm3, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1313 "psllw $7, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1314 "paddw %%mm1, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1315 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1316 "movq %%mm0, (%%edi, %%eax) \n\t" |
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 "addl $8, %%eax \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1319 // End |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1320 "9: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1321 // "int $3\n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1322 "leal 0b, %0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1323 "leal 1b, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1324 "leal 2b, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1325 "decl %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1326 "decl %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1327 "subl %0, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1328 "subl %0, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1329 "leal 9b, %3 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1330 "subl %0, %3 \n\t" |
5452 | 1331 |
1332 | |
1333 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), | |
1334 "=r" (fragmentLengthA) | |
1335 ); | |
1336 | |
1337 asm volatile( | |
1338 "jmp 9f \n\t" | |
1339 // Begin | |
1340 "0: \n\t" | |
1341 "movq (%%edx, %%eax), %%mm3 \n\t" | |
1342 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1343 "punpcklbw %%mm7, %%mm0 \n\t" | |
1344 "pshufw $0xFF, %%mm0, %%mm1 \n\t" | |
1345 "1: \n\t" | |
1346 "pshufw $0xFF, %%mm0, %%mm0 \n\t" | |
1347 "2: \n\t" | |
1348 "psubw %%mm1, %%mm0 \n\t" | |
1349 "movl 8(%%ebx, %%eax), %%esi \n\t" | |
1350 "pmullw %%mm3, %%mm0 \n\t" | |
1351 "psllw $7, %%mm1 \n\t" | |
1352 "paddw %%mm1, %%mm0 \n\t" | |
1353 | |
1354 "movq %%mm0, (%%edi, %%eax) \n\t" | |
1355 | |
1356 "addl $8, %%eax \n\t" | |
1357 // End | |
1358 "9: \n\t" | |
1359 // "int $3\n\t" | |
1360 "leal 0b, %0 \n\t" | |
1361 "leal 1b, %1 \n\t" | |
1362 "leal 2b, %2 \n\t" | |
1363 "decl %1 \n\t" | |
1364 "decl %2 \n\t" | |
1365 "subl %0, %1 \n\t" | |
1366 "subl %0, %2 \n\t" | |
1367 "leal 9b, %3 \n\t" | |
1368 "subl %0, %3 \n\t" | |
1369 | |
1370 | |
1371 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), | |
1372 "=r" (fragmentLengthB) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1373 ); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1374 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1375 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers |
5452 | 1376 fragmentPos=0; |
1377 | |
1378 for(i=0; i<dstW/numSplits; i++) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1379 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1380 int xx=xpos>>16; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1381 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1382 if((i&3) == 0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1383 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1384 int a=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1385 int b=((xpos+xInc)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1386 int c=((xpos+xInc*2)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1387 int d=((xpos+xInc*3)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1388 |
5452 | 1389 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9; |
1390 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9; | |
1391 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9; | |
1392 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9; | |
1393 filterPos[i/2]= xx; | |
1394 | |
1395 if(d+1<4) | |
1396 { | |
1397 int maxShift= 3-(d+1); | |
1398 int shift=0; | |
1399 | |
1400 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1401 |
5452 | 1402 funnyCode[fragmentPos + imm8OfPShufW1B]= |
1403 (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6); | |
1404 funnyCode[fragmentPos + imm8OfPShufW2B]= | |
1405 a | (b<<2) | (c<<4) | (d<<6); | |
1406 | |
1407 if(i+3>=dstW) shift=maxShift; //avoid overread | |
1408 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align | |
1409 | |
1410 if(shift && i>=shift) | |
1411 { | |
1412 funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift; | |
1413 funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift; | |
1414 filterPos[i/2]-=shift; | |
1415 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1416 |
5452 | 1417 fragmentPos+= fragmentLengthB; |
1418 } | |
1419 else | |
1420 { | |
1421 int maxShift= 3-d; | |
1422 int shift=0; | |
1423 | |
1424 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA); | |
1425 | |
1426 funnyCode[fragmentPos + imm8OfPShufW1A]= | |
1427 funnyCode[fragmentPos + imm8OfPShufW2A]= | |
1428 a | (b<<2) | (c<<4) | (d<<6); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1429 |
5452 | 1430 if(i+4>=dstW) shift=maxShift; //avoid overread |
1431 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align | |
1432 | |
1433 if(shift && i>=shift) | |
1434 { | |
1435 funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift; | |
1436 funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift; | |
1437 filterPos[i/2]-=shift; | |
1438 } | |
1439 | |
1440 fragmentPos+= fragmentLengthA; | |
1441 } | |
1442 | |
1443 funnyCode[fragmentPos]= RET; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1444 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1445 xpos+=xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1446 } |
5452 | 1447 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
|
1448 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1449 #endif // ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1450 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1451 //FIXME remove |
2216 | 1452 void SwScale_Init(){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1453 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1454 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1455 static void globalInit(){ |
2216 | 1456 // generating tables: |
1457 int i; | |
3344 | 1458 for(i=0; i<768; i++){ |
1459 int c= MIN(MAX(i-256, 0), 255); | |
1460 clip_table[i]=c; | |
2584 | 1461 } |
3344 | 1462 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1463 cpuCaps= gCpuCaps; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1464 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1465 #ifdef RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1466 #ifdef CAN_COMPILE_X86_ASM |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1467 // ordered per speed fasterst first |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1468 if(gCpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1469 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1470 else if(gCpuCaps.has3DNow) |
4281 | 1471 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1472 else if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1473 swScale= swScale_MMX; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1474 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1475 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1476 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1477 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1478 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1479 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1480 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1481 #else //RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1482 #ifdef HAVE_MMX2 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1483 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1484 cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1485 #elif defined (HAVE_3DNOW) |
4281 | 1486 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1487 cpuCaps.hasMMX2 = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1488 #elif defined (HAVE_MMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1489 swScale= swScale_MMX; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1490 cpuCaps.hasMMX2 = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1491 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1492 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1493 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1494 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1495 #endif //!RUNTIME_CPUDETECT |
3126 | 1496 } |
2584 | 1497 |
6492 | 1498 static void PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1499 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ | |
1500 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1501 /* Copy Y plane */ | |
1502 if(dstStride[0]==srcStride[0]) | |
1503 memcpy(dst, src[0], srcSliceH*dstStride[0]); | |
1504 else | |
1505 { | |
1506 int i; | |
1507 uint8_t *srcPtr= src[0]; | |
1508 uint8_t *dstPtr= dst; | |
1509 for(i=0; i<srcSliceH; i++) | |
1510 { | |
1511 memcpy(dstPtr, srcPtr, srcStride[0]); | |
1512 srcPtr+= srcStride[0]; | |
1513 dstPtr+= dstStride[0]; | |
1514 } | |
1515 } | |
1516 dst = dstParam[1] + dstStride[1]*srcSliceY; | |
1517 if(c->srcFormat==IMGFMT_YV12) | |
1518 interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] ); | |
1519 else /* I420 & IYUV */ | |
1520 interleaveBytes( src[2],src[1],dst,c->srcW,srcSliceH,srcStride[2],srcStride[1],dstStride[0] ); | |
1521 } | |
1522 | |
1523 | |
4554 | 1524 /* Warper functions for yuv2bgr */ |
1525 static void planarYuvToBgr(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
4555 | 1526 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1527 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
4554 | 1528 |
1529 if(c->srcFormat==IMGFMT_YV12) | |
4555 | 1530 yuv2rgb( dst,src[0],src[1],src[2],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
4554 | 1531 else /* I420 & IYUV */ |
4555 | 1532 yuv2rgb( dst,src[0],src[2],src[1],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
1533 } | |
1534 | |
6503 | 1535 static void PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
6492 | 1536 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1537 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1538 | |
1539 if(c->srcFormat==IMGFMT_YV12) | |
1540 yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1541 else /* I420 & IYUV */ | |
1542 yv12toyuy2( src[0],src[2],src[1],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1543 } | |
1544 | |
4567 | 1545 static void bgr24to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1546 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1547 | |
1548 if(dstStride[0]*3==srcStride[0]*4) | |
4581 | 1549 rgb24to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 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 { | |
4581 | 1558 rgb24to32(srcPtr, dstPtr, c->srcW*3); |
4567 | 1559 srcPtr+= srcStride[0]; |
1560 dstPtr+= dstStride[0]; | |
1561 } | |
1562 } | |
1563 } | |
1564 | |
6492 | 1565 static void bgr24to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1566 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1567 | |
1568 if(dstStride[0]*3==srcStride[0]*2) | |
1569 rgb24to16(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 rgb24to16(srcPtr, dstPtr, c->srcW*3); | |
1579 srcPtr+= srcStride[0]; | |
1580 dstPtr+= dstStride[0]; | |
1581 } | |
1582 } | |
1583 } | |
1584 | |
1585 static void bgr24to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1586 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1587 | |
1588 if(dstStride[0]*3==srcStride[0]*2) | |
1589 rgb24to15(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 rgb24to15(srcPtr, dstPtr, c->srcW*3); | |
1599 srcPtr+= srcStride[0]; | |
1600 dstPtr+= dstStride[0]; | |
1601 } | |
1602 } | |
1603 } | |
1604 | |
4567 | 1605 static void bgr32to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1606 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1607 | |
1608 if(dstStride[0]*4==srcStride[0]*3) | |
4581 | 1609 rgb32to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 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 { | |
4581 | 1618 rgb32to24(srcPtr, dstPtr, c->srcW<<2); |
1619 srcPtr+= srcStride[0]; | |
1620 dstPtr+= dstStride[0]; | |
1621 } | |
1622 } | |
1623 } | |
1624 | |
6492 | 1625 static void bgr32to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1626 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1627 | |
1628 if(dstStride[0]*4==srcStride[0]*2) | |
1629 rgb32to16(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 rgb32to16(srcPtr, dstPtr, c->srcW<<2); | |
1639 srcPtr+= srcStride[0]; | |
1640 dstPtr+= dstStride[0]; | |
1641 } | |
1642 } | |
1643 } | |
1644 | |
1645 static void bgr32to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1646 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1647 | |
1648 if(dstStride[0]*4==srcStride[0]*2) | |
1649 rgb32to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1650 else | |
1651 { | |
1652 int i; | |
1653 uint8_t *srcPtr= src[0]; | |
1654 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1655 | |
1656 for(i=0; i<srcSliceH; i++) | |
1657 { | |
1658 rgb32to15(srcPtr, dstPtr, c->srcW<<2); | |
1659 srcPtr+= srcStride[0]; | |
1660 dstPtr+= dstStride[0]; | |
1661 } | |
1662 } | |
1663 } | |
1664 | |
4581 | 1665 static void bgr15to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1666 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1667 | |
1668 if(dstStride[0]==srcStride[0]) | |
1669 rgb15to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1670 else | |
1671 { | |
1672 int i; | |
1673 uint8_t *srcPtr= src[0]; | |
1674 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1675 | |
1676 for(i=0; i<srcSliceH; i++) | |
1677 { | |
1678 rgb15to16(srcPtr, dstPtr, c->srcW<<1); | |
4567 | 1679 srcPtr+= srcStride[0]; |
1680 dstPtr+= dstStride[0]; | |
1681 } | |
1682 } | |
1683 } | |
1684 | |
6492 | 1685 static void bgr15to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1686 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1687 | |
1688 if(dstStride[0]*2==srcStride[0]*3) | |
1689 rgb15to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1690 else | |
1691 { | |
1692 int i; | |
1693 uint8_t *srcPtr= src[0]; | |
1694 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1695 | |
1696 for(i=0; i<srcSliceH; i++) | |
1697 { | |
1698 rgb15to24(srcPtr, dstPtr, c->srcW<<1); | |
1699 srcPtr+= srcStride[0]; | |
1700 dstPtr+= dstStride[0]; | |
1701 } | |
1702 } | |
1703 } | |
1704 | |
1705 static void bgr15to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1706 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1707 | |
1708 if(dstStride[0]*2==srcStride[0]*4) | |
1709 rgb15to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1710 else | |
1711 { | |
1712 int i; | |
1713 uint8_t *srcPtr= src[0]; | |
1714 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1715 | |
1716 for(i=0; i<srcSliceH; i++) | |
1717 { | |
1718 rgb15to32(srcPtr, dstPtr, c->srcW<<1); | |
1719 srcPtr+= srcStride[0]; | |
1720 dstPtr+= dstStride[0]; | |
1721 } | |
1722 } | |
1723 } | |
1724 | |
1725 static void bgr16to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1726 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1727 | |
1728 if(dstStride[0]*2==srcStride[0]*3) | |
1729 rgb16to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1730 else | |
1731 { | |
1732 int i; | |
1733 uint8_t *srcPtr= src[0]; | |
1734 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1735 | |
1736 for(i=0; i<srcSliceH; i++) | |
1737 { | |
1738 rgb16to24(srcPtr, dstPtr, c->srcW<<1); | |
1739 srcPtr+= srcStride[0]; | |
1740 dstPtr+= dstStride[0]; | |
1741 } | |
1742 } | |
1743 } | |
1744 | |
1745 static void bgr16to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1746 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1747 | |
1748 if(dstStride[0]*2==srcStride[0]*4) | |
1749 rgb16to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1750 else | |
1751 { | |
1752 int i; | |
1753 uint8_t *srcPtr= src[0]; | |
1754 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1755 | |
1756 for(i=0; i<srcSliceH; i++) | |
1757 { | |
1758 rgb16to32(srcPtr, dstPtr, c->srcW<<1); | |
1759 srcPtr+= srcStride[0]; | |
1760 dstPtr+= dstStride[0]; | |
1761 } | |
1762 } | |
1763 } | |
1764 | |
4623 | 1765 static void bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1766 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1767 | |
1768 rgb24toyv12( | |
1769 src[0], | |
1770 dst[0]+ srcSliceY *dstStride[0], | |
1771 dst[1]+(srcSliceY>>1)*dstStride[1], | |
1772 dst[2]+(srcSliceY>>1)*dstStride[2], | |
1773 c->srcW, srcSliceH, | |
1774 dstStride[0], dstStride[1], srcStride[0]); | |
1775 } | |
1776 | |
6503 | 1777 /** |
1778 * bring pointers in YUV order instead of YVU | |
1779 */ | |
6517 | 1780 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
|
1781 if(format == IMGFMT_YV12 || format == IMGFMT_YVU9){ |
6503 | 1782 sortedP[0]= p[0]; |
1783 sortedP[1]= p[1]; | |
1784 sortedP[2]= p[2]; | |
6517 | 1785 sortedStride[0]= stride[0]; |
1786 sortedStride[1]= stride[1]; | |
1787 sortedStride[2]= stride[2]; | |
6503 | 1788 } |
6517 | 1789 else if(isPacked(format) || isGray(format)) |
6503 | 1790 { |
1791 sortedP[0]= p[0]; | |
1792 sortedP[1]= | |
1793 sortedP[2]= NULL; | |
6517 | 1794 sortedStride[0]= stride[0]; |
6503 | 1795 sortedStride[1]= |
1796 sortedStride[2]= 0; | |
1797 } | |
1798 else /* I420 */ | |
1799 { | |
1800 sortedP[0]= p[0]; | |
1801 sortedP[1]= p[2]; | |
1802 sortedP[2]= p[1]; | |
6517 | 1803 sortedStride[0]= stride[0]; |
1804 sortedStride[1]= stride[2]; | |
1805 sortedStride[2]= stride[1]; | |
6503 | 1806 } |
1807 } | |
4567 | 1808 |
4555 | 1809 /* unscaled copy like stuff (assumes nearly identical formats) */ |
1810 static void simpleCopy(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY, | |
6503 | 1811 int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){ |
4555 | 1812 |
1813 int srcStride[3]; | |
6503 | 1814 int dstStride[3]; |
4555 | 1815 uint8_t *src[3]; |
1816 uint8_t *dst[3]; | |
1817 | |
6517 | 1818 orderYUV(c->srcFormat, src, srcStride, srcParam, srcStrideParam); |
1819 orderYUV(c->dstFormat, dst, dstStride, dstParam, dstStrideParam); | |
4555 | 1820 |
1821 if(isPacked(c->srcFormat)) | |
1822 { | |
1823 if(dstStride[0]==srcStride[0]) | |
1824 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]); | |
1825 else | |
1826 { | |
1827 int i; | |
1828 uint8_t *srcPtr= src[0]; | |
1829 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
4558 | 1830 int length=0; |
1831 | |
1832 /* universal length finder */ | |
4649 | 1833 while(length+c->srcW <= ABS(dstStride[0]) |
1834 && length+c->srcW <= ABS(srcStride[0])) length+= c->srcW; | |
4558 | 1835 ASSERT(length!=0); |
4555 | 1836 |
1837 for(i=0; i<srcSliceH; i++) | |
1838 { | |
1839 memcpy(dstPtr, srcPtr, length); | |
1840 srcPtr+= srcStride[0]; | |
1841 dstPtr+= dstStride[0]; | |
1842 } | |
1843 } | |
1844 } | |
1845 else | |
6503 | 1846 { /* Planar YUV or gray */ |
4555 | 1847 int plane; |
1848 for(plane=0; plane<3; plane++) | |
1849 { | |
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
|
1850 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
|
1851 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
|
1852 int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample); |
6503 | 1853 |
1854 if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0) | |
6492 | 1855 { |
6503 | 1856 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
|
1857 memset(dst[plane], 128, dstStride[plane]*height); |
6492 | 1858 } |
1859 else | |
1860 { | |
6503 | 1861 if(dstStride[plane]==srcStride[plane]) |
1862 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]); | |
1863 else | |
4555 | 1864 { |
6503 | 1865 int i; |
1866 uint8_t *srcPtr= src[plane]; | |
1867 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y; | |
1868 for(i=0; i<height; i++) | |
1869 { | |
1870 memcpy(dstPtr, srcPtr, length); | |
1871 srcPtr+= srcStride[plane]; | |
1872 dstPtr+= dstStride[plane]; | |
1873 } | |
4555 | 1874 } |
1875 } | |
1876 } | |
1877 } | |
4554 | 1878 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1879 |
6503 | 1880 static int remove_dup_fourcc(int fourcc) |
6492 | 1881 { |
1882 switch(fourcc) | |
1883 { | |
1884 case IMGFMT_IYUV: return IMGFMT_I420; | |
1885 case IMGFMT_Y8 : return IMGFMT_Y800; | |
6535
29f94210a67f
IF09 is alias for YVU9 (actually it has extra 4th plane containing MC change
arpi
parents:
6532
diff
changeset
|
1886 case IMGFMT_IF09: return IMGFMT_YVU9; |
6492 | 1887 default: return fourcc; |
1888 } | |
1889 } | |
1890 | |
6517 | 1891 static void getSubSampleFactors(int *h, int *v, int format){ |
1892 switch(format){ | |
1893 case IMGFMT_YUY2: | |
1894 *h=1; | |
1895 *v=0; | |
1896 break; | |
1897 case IMGFMT_YV12: | |
1898 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
|
1899 case IMGFMT_Y800: //FIXME remove after different subsamplings are fully implemented |
6517 | 1900 *h=1; |
1901 *v=1; | |
1902 break; | |
1903 case IMGFMT_YVU9: | |
1904 *h=2; | |
1905 *v=2; | |
1906 break; | |
1907 default: | |
1908 *h=0; | |
1909 *v=0; | |
1910 break; | |
1911 } | |
1912 } | |
1913 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1914 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
|
1915 SwsFilter *srcFilter, SwsFilter *dstFilter){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1916 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1917 SwsContext *c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1918 int i; |
4554 | 1919 int usesFilter; |
6503 | 1920 int unscaled; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1921 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
|
1922 #ifdef ARCH_X86 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1923 if(gCpuCaps.hasMMX) |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1924 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
|
1925 #endif |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1926 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
|
1927 //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
|
1928 //srcFormat= IMGFMT_YVU9; |
4481 | 1929 /* avoid dupplicate Formats, so we dont need to check to much */ |
6492 | 1930 srcFormat = remove_dup_fourcc(srcFormat); |
1931 dstFormat = remove_dup_fourcc(dstFormat); | |
6503 | 1932 |
1933 unscaled = (srcW == dstW && srcH == dstH); | |
1934 | |
1935 if(!isSupportedIn(srcFormat)) | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1936 { |
6503 | 1937 MSG_ERR("swScaler: %s is not supported as input format\n", vo_format_name(srcFormat)); |
1938 return NULL; | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1939 } |
6503 | 1940 if(!isSupportedOut(dstFormat)) |
1941 { | |
1942 MSG_ERR("swScaler: %s is not supported as output format\n", vo_format_name(dstFormat)); | |
1943 return NULL; | |
1944 } | |
1945 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1946 /* sanity check */ |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1947 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
|
1948 { |
6492 | 1949 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
|
1950 srcW, srcH, dstW, dstH); |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1951 return NULL; |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1952 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1953 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1954 if(!dstFilter) dstFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1955 if(!srcFilter) srcFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1956 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1957 c= memalign(64, sizeof(SwsContext)); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1958 memset(c, 0, sizeof(SwsContext)); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1959 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1960 c->srcW= srcW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1961 c->srcH= srcH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1962 c->dstW= dstW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1963 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
|
1964 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
|
1965 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1966 c->flags= flags; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1967 c->dstFormat= dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1968 c->srcFormat= srcFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1969 |
4554 | 1970 usesFilter=0; |
1971 if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesFilter=1; | |
1972 if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesFilter=1; | |
1973 if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesFilter=1; | |
1974 if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesFilter=1; | |
1975 if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesFilter=1; | |
1976 if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesFilter=1; | |
1977 if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesFilter=1; | |
1978 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
|
1979 |
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
|
1980 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
|
1981 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
|
1982 |
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
|
1983 // 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
|
1984 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
|
1985 |
6540 | 1986 // drop some chroma lines if the user wants it |
1987 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT; | |
1988 c->chrSrcVSubSample+= c->vChrDrop; | |
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
|
1989 |
6540 | 1990 // drop every 2. pixel for chroma calculation unless user wants full chroma |
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
|
1991 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
|
1992 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
|
1993 |
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
|
1994 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
|
1995 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
|
1996 |
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
|
1997 // 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
|
1998 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
|
1999 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
|
2000 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
|
2001 c->chrDstH= -((-dstH) >> c->chrDstVSubSample); |
4554 | 2002 |
6578 | 2003 if(isBGR(dstFormat)) |
2004 c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_RGB, c->table_rV, c->table_gU, c->table_gV, c->table_bU); | |
2005 if(isRGB(dstFormat)) | |
2006 c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_BGR, c->table_rV, c->table_gU, c->table_gV, c->table_bU); | |
2007 | |
4567 | 2008 /* unscaled special Cases */ |
6503 | 2009 if(unscaled && !usesFilter) |
4554 | 2010 { |
6492 | 2011 /* yv12_to_nv12 */ |
2012 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_NV12) | |
2013 { | |
2014 c->swScale= PlanarToNV12Wrapper; | |
2015 | |
2016 if(flags&SWS_PRINT_INFO) | |
2017 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2018 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2019 return c; | |
2020 } | |
2021 /* yv12_to_yuy2 */ | |
2022 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_YUY2) | |
2023 { | |
6503 | 2024 c->swScale= PlanarToYuy2Wrapper; |
6492 | 2025 |
2026 if(flags&SWS_PRINT_INFO) | |
2027 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2028 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2029 return c; | |
2030 } | |
4554 | 2031 /* 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
|
2032 if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_I420) && isBGR(dstFormat)) |
4554 | 2033 { |
2034 // FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics | |
6578 | 2035 //FIXME rgb vs. bgr ? |
4791 | 2036 #ifdef WORDS_BIGENDIAN |
5246 | 2037 if(dstFormat==IMGFMT_BGR32) |
2038 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR); | |
2039 else | |
2040 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); | |
4791 | 2041 #else |
4555 | 2042 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); |
4791 | 2043 #endif |
4554 | 2044 c->swScale= planarYuvToBgr; |
4555 | 2045 |
2046 if(flags&SWS_PRINT_INFO) | |
6492 | 2047 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4555 | 2048 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2049 return c; | |
2050 } | |
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
|
2051 #if 1 |
4555 | 2052 /* 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
|
2053 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
|
2054 || (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
|
2055 || (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
|
2056 || (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
|
2057 || (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
|
2058 ) |
4555 | 2059 { |
2060 c->swScale= simpleCopy; | |
2061 | |
4554 | 2062 if(flags&SWS_PRINT_INFO) |
6492 | 2063 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4554 | 2064 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2065 return c; | |
2066 } | |
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
|
2067 #endif |
4567 | 2068 /* bgr32to24 & rgb32to24*/ |
2069 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR24) | |
2070 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB24)) | |
2071 { | |
2072 c->swScale= bgr32to24Wrapper; | |
2073 | |
2074 if(flags&SWS_PRINT_INFO) | |
6492 | 2075 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2076 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2077 return c; | |
2078 } | |
2079 | |
2080 /* bgr32to16 & rgb32to16*/ | |
2081 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR16) | |
2082 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB16)) | |
2083 { | |
2084 c->swScale= bgr32to16Wrapper; | |
2085 | |
2086 if(flags&SWS_PRINT_INFO) | |
2087 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2088 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2089 return c; | |
2090 } | |
2091 | |
2092 /* bgr32to15 & rgb32to15*/ | |
2093 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR15) | |
2094 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB15)) | |
2095 { | |
2096 c->swScale= bgr32to15Wrapper; | |
2097 | |
2098 if(flags&SWS_PRINT_INFO) | |
2099 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 2100 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2101 return c; | |
2102 } | |
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
|
2103 |
4567 | 2104 /* bgr24to32 & rgb24to32*/ |
2105 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR32) | |
2106 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB32)) | |
2107 { | |
2108 c->swScale= bgr24to32Wrapper; | |
2109 | |
2110 if(flags&SWS_PRINT_INFO) | |
6492 | 2111 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2112 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2113 return c; | |
2114 } | |
2115 | |
2116 /* bgr24to16 & rgb24to16*/ | |
2117 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR16) | |
2118 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB16)) | |
2119 { | |
2120 c->swScale= bgr24to16Wrapper; | |
2121 | |
2122 if(flags&SWS_PRINT_INFO) | |
2123 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2124 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2125 return c; | |
2126 } | |
2127 | |
2128 /* bgr24to15 & rgb24to15*/ | |
2129 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR15) | |
2130 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB15)) | |
2131 { | |
2132 c->swScale= bgr24to15Wrapper; | |
2133 | |
2134 if(flags&SWS_PRINT_INFO) | |
2135 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 2136 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2137 return c; | |
2138 } | |
4581 | 2139 |
2140 /* bgr15to16 */ | |
2141 if(srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR16) | |
2142 { | |
2143 c->swScale= bgr15to16Wrapper; | |
2144 | |
2145 if(flags&SWS_PRINT_INFO) | |
6492 | 2146 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2147 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2148 return c; | |
2149 } | |
2150 | |
2151 /* bgr15to24 */ | |
2152 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR24) | |
2153 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB24)) | |
2154 { | |
2155 c->swScale= bgr15to24Wrapper; | |
2156 | |
2157 if(flags&SWS_PRINT_INFO) | |
2158 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2159 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2160 return c; | |
2161 } | |
2162 | |
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
|
2163 #if 0 //segfaults |
6492 | 2164 /* bgr15to32 */ |
2165 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR32) | |
2166 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB32)) | |
2167 { | |
2168 c->swScale= bgr15to32Wrapper; | |
2169 | |
2170 if(flags&SWS_PRINT_INFO) | |
2171 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2172 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2173 return c; | |
2174 } | |
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
|
2175 #endif |
6492 | 2176 /* bgr16to24 */ |
2177 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR24) | |
2178 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB24)) | |
2179 { | |
2180 c->swScale= bgr16to24Wrapper; | |
2181 | |
2182 if(flags&SWS_PRINT_INFO) | |
2183 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2184 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2185 return c; | |
2186 } | |
2187 | |
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
|
2188 #if 0 //segfaults |
6492 | 2189 /* bgr16to32 */ |
2190 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR32) | |
2191 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB32)) | |
2192 { | |
2193 c->swScale= bgr16to32Wrapper; | |
2194 | |
2195 if(flags&SWS_PRINT_INFO) | |
2196 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4581 | 2197 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2198 return c; | |
2199 } | |
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
|
2200 #endif |
4623 | 2201 /* bgr24toYV12 */ |
2202 if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12) | |
2203 { | |
2204 c->swScale= bgr24toyv12Wrapper; | |
2205 | |
2206 if(flags&SWS_PRINT_INFO) | |
6492 | 2207 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4623 | 2208 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2209 return c; | |
2210 } | |
4554 | 2211 } |
2212 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2213 if(cpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2214 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2215 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
|
2216 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
|
2217 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2218 if(flags&SWS_PRINT_INFO) |
6492 | 2219 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
|
2220 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2221 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2222 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2223 c->canMMX2BeUsed=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2224 |
4467 | 2225 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW; |
2226 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH; | |
2227 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2228 // 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
|
2229 // 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
|
2230 // n-2 is the last chrominance sample available |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2231 // 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
|
2232 // 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
|
2233 // first and last pixel |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2234 if(flags&SWS_FAST_BILINEAR) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2235 { |
4467 | 2236 if(c->canMMX2BeUsed) |
2237 { | |
2238 c->lumXInc+= 20; | |
2239 c->chrXInc+= 20; | |
2240 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2241 //we dont use the x86asm scaler if mmx is available |
4467 | 2242 else if(cpuCaps.hasMMX) |
2243 { | |
2244 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20; | |
2245 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20; | |
2246 } | |
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 /* precalculate horizontal scaler filter coefficients */ |
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 const int filterAlign= cpuCaps.hasMMX ? 4 : 1; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2252 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2253 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2254 srcW , dstW, filterAlign, 1<<14, flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2255 srcFilter->lumH, dstFilter->lumH); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2256 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
|
2257 c->chrSrcW, c->chrDstW, filterAlign, 1<<14, flags, |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2258 srcFilter->chrH, dstFilter->chrH); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2259 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2260 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2261 // cant downscale !!! |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2262 if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2263 { |
5452 | 2264 c->lumMmx2Filter = (int16_t*)memalign(8, (dstW /8+8)*sizeof(int16_t)); |
2265 c->chrMmx2Filter = (int16_t*)memalign(8, (c->chrDstW /4+8)*sizeof(int16_t)); | |
2266 c->lumMmx2FilterPos= (int32_t*)memalign(8, (dstW /2/8+8)*sizeof(int32_t)); | |
2267 c->chrMmx2FilterPos= (int32_t*)memalign(8, (c->chrDstW/2/4+8)*sizeof(int32_t)); | |
2268 | |
2269 initMMX2HScaler( dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8); | |
2270 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
|
2271 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2272 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2273 } // Init Horizontal stuff |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2274 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2275 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2276 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2277 /* precalculate vertical scaler filter coefficients */ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2278 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2279 srcH , dstH, 1, (1<<12)-4, flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2280 srcFilter->lumV, dstFilter->lumV); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2281 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
|
2282 c->chrSrcH, c->chrDstH, 1, (1<<12)-4, flags, |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2283 srcFilter->chrV, dstFilter->chrV); |
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 // 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
|
2286 c->vLumBufSize= c->vLumFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2287 c->vChrBufSize= c->vChrFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2288 for(i=0; i<dstH; i++) |
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 int chrI= i*c->chrDstH / dstH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2291 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
|
2292 ((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
|
2293 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
|
2294 if(c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2295 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
|
2296 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
|
2297 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
|
2298 } |
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 // 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
|
2301 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
|
2302 c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*)); |
4419 | 2303 //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
|
2304 for(i=0; i<c->vLumBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2305 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
|
2306 for(i=0; i<c->vChrBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2307 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
|
2308 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2309 //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
|
2310 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
|
2311 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
|
2312 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2313 ASSERT(c->chrDstH <= dstH) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2314 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2315 // pack filter data for mmx code |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2316 if(cpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2317 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2318 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
|
2319 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
|
2320 for(i=0; i<c->vLumFilterSize*dstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2321 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
|
2322 c->vLumFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2323 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
|
2324 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
|
2325 c->vChrFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2326 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2327 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2328 if(flags&SWS_PRINT_INFO) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2329 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2330 #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
|
2331 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
|
2332 #else |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2333 char *dither= ""; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2334 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2335 if(flags&SWS_FAST_BILINEAR) |
6492 | 2336 MSG_INFO("\nSwScaler: FAST_BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2337 else if(flags&SWS_BILINEAR) |
6492 | 2338 MSG_INFO("\nSwScaler: BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2339 else if(flags&SWS_BICUBIC) |
6492 | 2340 MSG_INFO("\nSwScaler: BICUBIC scaler, "); |
4467 | 2341 else if(flags&SWS_X) |
6492 | 2342 MSG_INFO("\nSwScaler: Experimental scaler, "); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2343 else if(flags&SWS_POINT) |
6492 | 2344 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
|
2345 else if(flags&SWS_AREA) |
6492 | 2346 MSG_INFO("\nSwScaler: Area Averageing scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2347 else |
6492 | 2348 MSG_INFO("\nSwScaler: ehh flags invalid?! "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2349 |
6492 | 2350 if(dstFormat==IMGFMT_BGR15 || dstFormat==IMGFMT_BGR16) |
2351 MSG_INFO("from %s to%s %s ", | |
2352 vo_format_name(srcFormat), dither, vo_format_name(dstFormat)); | |
2353 else | |
2354 MSG_INFO("from %s to %s ", | |
2355 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
|
2356 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2357 if(cpuCaps.hasMMX2) |
6492 | 2358 MSG_INFO("using MMX2\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2359 else if(cpuCaps.has3DNow) |
6492 | 2360 MSG_INFO("using 3DNOW\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2361 else if(cpuCaps.hasMMX) |
6492 | 2362 MSG_INFO("using MMX\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2363 else |
6492 | 2364 MSG_INFO("using C\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2365 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2366 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2367 if((flags & SWS_PRINT_INFO) && verbose) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2368 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2369 if(cpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2370 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2371 if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR)) |
6492 | 2372 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
|
2373 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2374 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2375 if(c->hLumFilterSize==4) |
6492 | 2376 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
|
2377 else if(c->hLumFilterSize==8) |
6492 | 2378 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
|
2379 else |
6492 | 2380 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
|
2381 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2382 if(c->hChrFilterSize==4) |
6492 | 2383 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
|
2384 else if(c->hChrFilterSize==8) |
6492 | 2385 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
|
2386 else |
6492 | 2387 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
|
2388 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2389 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2390 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2391 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2392 #ifdef ARCH_X86 |
6492 | 2393 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
|
2394 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2395 if(flags & SWS_FAST_BILINEAR) |
6492 | 2396 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
|
2397 else |
6492 | 2398 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
|
2399 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2400 } |
4419 | 2401 if(isPlanarYUV(dstFormat)) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2402 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2403 if(c->vLumFilterSize==1) |
6492 | 2404 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
|
2405 else |
6492 | 2406 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
|
2407 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2408 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2409 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2410 if(c->vLumFilterSize==1 && c->vChrFilterSize==2) |
6492 | 2411 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
|
2412 "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
|
2413 else if(c->vLumFilterSize==2 && c->vChrFilterSize==2) |
6492 | 2414 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
|
2415 else |
6492 | 2416 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
|
2417 } |
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 if(dstFormat==IMGFMT_BGR24) |
6492 | 2420 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
|
2421 cpuCaps.hasMMX2 ? "MMX2" : (cpuCaps.hasMMX ? "MMX" : "C")); |
4304 | 2422 else if(dstFormat==IMGFMT_BGR32) |
6492 | 2423 MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2424 else if(dstFormat==IMGFMT_BGR16) |
6492 | 2425 MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2426 else if(dstFormat==IMGFMT_BGR15) |
6492 | 2427 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
|
2428 |
6492 | 2429 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
|
2430 } |
4467 | 2431 if((flags & SWS_PRINT_INFO) && verbose>1) |
2432 { | |
6492 | 2433 MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2434 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); |
6492 | 2435 MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2436 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc); |
2437 } | |
4554 | 2438 |
2439 c->swScale= swScale; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2440 return c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2441 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2442 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2443 /** |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2444 * 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
|
2445 * 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
|
2446 */ |
4290
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 SwsVector *getGaussianVec(double variance, double quality){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2449 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
|
2450 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2451 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2452 double middle= (length-1)*0.5; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2453 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2454 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2455 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2456 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2457 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2458 for(i=0; i<length; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2459 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2460 double dist= i-middle; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2461 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
|
2462 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2463 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2464 normalizeVec(vec, 1.0); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2465 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2466 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2467 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2468 |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2469 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
|
2470 int i; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2471 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
|
2472 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
|
2473 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2474 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
|
2475 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
|
2476 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2477 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
|
2478 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
|
2479 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2480 return vec; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2481 } |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2482 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2483 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2484 SwsVector *getIdentityVec(void){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2485 double *coeff= memalign(sizeof(double), sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2486 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2487 coeff[0]= 1.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2488 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2489 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2490 vec->length= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2491 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2492 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2493 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2494 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2495 void normalizeVec(SwsVector *a, double height){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2496 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2497 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2498 double inv; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2499 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2500 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2501 sum+= a->coeff[i]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2502 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2503 inv= height/sum; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2504 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2505 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2506 a->coeff[i]*= height; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2507 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2508 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2509 void scaleVec(SwsVector *a, double scalar){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2510 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2511 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2512 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2513 a->coeff[i]*= scalar; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2514 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2515 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2516 static SwsVector *getConvVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2517 int length= a->length + b->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2518 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2519 int i, j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2520 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2521 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2522 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2523 vec->length= length; |
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<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2526 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2527 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2528 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2529 for(j=0; j<b->length; j++) |
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 coeff[i+j]+= a->coeff[i]*b->coeff[j]; |
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 } |
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 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2536 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2537 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2538 static SwsVector *sumVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2539 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2540 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2541 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2542 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2543 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2544 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2545 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2546 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2547 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2548 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2549 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
|
2550 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
|
2551 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2552 return vec; |
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 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2555 static SwsVector *diffVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2556 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2557 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2558 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2559 SwsVector *vec= malloc(sizeof(SwsVector)); |
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 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2562 vec->length= length; |
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 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2565 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2566 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
|
2567 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
|
2568 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2569 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2570 } |
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 /* 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
|
2573 static SwsVector *getShiftedVec(SwsVector *a, int shift){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2574 int length= a->length + ABS(shift)*2; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2575 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2576 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2577 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2578 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2579 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2580 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2581 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2582 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
|
2583 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2584 for(i=0; i<a->length; i++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2585 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2586 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
|
2587 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2588 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2589 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2590 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2591 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2592 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
|
2593 SwsVector *shifted= getShiftedVec(a, shift); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2594 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2595 a->coeff= shifted->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2596 a->length= shifted->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2597 free(shifted); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2598 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2599 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2600 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
|
2601 SwsVector *sum= sumVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2602 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2603 a->coeff= sum->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2604 a->length= sum->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2605 free(sum); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2606 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2607 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2608 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
|
2609 SwsVector *diff= diffVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2610 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2611 a->coeff= diff->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2612 a->length= diff->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2613 free(diff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2614 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2615 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2616 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
|
2617 SwsVector *conv= getConvVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2618 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2619 a->coeff= conv->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2620 a->length= conv->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2621 free(conv); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2622 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2623 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2624 SwsVector *cloneVec(SwsVector *a){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2625 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
|
2626 int i; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2627 SwsVector *vec= malloc(sizeof(SwsVector)); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2628 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2629 vec->coeff= coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2630 vec->length= a->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2631 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2632 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
|
2633 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2634 return vec; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2635 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2636 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2637 void printVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2638 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2639 double max=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2640 double min=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2641 double range; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2642 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2643 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2644 if(a->coeff[i]>max) max= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2645 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2646 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2647 if(a->coeff[i]<min) min= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2648 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2649 range= max - min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2650 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2651 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2652 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2653 int x= (int)((a->coeff[i]-min)*60.0/range +0.5); |
6492 | 2654 MSG_DBG2("%1.3f ", a->coeff[i]); |
2655 for(;x>0; x--) MSG_DBG2(" "); | |
2656 MSG_DBG2("|\n"); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2657 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2658 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2659 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2660 void freeVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2661 if(!a) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2662 if(a->coeff) free(a->coeff); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2663 a->coeff=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2664 a->length=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2665 free(a); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2666 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2667 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2668 void freeSwsContext(SwsContext *c){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2669 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2670 if(!c) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2671 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2672 if(c->lumPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2673 { |
4419 | 2674 for(i=0; i<c->vLumBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2675 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2676 if(c->lumPixBuf[i]) free(c->lumPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2677 c->lumPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2678 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2679 free(c->lumPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2680 c->lumPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2681 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2682 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2683 if(c->chrPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2684 { |
4419 | 2685 for(i=0; i<c->vChrBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2686 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2687 if(c->chrPixBuf[i]) free(c->chrPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2688 c->chrPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2689 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2690 free(c->chrPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2691 c->chrPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2692 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2693 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2694 if(c->vLumFilter) free(c->vLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2695 c->vLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2696 if(c->vChrFilter) free(c->vChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2697 c->vChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2698 if(c->hLumFilter) free(c->hLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2699 c->hLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2700 if(c->hChrFilter) free(c->hChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2701 c->hChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2702 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2703 if(c->vLumFilterPos) free(c->vLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2704 c->vLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2705 if(c->vChrFilterPos) free(c->vChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2706 c->vChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2707 if(c->hLumFilterPos) free(c->hLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2708 c->hLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2709 if(c->hChrFilterPos) free(c->hChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2710 c->hChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2711 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2712 if(c->lumMmxFilter) free(c->lumMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2713 c->lumMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2714 if(c->chrMmxFilter) free(c->chrMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2715 c->chrMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2716 |
5452 | 2717 if(c->lumMmx2Filter) free(c->lumMmx2Filter); |
2718 c->lumMmx2Filter=NULL; | |
2719 if(c->chrMmx2Filter) free(c->chrMmx2Filter); | |
2720 c->chrMmx2Filter=NULL; | |
2721 if(c->lumMmx2FilterPos) free(c->lumMmx2FilterPos); | |
2722 c->lumMmx2FilterPos=NULL; | |
2723 if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos); | |
2724 c->chrMmx2FilterPos=NULL; | |
6578 | 2725 if(c->yuvTable) free(c->yuvTable); |
2726 c->yuvTable=NULL; | |
5452 | 2727 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2728 free(c); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2729 } |
4281 | 2730 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2731 |