Mercurial > mplayer.hg
annotate postproc/swscale.c @ 6582:f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
author | michael |
---|---|
date | Thu, 27 Jun 2002 23:48:53 +0000 |
parents | 5059143aa020 |
children | 0b5a789d7fab |
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, | |
6580 | 260 IMGFMT_RGB24, |
261 IMGFMT_RGB32, | |
6517 | 262 //IMGFMT_Y8, |
263 IMGFMT_Y800, | |
264 //IMGFMT_YUY2, | |
265 0 | |
266 }; | |
267 | |
268 static uint64_t getSSD(uint8_t *src1, uint8_t *src2, int stride1, int stride2, int w, int h){ | |
269 int x,y; | |
270 uint64_t ssd=0; | |
271 | |
272 for(y=0; y<h; y++){ | |
273 for(x=0; x<w; x++){ | |
274 int d= src1[x + y*stride1] - src2[x + y*stride2]; | |
275 ssd+= d*d; | |
276 } | |
277 } | |
278 return ssd; | |
279 } | |
280 | |
281 // test by ref -> src -> dst -> out & compare out against ref | |
282 // ref & out are YV12 | |
283 static void doTest(uint8_t *ref[3], int refStride[3], int w, int h, int srcFormat, int dstFormat, | |
284 int srcW, int srcH, int dstW, int dstH, int flags){ | |
285 uint8_t *src[3]; | |
286 uint8_t *dst[3]; | |
287 uint8_t *out[3]; | |
288 int srcStride[3], dstStride[3]; | |
289 int i; | |
290 uint64_t ssdY, ssdU, ssdV; | |
291 SwsContext *srcContext, *dstContext, *outContext; | |
292 | |
293 for(i=0; i<3; i++){ | |
294 srcStride[i]= srcW*4; | |
295 dstStride[i]= dstW*4; | |
296 src[i]= malloc(srcStride[i]*srcH); | |
297 dst[i]= malloc(dstStride[i]*dstH); | |
298 out[i]= malloc(refStride[i]*h); | |
299 } | |
300 | |
301 srcContext= getSwsContext(w, h, IMGFMT_YV12, srcW, srcH, srcFormat, flags, NULL, NULL); | |
302 dstContext= getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, NULL, NULL); | |
303 outContext= getSwsContext(dstW, dstH, dstFormat, w, h, IMGFMT_YV12, flags, NULL, NULL); | |
304 if(srcContext==NULL ||dstContext==NULL ||outContext==NULL){ | |
305 printf("Failed allocating swsContext\n"); | |
306 goto end; | |
307 } | |
308 // printf("test %X %X %X -> %X %X %X\n", (int)ref[0], (int)ref[1], (int)ref[2], | |
309 // (int)src[0], (int)src[1], (int)src[2]); | |
310 | |
311 srcContext->swScale(srcContext, ref, refStride, 0, h , src, srcStride); | |
312 dstContext->swScale(dstContext, src, srcStride, 0, srcH, dst, dstStride); | |
313 outContext->swScale(outContext, dst, dstStride, 0, dstH, out, refStride); | |
314 | |
315 ssdY= getSSD(ref[0], out[0], refStride[0], refStride[0], w, h); | |
316 ssdU= getSSD(ref[1], out[1], refStride[1], refStride[1], (w+1)>>1, (h+1)>>1); | |
317 ssdV= getSSD(ref[2], out[2], refStride[2], refStride[2], (w+1)>>1, (h+1)>>1); | |
318 | |
319 if(isGray(srcFormat) || isGray(dstFormat)) ssdU=ssdV=0; //FIXME check that output is really gray | |
320 | |
321 ssdY/= w*h; | |
322 ssdU/= w*h/4; | |
323 ssdV/= w*h/4; | |
324 | |
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
|
325 if(ssdY>100 || ssdU>50 || ssdV>50){ |
6517 | 326 printf(" %s %dx%d -> %s %4dx%4d flags=%2d SSD=%5lld,%5lld,%5lld\n", |
327 vo_format_name(srcFormat), srcW, srcH, | |
328 vo_format_name(dstFormat), dstW, dstH, | |
329 flags, | |
330 ssdY, ssdU, ssdV); | |
331 } | |
332 | |
333 end: | |
334 | |
335 freeSwsContext(srcContext); | |
336 freeSwsContext(dstContext); | |
337 freeSwsContext(outContext); | |
338 | |
339 for(i=0; i<3; i++){ | |
340 free(src[i]); | |
341 free(dst[i]); | |
342 free(out[i]); | |
343 } | |
344 } | |
345 | |
346 static void selfTest(uint8_t *src[3], int stride[3], int w, int h){ | |
347 int srcFormat, dstFormat, srcFormatIndex, dstFormatIndex; | |
348 int srcW, srcH, dstW, dstH; | |
349 int flags; | |
350 | |
351 for(srcFormatIndex=0; ;srcFormatIndex++){ | |
352 srcFormat= testFormat[srcFormatIndex]; | |
353 if(!srcFormat) break; | |
354 for(dstFormatIndex=0; ;dstFormatIndex++){ | |
355 dstFormat= testFormat[dstFormatIndex]; | |
356 if(!dstFormat) break; | |
357 if(!isSupportedOut(dstFormat)) continue; | |
358 | |
359 srcW= w+w/3; | |
360 srcH= h+h/3; | |
361 for(dstW=w; dstW<w*2; dstW+= dstW/3){ | |
362 for(dstH=h; dstH<h*2; dstH+= dstH/3){ | |
363 for(flags=1; flags<33; flags*=2) | |
364 doTest(src, stride, w, h, srcFormat, dstFormat, | |
365 srcW, srcH, dstW, dstH, flags); | |
366 } | |
367 } | |
368 } | |
369 } | |
370 } | |
371 | |
6540 | 372 static inline void yuv2yuvXinC(int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, |
3352 | 373 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, |
6540 | 374 uint8_t *dest, uint8_t *uDest, uint8_t *vDest, int dstW, int chrDstW) |
3352 | 375 { |
376 //FIXME Optimize (just quickly writen not opti..) | |
377 int i; | |
6540 | 378 for(i=0; i<dstW; i++) |
3352 | 379 { |
380 int val=0; | |
381 int j; | |
382 for(j=0; j<lumFilterSize; j++) | |
383 val += lumSrc[j][i] * lumFilter[j]; | |
384 | |
385 dest[i]= MIN(MAX(val>>19, 0), 255); | |
386 } | |
387 | |
388 if(uDest != NULL) | |
6540 | 389 for(i=0; i<chrDstW; i++) |
3352 | 390 { |
391 int u=0; | |
392 int v=0; | |
393 int j; | |
3641 | 394 for(j=0; j<chrFilterSize; j++) |
3352 | 395 { |
396 u += chrSrc[j][i] * chrFilter[j]; | |
397 v += chrSrc[j][i + 2048] * chrFilter[j]; | |
398 } | |
399 | |
400 uDest[i]= MIN(MAX(u>>19, 0), 255); | |
401 vDest[i]= MIN(MAX(v>>19, 0), 255); | |
402 } | |
403 } | |
404 | |
6578 | 405 #define YSCALE_YUV_2_RGBX_C(type) \ |
406 for(i=0; i<(dstW>>1); i++){\ | |
407 int j;\ | |
408 int Y1=0;\ | |
409 int Y2=0;\ | |
410 int U=0;\ | |
411 int V=0;\ | |
412 type *r, *b, *g;\ | |
413 const int i2= 2*i;\ | |
414 \ | |
415 for(j=0; j<lumFilterSize; j++)\ | |
416 {\ | |
417 Y1 += lumSrc[j][i2] * lumFilter[j];\ | |
418 Y2 += lumSrc[j][i2+1] * lumFilter[j];\ | |
419 }\ | |
420 for(j=0; j<chrFilterSize; j++)\ | |
421 {\ | |
422 U += chrSrc[j][i] * chrFilter[j];\ | |
423 V += chrSrc[j][i+2048] * chrFilter[j];\ | |
424 }\ | |
425 Y1>>=19;\ | |
426 Y2>>=19;\ | |
427 U >>=19;\ | |
428 V >>=19;\ | |
429 if((Y1|Y2|U|V)&256)\ | |
430 {\ | |
431 if(Y1>255) Y1=255;\ | |
432 else if(Y1<0)Y1=0;\ | |
433 if(Y2>255) Y2=255;\ | |
434 else if(Y2<0)Y2=0;\ | |
435 if(U>255) U=255;\ | |
436 else if(U<0) U=0;\ | |
437 if(V>255) V=255;\ | |
438 else if(V<0) V=0;\ | |
439 }\ | |
440 r = c->table_rV[V];\ | |
441 g = c->table_gU[U] + c->table_gV[V];\ | |
442 b = c->table_bU[U];\ | |
3352 | 443 |
6578 | 444 #define YSCALE_YUV_2_RGB2_C(type) \ |
445 for(i=0; i<(dstW>>1); i++){\ | |
446 const int i2= 2*i;\ | |
447 int Y1= (buf0[i2 ]*yalpha1+buf1[i2 ]*yalpha)>>19;\ | |
448 int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;\ | |
449 int U= (uvbuf0[i ]*uvalpha1+uvbuf1[i ]*uvalpha)>>19;\ | |
450 int V= (uvbuf0[i+2048]*uvalpha1+uvbuf1[i+2048]*uvalpha)>>19;\ | |
451 type *r, *b, *g;\ | |
452 r = c->table_rV[V];\ | |
453 g = c->table_gU[U] + c->table_gV[V];\ | |
454 b = c->table_bU[U];\ | |
3352 | 455 |
6578 | 456 #define YSCALE_YUV_2_RGB1_C(type) \ |
457 for(i=0; i<(dstW>>1); i++){\ | |
458 const int i2= 2*i;\ | |
459 int Y1= buf0[i2 ]>>7;\ | |
460 int Y2= buf0[i2+1]>>7;\ | |
461 int U= (uvbuf1[i ])>>7;\ | |
462 int V= (uvbuf1[i+2048])>>7;\ | |
463 type *r, *b, *g;\ | |
464 r = c->table_rV[V];\ | |
465 g = c->table_gU[U] + c->table_gV[V];\ | |
466 b = c->table_bU[U];\ | |
3352 | 467 |
6578 | 468 #define YSCALE_YUV_2_RGB1B_C(type) \ |
469 for(i=0; i<(dstW>>1); i++){\ | |
470 const int i2= 2*i;\ | |
471 int Y1= buf0[i2 ]>>7;\ | |
472 int Y2= buf0[i2+1]>>7;\ | |
473 int U= (uvbuf0[i ] + uvbuf1[i ])>>8;\ | |
474 int V= (uvbuf0[i+2048] + uvbuf1[i+2048])>>8;\ | |
475 type *r, *b, *g;\ | |
476 r = c->table_rV[V];\ | |
477 g = c->table_gU[U] + c->table_gV[V];\ | |
478 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
|
479 |
6578 | 480 #define YSCALE_YUV_2_ANYRGB_C(func)\ |
481 switch(c->dstFormat)\ | |
482 {\ | |
483 case IMGFMT_BGR32:\ | |
484 case IMGFMT_RGB32:\ | |
485 func(uint32_t)\ | |
486 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\ | |
487 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\ | |
488 } \ | |
489 break;\ | |
490 case IMGFMT_RGB24:\ | |
491 func(uint8_t)\ | |
492 ((uint8_t*)dest)[0]= r[Y1];\ | |
493 ((uint8_t*)dest)[1]= g[Y1];\ | |
494 ((uint8_t*)dest)[2]= b[Y1];\ | |
495 ((uint8_t*)dest)[3]= r[Y2];\ | |
496 ((uint8_t*)dest)[4]= g[Y2];\ | |
497 ((uint8_t*)dest)[5]= b[Y2];\ | |
498 ((uint8_t*)dest)+=6;\ | |
499 }\ | |
500 break;\ | |
501 case IMGFMT_BGR24:\ | |
502 func(uint8_t)\ | |
503 ((uint8_t*)dest)[0]= b[Y1];\ | |
504 ((uint8_t*)dest)[1]= g[Y1];\ | |
505 ((uint8_t*)dest)[2]= r[Y1];\ | |
506 ((uint8_t*)dest)[3]= b[Y2];\ | |
507 ((uint8_t*)dest)[4]= g[Y2];\ | |
508 ((uint8_t*)dest)[5]= r[Y2];\ | |
509 ((uint8_t*)dest)+=6;\ | |
510 }\ | |
511 break;\ | |
512 case IMGFMT_RGB16:\ | |
513 case IMGFMT_BGR16:\ | |
514 {\ | |
515 const int dr1= dither_2x2_8[y&1 ][0];\ | |
516 const int dg1= dither_2x2_4[y&1 ][0];\ | |
517 const int db1= dither_2x2_8[(y&1)^1][0];\ | |
518 const int dr2= dither_2x2_8[y&1 ][1];\ | |
519 const int dg2= dither_2x2_4[y&1 ][1];\ | |
520 const int db2= dither_2x2_8[(y&1)^1][1];\ | |
521 func(uint16_t)\ | |
522 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\ | |
523 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\ | |
524 }\ | |
525 }\ | |
526 break;\ | |
527 case IMGFMT_RGB15:\ | |
528 case IMGFMT_BGR15:\ | |
529 {\ | |
530 const int dr1= dither_2x2_8[y&1 ][0];\ | |
531 const int dg1= dither_2x2_8[y&1 ][1];\ | |
532 const int db1= dither_2x2_8[(y&1)^1][0];\ | |
533 const int dr2= dither_2x2_8[y&1 ][1];\ | |
534 const int dg2= dither_2x2_8[y&1 ][0];\ | |
535 const int db2= dither_2x2_8[(y&1)^1][1];\ | |
536 func(uint16_t)\ | |
537 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\ | |
538 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\ | |
539 }\ | |
540 }\ | |
541 break;\ | |
542 case IMGFMT_RGB8:\ | |
543 case IMGFMT_BGR8:\ | |
544 {\ | |
545 const uint8_t * const d64= dither_8x8_73[y&7];\ | |
546 const uint8_t * const d32= dither_8x8_32[y&7];\ | |
547 func(uint8_t)\ | |
548 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\ | |
549 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\ | |
550 }\ | |
551 }\ | |
552 break;\ | |
553 case IMGFMT_RGB4:\ | |
554 case IMGFMT_BGR4:\ | |
555 {\ | |
556 const uint8_t * const d64= dither_8x8_73 [y&7];\ | |
557 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
558 func(uint8_t)\ | |
559 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\ | |
560 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\ | |
561 }\ | |
562 }\ | |
563 break;\ | |
564 case IMGFMT_RGB1:\ | |
565 case IMGFMT_BGR1:\ | |
566 {\ | |
567 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
568 uint8_t *g= c->table_gU[128] + c->table_gV[128];\ | |
569 for(i=0; i<dstW-7; i+=8){\ | |
570 int acc;\ | |
571 acc = g[((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19) + d128[0]];\ | |
572 acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\ | |
573 acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\ | |
574 acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\ | |
575 acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\ | |
576 acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\ | |
577 acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\ | |
578 acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\ | |
579 ((uint8_t*)dest)[0]= acc;\ | |
580 ((uint8_t*)dest)++;\ | |
581 }\ | |
582 \ | |
583 /*\ | |
584 ((uint8_t*)dest)-= dstW>>4;\ | |
585 {\ | |
586 int acc=0;\ | |
587 int left=0;\ | |
588 static int top[1024];\ | |
589 static int last_new[1024][1024];\ | |
590 static int last_in3[1024][1024];\ | |
591 static int drift[1024][1024];\ | |
592 int topLeft=0;\ | |
593 int shift=0;\ | |
594 int count=0;\ | |
595 const uint8_t * const d128=dither_8x8_220[y&7];\ | |
596 int error_new=0;\ | |
597 int error_in3=0;\ | |
598 int f=0;\ | |
599 \ | |
600 for(i=dstW>>1; i<dstW; i++){\ | |
601 int in= ((buf0[i ]*yalpha1+buf1[i ]*yalpha)>>19);\ | |
602 int in2 = (76309 * (in - 16) + 32768) >> 16;\ | |
603 int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\ | |
604 int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\ | |
605 + (last_new[y][i] - in3)*f/256;\ | |
606 int new= old> 128 ? 255 : 0;\ | |
607 \ | |
608 error_new+= ABS(last_new[y][i] - new);\ | |
609 error_in3+= ABS(last_in3[y][i] - in3);\ | |
610 f= error_new - error_in3*4;\ | |
611 if(f<0) f=0;\ | |
612 if(f>256) f=256;\ | |
613 \ | |
614 topLeft= top[i];\ | |
615 left= top[i]= old - new;\ | |
616 last_new[y][i]= new;\ | |
617 last_in3[y][i]= in3;\ | |
618 \ | |
619 acc+= acc + (new&1);\ | |
620 if((i&7)==6){\ | |
621 ((uint8_t*)dest)[0]= acc;\ | |
622 ((uint8_t*)dest)++;\ | |
623 }\ | |
624 }\ | |
625 }\ | |
626 */\ | |
627 }\ | |
628 break;\ | |
629 }\ | |
3352 | 630 |
631 | |
6578 | 632 static inline void yuv2rgbXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize, |
633 int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize, | |
634 uint8_t *dest, int dstW, int y) | |
635 { | |
636 int i; | |
637 switch(c->dstFormat) | |
3352 | 638 { |
6578 | 639 case IMGFMT_RGB32: |
640 case IMGFMT_BGR32: | |
641 YSCALE_YUV_2_RGBX_C(uint32_t) | |
642 ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1]; | |
643 ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2]; | |
644 } | |
645 break; | |
646 case IMGFMT_RGB24: | |
647 YSCALE_YUV_2_RGBX_C(uint8_t) | |
648 ((uint8_t*)dest)[0]= r[Y1]; | |
649 ((uint8_t*)dest)[1]= g[Y1]; | |
650 ((uint8_t*)dest)[2]= b[Y1]; | |
651 ((uint8_t*)dest)[3]= r[Y2]; | |
652 ((uint8_t*)dest)[4]= g[Y2]; | |
653 ((uint8_t*)dest)[5]= b[Y2]; | |
654 ((uint8_t*)dest)+=6; | |
655 } | |
656 break; | |
657 case IMGFMT_BGR24: | |
658 YSCALE_YUV_2_RGBX_C(uint8_t) | |
659 ((uint8_t*)dest)[0]= b[Y1]; | |
660 ((uint8_t*)dest)[1]= g[Y1]; | |
661 ((uint8_t*)dest)[2]= r[Y1]; | |
662 ((uint8_t*)dest)[3]= b[Y2]; | |
663 ((uint8_t*)dest)[4]= g[Y2]; | |
664 ((uint8_t*)dest)[5]= r[Y2]; | |
665 ((uint8_t*)dest)+=6; | |
666 } | |
667 break; | |
668 case IMGFMT_RGB16: | |
669 case IMGFMT_BGR16: | |
670 { | |
671 const int dr1= dither_2x2_8[y&1 ][0]; | |
672 const int dg1= dither_2x2_4[y&1 ][0]; | |
673 const int db1= dither_2x2_8[(y&1)^1][0]; | |
674 const int dr2= dither_2x2_8[y&1 ][1]; | |
675 const int dg2= dither_2x2_4[y&1 ][1]; | |
676 const int db2= dither_2x2_8[(y&1)^1][1]; | |
677 YSCALE_YUV_2_RGBX_C(uint16_t) | |
678 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1]; | |
679 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2]; | |
680 } | |
681 } | |
682 break; | |
683 case IMGFMT_RGB15: | |
684 case IMGFMT_BGR15: | |
685 { | |
686 const int dr1= dither_2x2_8[y&1 ][0]; | |
687 const int dg1= dither_2x2_8[y&1 ][1]; | |
688 const int db1= dither_2x2_8[(y&1)^1][0]; | |
689 const int dr2= dither_2x2_8[y&1 ][1]; | |
690 const int dg2= dither_2x2_8[y&1 ][0]; | |
691 const int db2= dither_2x2_8[(y&1)^1][1]; | |
692 YSCALE_YUV_2_RGBX_C(uint16_t) | |
693 ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1]; | |
694 ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2]; | |
695 } | |
696 } | |
697 break; | |
698 case IMGFMT_RGB8: | |
699 case IMGFMT_BGR8: | |
700 { | |
701 const uint8_t * const d64= dither_8x8_73[y&7]; | |
702 const uint8_t * const d32= dither_8x8_32[y&7]; | |
703 YSCALE_YUV_2_RGBX_C(uint8_t) | |
704 ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]]; | |
705 ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]]; | |
706 } | |
707 } | |
708 break; | |
709 case IMGFMT_RGB4: | |
710 case IMGFMT_BGR4: | |
711 { | |
712 const uint8_t * const d64= dither_8x8_73 [y&7]; | |
713 const uint8_t * const d128=dither_8x8_220[y&7]; | |
714 YSCALE_YUV_2_RGBX_C(uint8_t) | |
715 ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]; | |
716 ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]]; | |
717 } | |
718 } | |
719 break; | |
720 case IMGFMT_RGB1: | |
721 case IMGFMT_BGR1: | |
722 { | |
723 const uint8_t * const d128=dither_8x8_220[y&7]; | |
724 uint8_t *g= c->table_gU[128] + c->table_gV[128]; | |
725 int acc=0; | |
726 for(i=0; i<dstW-1; i+=2){ | |
727 int j; | |
728 int Y1=0; | |
729 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
|
730 |
6578 | 731 for(j=0; j<lumFilterSize; j++) |
732 { | |
733 Y1 += lumSrc[j][i] * lumFilter[j]; | |
734 Y2 += lumSrc[j][i+1] * lumFilter[j]; | |
735 } | |
736 Y1>>=19; | |
737 Y2>>=19; | |
738 if((Y1|Y2)&256) | |
739 { | |
740 if(Y1>255) Y1=255; | |
741 else if(Y1<0)Y1=0; | |
742 if(Y2>255) Y2=255; | |
743 else if(Y2<0)Y2=0; | |
744 } | |
745 acc+= acc + g[Y1+d128[(i+0)&7]]; | |
746 acc+= acc + g[Y2+d128[(i+1)&7]]; | |
747 if((i&7)==6){ | |
748 ((uint8_t*)dest)[0]= acc; | |
749 ((uint8_t*)dest)++; | |
750 } | |
3352 | 751 } |
752 } | |
6578 | 753 break; |
3352 | 754 } |
755 } | |
756 | |
757 | |
3126 | 758 //Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one |
759 //Plain C versions | |
3152 | 760 #if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT) |
761 #define COMPILE_C | |
762 #endif | |
763 | |
764 #ifdef CAN_COMPILE_X86_ASM | |
765 | |
766 #if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
767 #define COMPILE_MMX | |
768 #endif | |
769 | |
770 #if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT) | |
771 #define COMPILE_MMX2 | |
772 #endif | |
773 | |
774 #if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT) | |
775 #define COMPILE_3DNOW | |
776 #endif | |
777 #endif //CAN_COMPILE_X86_ASM | |
778 | |
779 #undef HAVE_MMX | |
780 #undef HAVE_MMX2 | |
781 #undef HAVE_3DNOW | |
782 | |
783 #ifdef COMPILE_C | |
3126 | 784 #undef HAVE_MMX |
785 #undef HAVE_MMX2 | |
786 #undef HAVE_3DNOW | |
787 #define RENAME(a) a ## _C | |
788 #include "swscale_template.c" | |
3152 | 789 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
790 |
3126 | 791 #ifdef CAN_COMPILE_X86_ASM |
2576 | 792 |
3126 | 793 //X86 versions |
794 /* | |
795 #undef RENAME | |
796 #undef HAVE_MMX | |
797 #undef HAVE_MMX2 | |
798 #undef HAVE_3DNOW | |
799 #define ARCH_X86 | |
800 #define RENAME(a) a ## _X86 | |
801 #include "swscale_template.c" | |
802 */ | |
803 //MMX versions | |
3152 | 804 #ifdef COMPILE_MMX |
3126 | 805 #undef RENAME |
806 #define HAVE_MMX | |
807 #undef HAVE_MMX2 | |
808 #undef HAVE_3DNOW | |
809 #define RENAME(a) a ## _MMX | |
810 #include "swscale_template.c" | |
3152 | 811 #endif |
2316
bcb229557e9b
fixed alignment (static variables where sometimes not 8-byte aligned)
michael
parents:
2297
diff
changeset
|
812 |
3126 | 813 //MMX2 versions |
3152 | 814 #ifdef COMPILE_MMX2 |
3126 | 815 #undef RENAME |
816 #define HAVE_MMX | |
817 #define HAVE_MMX2 | |
818 #undef HAVE_3DNOW | |
819 #define RENAME(a) a ## _MMX2 | |
820 #include "swscale_template.c" | |
3152 | 821 #endif |
2469 | 822 |
3126 | 823 //3DNOW versions |
3152 | 824 #ifdef COMPILE_3DNOW |
3126 | 825 #undef RENAME |
826 #define HAVE_MMX | |
827 #undef HAVE_MMX2 | |
828 #define HAVE_3DNOW | |
829 #define RENAME(a) a ## _3DNow | |
830 #include "swscale_template.c" | |
3152 | 831 #endif |
2469 | 832 |
3126 | 833 #endif //CAN_COMPILE_X86_ASM |
2469 | 834 |
3126 | 835 // 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
|
836 |
2232
65996b3467d7
MMX & MMX2 optimizations (MMX2 is buggy and commented out)
michael
parents:
2230
diff
changeset
|
837 |
4419 | 838 // 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
|
839 // will use sws_flags from the command line |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
840 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
|
841 int srcSliceH, uint8_t* dst[], int dstStride, int dstbpp, |
3209 | 842 int srcW, int srcH, int dstW, int dstH){ |
2216 | 843 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
844 static SwsContext *context=NULL; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
845 int dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
846 int dstStride3[3]= {dstStride, dstStride>>1, dstStride>>1}; |
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 switch(dstbpp) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
849 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
850 case 8 : dstFormat= IMGFMT_Y8; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
851 case 12: dstFormat= IMGFMT_YV12; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
852 case 15: dstFormat= IMGFMT_BGR15; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
853 case 16: dstFormat= IMGFMT_BGR16; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
854 case 24: dstFormat= IMGFMT_BGR24; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
855 case 32: dstFormat= IMGFMT_BGR32; break; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
856 default: return; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
857 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
858 |
4419 | 859 if(!context) context=getSwsContextFromCmdLine(srcW, srcH, IMGFMT_YV12, dstW, dstH, dstFormat); |
860 | |
4555 | 861 context->swScale(context, src, srcStride, srcSliceY, srcSliceH, dst, dstStride3); |
4419 | 862 } |
863 | |
6540 | 864 void swsGetFlagsAndFilterFromCmdLine(int *flags, SwsFilter **srcFilterParam, SwsFilter **dstFilterParam) |
4419 | 865 { |
866 static int firstTime=1; | |
6540 | 867 *flags=0; |
4419 | 868 |
869 #ifdef ARCH_X86 | |
870 if(gCpuCaps.hasMMX) | |
871 asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions) | |
872 #endif | |
873 if(firstTime) | |
874 { | |
875 firstTime=0; | |
6540 | 876 *flags= SWS_PRINT_INFO; |
4419 | 877 } |
6540 | 878 else if(verbose>1) *flags= SWS_PRINT_INFO; |
4419 | 879 |
880 if(src_filter.lumH) freeVec(src_filter.lumH); | |
881 if(src_filter.lumV) freeVec(src_filter.lumV); | |
882 if(src_filter.chrH) freeVec(src_filter.chrH); | |
883 if(src_filter.chrV) freeVec(src_filter.chrV); | |
884 | |
885 if(sws_lum_gblur!=0.0){ | |
886 src_filter.lumH= getGaussianVec(sws_lum_gblur, 3.0); | |
887 src_filter.lumV= getGaussianVec(sws_lum_gblur, 3.0); | |
888 }else{ | |
889 src_filter.lumH= getIdentityVec(); | |
890 src_filter.lumV= getIdentityVec(); | |
891 } | |
892 | |
893 if(sws_chr_gblur!=0.0){ | |
894 src_filter.chrH= getGaussianVec(sws_chr_gblur, 3.0); | |
895 src_filter.chrV= getGaussianVec(sws_chr_gblur, 3.0); | |
896 }else{ | |
897 src_filter.chrH= getIdentityVec(); | |
898 src_filter.chrV= getIdentityVec(); | |
899 } | |
900 | |
901 if(sws_chr_sharpen!=0.0){ | |
902 SwsVector *g= getConstVec(-1.0, 3); | |
903 SwsVector *id= getConstVec(10.0/sws_chr_sharpen, 1); | |
904 g->coeff[1]=2.0; | |
905 addVec(id, g); | |
906 convVec(src_filter.chrH, id); | |
907 convVec(src_filter.chrV, id); | |
908 freeVec(g); | |
909 freeVec(id); | |
910 } | |
911 | |
912 if(sws_lum_sharpen!=0.0){ | |
913 SwsVector *g= getConstVec(-1.0, 3); | |
914 SwsVector *id= getConstVec(10.0/sws_lum_sharpen, 1); | |
915 g->coeff[1]=2.0; | |
916 addVec(id, g); | |
917 convVec(src_filter.lumH, id); | |
918 convVec(src_filter.lumV, id); | |
919 freeVec(g); | |
920 freeVec(id); | |
921 } | |
922 | |
923 if(sws_chr_hshift) | |
924 shiftVec(src_filter.chrH, sws_chr_hshift); | |
925 | |
926 if(sws_chr_vshift) | |
927 shiftVec(src_filter.chrV, sws_chr_vshift); | |
928 | |
929 normalizeVec(src_filter.chrH, 1.0); | |
930 normalizeVec(src_filter.chrV, 1.0); | |
931 normalizeVec(src_filter.lumH, 1.0); | |
932 normalizeVec(src_filter.lumV, 1.0); | |
933 | |
934 if(verbose > 1) printVec(src_filter.chrH); | |
935 if(verbose > 1) printVec(src_filter.lumH); | |
936 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
937 switch(sws_flags) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
938 { |
6540 | 939 case 0: *flags|= SWS_FAST_BILINEAR; break; |
940 case 1: *flags|= SWS_BILINEAR; break; | |
941 case 2: *flags|= SWS_BICUBIC; break; | |
942 case 3: *flags|= SWS_X; break; | |
943 case 4: *flags|= SWS_POINT; break; | |
944 case 5: *flags|= SWS_AREA; break; | |
6580 | 945 case 6: *flags|= SWS_BICUBLIN; break; |
6540 | 946 default:*flags|= SWS_BILINEAR; break; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
947 } |
6540 | 948 |
949 *srcFilterParam= &src_filter; | |
950 *dstFilterParam= NULL; | |
951 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
952 |
6540 | 953 // will use sws_flags & src_filter (from cmd line) |
954 SwsContext *getSwsContextFromCmdLine(int srcW, int srcH, int srcFormat, int dstW, int dstH, int dstFormat) | |
955 { | |
956 int flags; | |
957 SwsFilter *dstFilterParam, *srcFilterParam; | |
958 swsGetFlagsAndFilterFromCmdLine(&flags, &srcFilterParam, &dstFilterParam); | |
959 | |
960 return getSwsContext(srcW, srcH, srcFormat, dstW, dstH, dstFormat, flags, srcFilterParam, dstFilterParam); | |
4419 | 961 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
962 |
2216 | 963 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
964 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
|
965 int srcW, int dstW, int filterAlign, int one, int flags, |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
966 SwsVector *srcFilter, SwsVector *dstFilter) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
967 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
968 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
969 int filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
970 int filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
971 int minFilterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
972 double *filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
973 double *filter2=NULL; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
974 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
975 if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
976 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
|
977 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
978 |
4563
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
979 // Note the +1 is for the MMXscaler which reads over the end |
4419 | 980 *filterPos = (int16_t*)memalign(8, (dstW+1)*sizeof(int16_t)); |
981 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
982 if(ABS(xInc - 0x10000) <10) // unscaled |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
983 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
984 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
985 filterSize= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
986 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
987 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
|
988 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
989 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
990 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
991 filter[i*filterSize]=1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
992 (*filterPos)[i]=i; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
993 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
994 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
995 } |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
996 else if(flags&SWS_POINT) // lame looking point sampling mode |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
997 { |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
998 int i; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
999 int xDstInSrc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1000 filterSize= 1; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1001 filter= (double*)memalign(8, dstW*sizeof(double)*filterSize); |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1002 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1003 xDstInSrc= xInc/2 - 0x8000; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1004 for(i=0; i<dstW; i++) |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1005 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
1006 int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16; |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1007 |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1008 (*filterPos)[i]= xx; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1009 filter[i]= 1.0; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1010 xDstInSrc+= xInc; |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1011 } |
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
1012 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1013 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
|
1014 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1015 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1016 int xDstInSrc; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1017 if (flags&SWS_BICUBIC) filterSize= 4; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1018 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
|
1019 else filterSize= 2; // SWS_BILINEAR / SWS_AREA |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1020 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
|
1021 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1022 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1023 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1024 { |
4492
8303419cc60f
filterPos bugfix (only nearest neighbor scaling was affected)
michael
parents:
4481
diff
changeset
|
1025 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
|
1026 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1027 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1028 (*filterPos)[i]= xx; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1029 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1030 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1031 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
|
1032 double y1,y2,y3,y4; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1033 double A= -0.6; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1034 if(flags & SWS_BICUBIC){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1035 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1036 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
|
1037 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
|
1038 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
|
1039 y4 = ( + A*d*d - A*d*d*d); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1040 }else{ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1041 // cubic interpolation (derived it myself) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1042 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
|
1043 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
|
1044 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
|
1045 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
|
1046 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1047 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1048 filter[i*filterSize + 0]= y1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1049 filter[i*filterSize + 1]= y2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1050 filter[i*filterSize + 2]= y3; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1051 filter[i*filterSize + 3]= y4; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1052 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1053 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1054 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1055 //Bilinear upscale / linear interpolate / Area averaging |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1056 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1057 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1058 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
|
1059 double coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1060 if(coeff<0) coeff=0; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1061 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1062 xx++; |
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 xDstInSrc+= xInc; |
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 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1068 else // downscale |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1069 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1070 int xDstInSrc; |
4806 | 1071 ASSERT(dstW <= srcW) |
1072 | |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1073 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
|
1074 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
|
1075 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
|
1076 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
|
1077 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
|
1078 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1079 xDstInSrc= xInc/2 - 0x8000; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1080 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1081 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1082 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
|
1083 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1084 (*filterPos)[i]= xx; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1085 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1086 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1087 double d= ABS((xx<<16) - xDstInSrc)/(double)xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1088 double coeff; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1089 if((flags & SWS_BICUBIC) || (flags & SWS_X)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1090 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1091 double A= -0.75; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1092 // d*=2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1093 // Equation is from VirtualDub |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1094 if(d<1.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1095 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
|
1096 else if(d<2.0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1097 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
|
1098 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1099 coeff=0.0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1100 } |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1101 else if(flags & SWS_AREA) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1102 { |
4402
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1103 double srcPixelSize= (1<<16)/(double)xInc; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1104 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
|
1105 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
|
1106 else coeff=0.0; |
67abbf501b02
area averageing scaling support (-sws 5) (is identical to bilinear for upscale)
michael
parents:
4401
diff
changeset
|
1107 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1108 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1109 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1110 coeff= 1.0 - d; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1111 if(coeff<0) coeff=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1112 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1113 filter[i*filterSize + j]= coeff; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1114 xx++; |
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 xDstInSrc+= xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1117 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1118 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1119 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1120 /* apply src & dst Filter to filter -> filter2 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1121 free(filter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1122 */ |
4806 | 1123 ASSERT(filterSize>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1124 filter2Size= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1125 if(srcFilter) filter2Size+= srcFilter->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1126 if(dstFilter) filter2Size+= dstFilter->length - 1; |
4806 | 1127 ASSERT(filter2Size>0) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1128 filter2= (double*)memalign(8, filter2Size*dstW*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1129 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1130 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1131 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1132 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1133 SwsVector scaleFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1134 SwsVector *outVec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1135 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1136 scaleFilter.coeff= filter + i*filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1137 scaleFilter.length= filterSize; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1138 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1139 if(srcFilter) outVec= getConvVec(srcFilter, &scaleFilter); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1140 else outVec= &scaleFilter; |
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 ASSERT(outVec->length == filter2Size) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1143 //FIXME dstFilter |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1144 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1145 for(j=0; j<outVec->length; j++) |
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 filter2[i*filter2Size + j]= outVec->coeff[j]; |
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 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1150 (*filterPos)[i]+= (filterSize-1)/2 - (filter2Size-1)/2; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1151 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1152 if(outVec != &scaleFilter) freeVec(outVec); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1153 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1154 free(filter); filter=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1155 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1156 /* 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
|
1157 // 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
|
1158 minFilterSize= 0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1159 for(i=dstW-1; i>=0; i--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1160 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1161 int min= filter2Size; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1162 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1163 double cutOff=0.0; |
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 /* 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
|
1166 for(j=0; j<filter2Size; j++) |
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 int k; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1169 cutOff += ABS(filter2[i*filter2Size]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1170 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1171 if(cutOff > SWS_MAX_REDUCE_CUTOFF) 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 /* preserve Monotonicity because the core cant handle the filter otherwise */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1174 if(i<dstW-1 && (*filterPos)[i] >= (*filterPos)[i+1]) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1175 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1176 // Move filter coeffs left |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1177 for(k=1; k<filter2Size; k++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1178 filter2[i*filter2Size + k - 1]= filter2[i*filter2Size + k]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1179 filter2[i*filter2Size + k - 1]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1180 (*filterPos)[i]++; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1181 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1182 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1183 cutOff=0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1184 /* count near zeros on the right */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1185 for(j=filter2Size-1; j>0; j--) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1186 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1187 cutOff += ABS(filter2[i*filter2Size + j]); |
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 if(cutOff > SWS_MAX_REDUCE_CUTOFF) break; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1190 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 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1193 if(min>minFilterSize) minFilterSize= min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1194 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1195 |
4806 | 1196 ASSERT(minFilterSize > 0) |
4419 | 1197 filterSize= (minFilterSize +(filterAlign-1)) & (~(filterAlign-1)); |
4806 | 1198 ASSERT(filterSize > 0) |
4419 | 1199 filter= (double*)memalign(8, filterSize*dstW*sizeof(double)); |
1200 *outFilterSize= filterSize; | |
1201 | |
5937 | 1202 if(flags&SWS_PRINT_INFO) |
6492 | 1203 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
|
1204 /* try to reduce the filter-size (step2 reduce it) */ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1205 for(i=0; i<dstW; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1206 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1207 int j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1208 |
4419 | 1209 for(j=0; j<filterSize; j++) |
1210 { | |
1211 if(j>=filter2Size) filter[i*filterSize + j]= 0.0; | |
1212 else filter[i*filterSize + j]= filter2[i*filter2Size + j]; | |
1213 } | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1214 } |
4419 | 1215 free(filter2); filter2=NULL; |
1216 | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1217 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1218 //FIXME try to align filterpos if possible |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1219 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1220 //fix borders |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1221 for(i=0; i<dstW; i++) |
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 int j; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1224 if((*filterPos)[i] < 0) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1225 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1226 // Move filter coeffs left to compensate for filterPos |
4419 | 1227 for(j=1; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1228 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1229 int left= MAX(j + (*filterPos)[i], 0); |
4419 | 1230 filter[i*filterSize + left] += filter[i*filterSize + j]; |
1231 filter[i*filterSize + j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1232 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1233 (*filterPos)[i]= 0; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1234 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1235 |
4419 | 1236 if((*filterPos)[i] + filterSize > srcW) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1237 { |
4419 | 1238 int shift= (*filterPos)[i] + filterSize - srcW; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1239 // Move filter coeffs right to compensate for filterPos |
4419 | 1240 for(j=filterSize-2; j>=0; j--) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1241 { |
4419 | 1242 int right= MIN(j + shift, filterSize-1); |
1243 filter[i*filterSize +right] += filter[i*filterSize +j]; | |
1244 filter[i*filterSize +j]=0; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1245 } |
4419 | 1246 (*filterPos)[i]= srcW - filterSize; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1247 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1248 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1249 |
4419 | 1250 // Note the +1 is for the MMXscaler which reads over the end |
1251 *outFilter= (int16_t*)memalign(8, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
1252 memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t)); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1253 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1254 /* Normalize & Store in outFilter */ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1255 for(i=0; i<dstW; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1256 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1257 int j; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1258 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1259 double scale= one; |
4419 | 1260 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1261 { |
4419 | 1262 sum+= filter[i*filterSize + j]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1263 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1264 scale/= sum; |
4419 | 1265 for(j=0; j<filterSize; j++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1266 { |
4419 | 1267 (*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
|
1268 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1269 } |
4563
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 (*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
|
1272 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
|
1273 { |
835d4fce67d5
fixing green line at right side bug (reported by Nick Kurshev <nickols_k@mail.ru>)
michael
parents:
4558
diff
changeset
|
1274 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
|
1275 (*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
|
1276 } |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1277 |
4419 | 1278 free(filter); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1279 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1280 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1281 #ifdef ARCH_X86 |
5452 | 1282 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
|
1283 { |
5452 | 1284 uint8_t *fragmentA; |
1285 int imm8OfPShufW1A; | |
1286 int imm8OfPShufW2A; | |
1287 int fragmentLengthA; | |
1288 uint8_t *fragmentB; | |
1289 int imm8OfPShufW1B; | |
1290 int imm8OfPShufW2B; | |
1291 int fragmentLengthB; | |
1292 int fragmentPos; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1293 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1294 int xpos, i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1295 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1296 // create an optimized horizontal scaling routine |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1297 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1298 //code fragment |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1299 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1300 asm volatile( |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1301 "jmp 9f \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1302 // Begin |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1303 "0: \n\t" |
5452 | 1304 "movq (%%edx, %%eax), %%mm3 \n\t" |
1305 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1306 "movd 1(%%ecx, %%esi), %%mm1 \n\t" | |
1307 "punpcklbw %%mm7, %%mm1 \n\t" | |
1308 "punpcklbw %%mm7, %%mm0 \n\t" | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1309 "pshufw $0xFF, %%mm1, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1310 "1: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1311 "pshufw $0xFF, %%mm0, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1312 "2: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1313 "psubw %%mm1, %%mm0 \n\t" |
5452 | 1314 "movl 8(%%ebx, %%eax), %%esi \n\t" |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1315 "pmullw %%mm3, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1316 "psllw $7, %%mm1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1317 "paddw %%mm1, %%mm0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1318 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1319 "movq %%mm0, (%%edi, %%eax) \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1320 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1321 "addl $8, %%eax \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1322 // End |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1323 "9: \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1324 // "int $3\n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1325 "leal 0b, %0 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1326 "leal 1b, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1327 "leal 2b, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1328 "decl %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1329 "decl %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1330 "subl %0, %1 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1331 "subl %0, %2 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1332 "leal 9b, %3 \n\t" |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1333 "subl %0, %3 \n\t" |
5452 | 1334 |
1335 | |
1336 :"=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), | |
1337 "=r" (fragmentLengthA) | |
1338 ); | |
1339 | |
1340 asm volatile( | |
1341 "jmp 9f \n\t" | |
1342 // Begin | |
1343 "0: \n\t" | |
1344 "movq (%%edx, %%eax), %%mm3 \n\t" | |
1345 "movd (%%ecx, %%esi), %%mm0 \n\t" | |
1346 "punpcklbw %%mm7, %%mm0 \n\t" | |
1347 "pshufw $0xFF, %%mm0, %%mm1 \n\t" | |
1348 "1: \n\t" | |
1349 "pshufw $0xFF, %%mm0, %%mm0 \n\t" | |
1350 "2: \n\t" | |
1351 "psubw %%mm1, %%mm0 \n\t" | |
1352 "movl 8(%%ebx, %%eax), %%esi \n\t" | |
1353 "pmullw %%mm3, %%mm0 \n\t" | |
1354 "psllw $7, %%mm1 \n\t" | |
1355 "paddw %%mm1, %%mm0 \n\t" | |
1356 | |
1357 "movq %%mm0, (%%edi, %%eax) \n\t" | |
1358 | |
1359 "addl $8, %%eax \n\t" | |
1360 // End | |
1361 "9: \n\t" | |
1362 // "int $3\n\t" | |
1363 "leal 0b, %0 \n\t" | |
1364 "leal 1b, %1 \n\t" | |
1365 "leal 2b, %2 \n\t" | |
1366 "decl %1 \n\t" | |
1367 "decl %2 \n\t" | |
1368 "subl %0, %1 \n\t" | |
1369 "subl %0, %2 \n\t" | |
1370 "leal 9b, %3 \n\t" | |
1371 "subl %0, %3 \n\t" | |
1372 | |
1373 | |
1374 :"=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), | |
1375 "=r" (fragmentLengthB) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1376 ); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1377 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1378 xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers |
5452 | 1379 fragmentPos=0; |
1380 | |
1381 for(i=0; i<dstW/numSplits; i++) | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1382 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1383 int xx=xpos>>16; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1384 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1385 if((i&3) == 0) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1386 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1387 int a=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1388 int b=((xpos+xInc)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1389 int c=((xpos+xInc*2)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1390 int d=((xpos+xInc*3)>>16) - xx; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1391 |
5452 | 1392 filter[i ] = (( xpos & 0xFFFF) ^ 0xFFFF)>>9; |
1393 filter[i+1] = (((xpos+xInc ) & 0xFFFF) ^ 0xFFFF)>>9; | |
1394 filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9; | |
1395 filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9; | |
1396 filterPos[i/2]= xx; | |
1397 | |
1398 if(d+1<4) | |
1399 { | |
1400 int maxShift= 3-(d+1); | |
1401 int shift=0; | |
1402 | |
1403 memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1404 |
5452 | 1405 funnyCode[fragmentPos + imm8OfPShufW1B]= |
1406 (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6); | |
1407 funnyCode[fragmentPos + imm8OfPShufW2B]= | |
1408 a | (b<<2) | (c<<4) | (d<<6); | |
1409 | |
1410 if(i+3>=dstW) shift=maxShift; //avoid overread | |
1411 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align | |
1412 | |
1413 if(shift && i>=shift) | |
1414 { | |
1415 funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift; | |
1416 funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift; | |
1417 filterPos[i/2]-=shift; | |
1418 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1419 |
5452 | 1420 fragmentPos+= fragmentLengthB; |
1421 } | |
1422 else | |
1423 { | |
1424 int maxShift= 3-d; | |
1425 int shift=0; | |
1426 | |
1427 memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA); | |
1428 | |
1429 funnyCode[fragmentPos + imm8OfPShufW1A]= | |
1430 funnyCode[fragmentPos + imm8OfPShufW2A]= | |
1431 a | (b<<2) | (c<<4) | (d<<6); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1432 |
5452 | 1433 if(i+4>=dstW) shift=maxShift; //avoid overread |
1434 else if((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align | |
1435 | |
1436 if(shift && i>=shift) | |
1437 { | |
1438 funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift; | |
1439 funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift; | |
1440 filterPos[i/2]-=shift; | |
1441 } | |
1442 | |
1443 fragmentPos+= fragmentLengthA; | |
1444 } | |
1445 | |
1446 funnyCode[fragmentPos]= RET; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1447 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1448 xpos+=xInc; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1449 } |
5452 | 1450 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
|
1451 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1452 #endif // ARCH_X86 |
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 //FIXME remove |
2216 | 1455 void SwScale_Init(){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1456 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1457 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1458 static void globalInit(){ |
2216 | 1459 // generating tables: |
1460 int i; | |
3344 | 1461 for(i=0; i<768; i++){ |
1462 int c= MIN(MAX(i-256, 0), 255); | |
1463 clip_table[i]=c; | |
2584 | 1464 } |
3344 | 1465 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1466 cpuCaps= gCpuCaps; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1467 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1468 #ifdef RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1469 #ifdef CAN_COMPILE_X86_ASM |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1470 // ordered per speed fasterst first |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1471 if(gCpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1472 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1473 else if(gCpuCaps.has3DNow) |
4281 | 1474 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1475 else if(gCpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1476 swScale= swScale_MMX; |
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 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1480 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1481 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1482 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1483 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1484 #else //RUNTIME_CPUDETECT |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1485 #ifdef HAVE_MMX2 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1486 swScale= swScale_MMX2; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1487 cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1488 #elif defined (HAVE_3DNOW) |
4281 | 1489 swScale= swScale_3DNow; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1490 cpuCaps.hasMMX2 = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1491 #elif defined (HAVE_MMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1492 swScale= swScale_MMX; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1493 cpuCaps.hasMMX2 = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1494 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1495 swScale= swScale_C; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1496 cpuCaps.hasMMX2 = cpuCaps.hasMMX = cpuCaps.has3DNow = 0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1497 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1498 #endif //!RUNTIME_CPUDETECT |
3126 | 1499 } |
2584 | 1500 |
6492 | 1501 static void PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1502 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ | |
1503 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1504 /* Copy Y plane */ | |
1505 if(dstStride[0]==srcStride[0]) | |
1506 memcpy(dst, src[0], srcSliceH*dstStride[0]); | |
1507 else | |
1508 { | |
1509 int i; | |
1510 uint8_t *srcPtr= src[0]; | |
1511 uint8_t *dstPtr= dst; | |
1512 for(i=0; i<srcSliceH; i++) | |
1513 { | |
1514 memcpy(dstPtr, srcPtr, srcStride[0]); | |
1515 srcPtr+= srcStride[0]; | |
1516 dstPtr+= dstStride[0]; | |
1517 } | |
1518 } | |
1519 dst = dstParam[1] + dstStride[1]*srcSliceY; | |
1520 if(c->srcFormat==IMGFMT_YV12) | |
1521 interleaveBytes( src[1],src[2],dst,c->srcW,srcSliceH,srcStride[1],srcStride[2],dstStride[0] ); | |
1522 else /* I420 & IYUV */ | |
1523 interleaveBytes( src[2],src[1],dst,c->srcW,srcSliceH,srcStride[2],srcStride[1],dstStride[0] ); | |
1524 } | |
1525 | |
1526 | |
4554 | 1527 /* Warper functions for yuv2bgr */ |
1528 static void planarYuvToBgr(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
4555 | 1529 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1530 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
4554 | 1531 |
1532 if(c->srcFormat==IMGFMT_YV12) | |
4555 | 1533 yuv2rgb( dst,src[0],src[1],src[2],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
4554 | 1534 else /* I420 & IYUV */ |
4555 | 1535 yuv2rgb( dst,src[0],src[2],src[1],c->srcW,srcSliceH,dstStride[0],srcStride[0],srcStride[1] ); |
1536 } | |
1537 | |
6503 | 1538 static void PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
6492 | 1539 int srcSliceH, uint8_t* dstParam[], int dstStride[]){ |
1540 uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY; | |
1541 | |
1542 if(c->srcFormat==IMGFMT_YV12) | |
1543 yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1544 else /* I420 & IYUV */ | |
1545 yv12toyuy2( src[0],src[2],src[1],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] ); | |
1546 } | |
1547 | |
4567 | 1548 static void bgr24to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1549 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1550 | |
1551 if(dstStride[0]*3==srcStride[0]*4) | |
4581 | 1552 rgb24to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 1553 else |
1554 { | |
1555 int i; | |
1556 uint8_t *srcPtr= src[0]; | |
1557 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1558 | |
1559 for(i=0; i<srcSliceH; i++) | |
1560 { | |
4581 | 1561 rgb24to32(srcPtr, dstPtr, c->srcW*3); |
4567 | 1562 srcPtr+= srcStride[0]; |
1563 dstPtr+= dstStride[0]; | |
1564 } | |
1565 } | |
1566 } | |
1567 | |
6492 | 1568 static void bgr24to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1569 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1570 | |
1571 if(dstStride[0]*3==srcStride[0]*2) | |
1572 rgb24to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1573 else | |
1574 { | |
1575 int i; | |
1576 uint8_t *srcPtr= src[0]; | |
1577 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1578 | |
1579 for(i=0; i<srcSliceH; i++) | |
1580 { | |
1581 rgb24to16(srcPtr, dstPtr, c->srcW*3); | |
1582 srcPtr+= srcStride[0]; | |
1583 dstPtr+= dstStride[0]; | |
1584 } | |
1585 } | |
1586 } | |
1587 | |
1588 static void bgr24to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1589 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1590 | |
1591 if(dstStride[0]*3==srcStride[0]*2) | |
1592 rgb24to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1593 else | |
1594 { | |
1595 int i; | |
1596 uint8_t *srcPtr= src[0]; | |
1597 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1598 | |
1599 for(i=0; i<srcSliceH; i++) | |
1600 { | |
1601 rgb24to15(srcPtr, dstPtr, c->srcW*3); | |
1602 srcPtr+= srcStride[0]; | |
1603 dstPtr+= dstStride[0]; | |
1604 } | |
1605 } | |
1606 } | |
1607 | |
4567 | 1608 static void bgr32to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1609 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1610 | |
1611 if(dstStride[0]*4==srcStride[0]*3) | |
4581 | 1612 rgb32to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); |
4567 | 1613 else |
1614 { | |
1615 int i; | |
1616 uint8_t *srcPtr= src[0]; | |
1617 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1618 | |
1619 for(i=0; i<srcSliceH; i++) | |
1620 { | |
4581 | 1621 rgb32to24(srcPtr, dstPtr, c->srcW<<2); |
1622 srcPtr+= srcStride[0]; | |
1623 dstPtr+= dstStride[0]; | |
1624 } | |
1625 } | |
1626 } | |
1627 | |
6492 | 1628 static void bgr32to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1629 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1630 | |
1631 if(dstStride[0]*4==srcStride[0]*2) | |
1632 rgb32to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1633 else | |
1634 { | |
1635 int i; | |
1636 uint8_t *srcPtr= src[0]; | |
1637 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1638 | |
1639 for(i=0; i<srcSliceH; i++) | |
1640 { | |
1641 rgb32to16(srcPtr, dstPtr, c->srcW<<2); | |
1642 srcPtr+= srcStride[0]; | |
1643 dstPtr+= dstStride[0]; | |
1644 } | |
1645 } | |
1646 } | |
1647 | |
1648 static void bgr32to15Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1649 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1650 | |
1651 if(dstStride[0]*4==srcStride[0]*2) | |
1652 rgb32to15(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1653 else | |
1654 { | |
1655 int i; | |
1656 uint8_t *srcPtr= src[0]; | |
1657 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1658 | |
1659 for(i=0; i<srcSliceH; i++) | |
1660 { | |
1661 rgb32to15(srcPtr, dstPtr, c->srcW<<2); | |
1662 srcPtr+= srcStride[0]; | |
1663 dstPtr+= dstStride[0]; | |
1664 } | |
1665 } | |
1666 } | |
1667 | |
4581 | 1668 static void bgr15to16Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1669 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1670 | |
1671 if(dstStride[0]==srcStride[0]) | |
1672 rgb15to16(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1673 else | |
1674 { | |
1675 int i; | |
1676 uint8_t *srcPtr= src[0]; | |
1677 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1678 | |
1679 for(i=0; i<srcSliceH; i++) | |
1680 { | |
1681 rgb15to16(srcPtr, dstPtr, c->srcW<<1); | |
4567 | 1682 srcPtr+= srcStride[0]; |
1683 dstPtr+= dstStride[0]; | |
1684 } | |
1685 } | |
1686 } | |
1687 | |
6492 | 1688 static void bgr15to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1689 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1690 | |
1691 if(dstStride[0]*2==srcStride[0]*3) | |
1692 rgb15to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1693 else | |
1694 { | |
1695 int i; | |
1696 uint8_t *srcPtr= src[0]; | |
1697 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1698 | |
1699 for(i=0; i<srcSliceH; i++) | |
1700 { | |
1701 rgb15to24(srcPtr, dstPtr, c->srcW<<1); | |
1702 srcPtr+= srcStride[0]; | |
1703 dstPtr+= dstStride[0]; | |
1704 } | |
1705 } | |
1706 } | |
1707 | |
1708 static void bgr15to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1709 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1710 | |
1711 if(dstStride[0]*2==srcStride[0]*4) | |
1712 rgb15to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1713 else | |
1714 { | |
1715 int i; | |
1716 uint8_t *srcPtr= src[0]; | |
1717 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1718 | |
1719 for(i=0; i<srcSliceH; i++) | |
1720 { | |
1721 rgb15to32(srcPtr, dstPtr, c->srcW<<1); | |
1722 srcPtr+= srcStride[0]; | |
1723 dstPtr+= dstStride[0]; | |
1724 } | |
1725 } | |
1726 } | |
1727 | |
1728 static void bgr16to24Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1729 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1730 | |
1731 if(dstStride[0]*2==srcStride[0]*3) | |
1732 rgb16to24(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1733 else | |
1734 { | |
1735 int i; | |
1736 uint8_t *srcPtr= src[0]; | |
1737 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1738 | |
1739 for(i=0; i<srcSliceH; i++) | |
1740 { | |
1741 rgb16to24(srcPtr, dstPtr, c->srcW<<1); | |
1742 srcPtr+= srcStride[0]; | |
1743 dstPtr+= dstStride[0]; | |
1744 } | |
1745 } | |
1746 } | |
1747 | |
1748 static void bgr16to32Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, | |
1749 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1750 | |
1751 if(dstStride[0]*2==srcStride[0]*4) | |
1752 rgb16to32(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]); | |
1753 else | |
1754 { | |
1755 int i; | |
1756 uint8_t *srcPtr= src[0]; | |
1757 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
1758 | |
1759 for(i=0; i<srcSliceH; i++) | |
1760 { | |
1761 rgb16to32(srcPtr, dstPtr, c->srcW<<1); | |
1762 srcPtr+= srcStride[0]; | |
1763 dstPtr+= dstStride[0]; | |
1764 } | |
1765 } | |
1766 } | |
1767 | |
4623 | 1768 static void bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
1769 int srcSliceH, uint8_t* dst[], int dstStride[]){ | |
1770 | |
1771 rgb24toyv12( | |
1772 src[0], | |
1773 dst[0]+ srcSliceY *dstStride[0], | |
1774 dst[1]+(srcSliceY>>1)*dstStride[1], | |
1775 dst[2]+(srcSliceY>>1)*dstStride[2], | |
1776 c->srcW, srcSliceH, | |
1777 dstStride[0], dstStride[1], srcStride[0]); | |
1778 } | |
1779 | |
6582
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1780 static void yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY, |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1781 int srcSliceH, uint8_t* dst[], int dstStride[]){ |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1782 int i; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1783 |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1784 /* copy Y */ |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1785 if(srcStride[0]==dstStride[0]) |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1786 memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1787 else{ |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1788 uint8_t *srcPtr= src[0]; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1789 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1790 |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1791 for(i=0; i<srcSliceH; i++) |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1792 { |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1793 memcpy(dstPtr, srcPtr, c->srcW); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1794 srcPtr+= srcStride[0]; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1795 dstPtr+= dstStride[0]; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1796 } |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1797 } |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1798 |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1799 if(c->dstFormat==IMGFMT_YV12){ |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1800 planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1801 planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1802 }else{ |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1803 planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1804 planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1805 } |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1806 } |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
1807 |
6503 | 1808 /** |
1809 * bring pointers in YUV order instead of YVU | |
1810 */ | |
6517 | 1811 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
|
1812 if(format == IMGFMT_YV12 || format == IMGFMT_YVU9){ |
6503 | 1813 sortedP[0]= p[0]; |
1814 sortedP[1]= p[1]; | |
1815 sortedP[2]= p[2]; | |
6517 | 1816 sortedStride[0]= stride[0]; |
1817 sortedStride[1]= stride[1]; | |
1818 sortedStride[2]= stride[2]; | |
6503 | 1819 } |
6517 | 1820 else if(isPacked(format) || isGray(format)) |
6503 | 1821 { |
1822 sortedP[0]= p[0]; | |
1823 sortedP[1]= | |
1824 sortedP[2]= NULL; | |
6517 | 1825 sortedStride[0]= stride[0]; |
6503 | 1826 sortedStride[1]= |
1827 sortedStride[2]= 0; | |
1828 } | |
1829 else /* I420 */ | |
1830 { | |
1831 sortedP[0]= p[0]; | |
1832 sortedP[1]= p[2]; | |
1833 sortedP[2]= p[1]; | |
6517 | 1834 sortedStride[0]= stride[0]; |
1835 sortedStride[1]= stride[2]; | |
1836 sortedStride[2]= stride[1]; | |
6503 | 1837 } |
1838 } | |
4567 | 1839 |
4555 | 1840 /* unscaled copy like stuff (assumes nearly identical formats) */ |
1841 static void simpleCopy(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY, | |
6503 | 1842 int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){ |
4555 | 1843 |
1844 int srcStride[3]; | |
6503 | 1845 int dstStride[3]; |
4555 | 1846 uint8_t *src[3]; |
1847 uint8_t *dst[3]; | |
1848 | |
6517 | 1849 orderYUV(c->srcFormat, src, srcStride, srcParam, srcStrideParam); |
1850 orderYUV(c->dstFormat, dst, dstStride, dstParam, dstStrideParam); | |
4555 | 1851 |
1852 if(isPacked(c->srcFormat)) | |
1853 { | |
1854 if(dstStride[0]==srcStride[0]) | |
1855 memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]); | |
1856 else | |
1857 { | |
1858 int i; | |
1859 uint8_t *srcPtr= src[0]; | |
1860 uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY; | |
4558 | 1861 int length=0; |
1862 | |
1863 /* universal length finder */ | |
4649 | 1864 while(length+c->srcW <= ABS(dstStride[0]) |
1865 && length+c->srcW <= ABS(srcStride[0])) length+= c->srcW; | |
4558 | 1866 ASSERT(length!=0); |
4555 | 1867 |
1868 for(i=0; i<srcSliceH; i++) | |
1869 { | |
1870 memcpy(dstPtr, srcPtr, length); | |
1871 srcPtr+= srcStride[0]; | |
1872 dstPtr+= dstStride[0]; | |
1873 } | |
1874 } | |
1875 } | |
1876 else | |
6503 | 1877 { /* Planar YUV or gray */ |
4555 | 1878 int plane; |
1879 for(plane=0; plane<3; plane++) | |
1880 { | |
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
|
1881 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
|
1882 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
|
1883 int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample); |
6503 | 1884 |
1885 if((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0) | |
6492 | 1886 { |
6503 | 1887 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
|
1888 memset(dst[plane], 128, dstStride[plane]*height); |
6492 | 1889 } |
1890 else | |
1891 { | |
6503 | 1892 if(dstStride[plane]==srcStride[plane]) |
1893 memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]); | |
1894 else | |
4555 | 1895 { |
6503 | 1896 int i; |
1897 uint8_t *srcPtr= src[plane]; | |
1898 uint8_t *dstPtr= dst[plane] + dstStride[plane]*y; | |
1899 for(i=0; i<height; i++) | |
1900 { | |
1901 memcpy(dstPtr, srcPtr, length); | |
1902 srcPtr+= srcStride[plane]; | |
1903 dstPtr+= dstStride[plane]; | |
1904 } | |
4555 | 1905 } |
1906 } | |
1907 } | |
1908 } | |
4554 | 1909 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1910 |
6503 | 1911 static int remove_dup_fourcc(int fourcc) |
6492 | 1912 { |
1913 switch(fourcc) | |
1914 { | |
1915 case IMGFMT_IYUV: return IMGFMT_I420; | |
1916 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
|
1917 case IMGFMT_IF09: return IMGFMT_YVU9; |
6492 | 1918 default: return fourcc; |
1919 } | |
1920 } | |
1921 | |
6517 | 1922 static void getSubSampleFactors(int *h, int *v, int format){ |
1923 switch(format){ | |
1924 case IMGFMT_YUY2: | |
1925 *h=1; | |
1926 *v=0; | |
1927 break; | |
1928 case IMGFMT_YV12: | |
1929 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
|
1930 case IMGFMT_Y800: //FIXME remove after different subsamplings are fully implemented |
6517 | 1931 *h=1; |
1932 *v=1; | |
1933 break; | |
1934 case IMGFMT_YVU9: | |
1935 *h=2; | |
1936 *v=2; | |
1937 break; | |
1938 default: | |
1939 *h=0; | |
1940 *v=0; | |
1941 break; | |
1942 } | |
1943 } | |
1944 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1945 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
|
1946 SwsFilter *srcFilter, SwsFilter *dstFilter){ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1947 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1948 SwsContext *c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1949 int i; |
4554 | 1950 int usesFilter; |
6503 | 1951 int unscaled; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1952 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
|
1953 #ifdef ARCH_X86 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1954 if(gCpuCaps.hasMMX) |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
1955 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
|
1956 #endif |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1957 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
|
1958 //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
|
1959 //srcFormat= IMGFMT_YVU9; |
4481 | 1960 /* avoid dupplicate Formats, so we dont need to check to much */ |
6492 | 1961 srcFormat = remove_dup_fourcc(srcFormat); |
1962 dstFormat = remove_dup_fourcc(dstFormat); | |
6503 | 1963 |
1964 unscaled = (srcW == dstW && srcH == dstH); | |
1965 | |
1966 if(!isSupportedIn(srcFormat)) | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1967 { |
6503 | 1968 MSG_ERR("swScaler: %s is not supported as input format\n", vo_format_name(srcFormat)); |
1969 return NULL; | |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1970 } |
6503 | 1971 if(!isSupportedOut(dstFormat)) |
1972 { | |
1973 MSG_ERR("swScaler: %s is not supported as output format\n", vo_format_name(dstFormat)); | |
1974 return NULL; | |
1975 } | |
1976 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1977 /* sanity check */ |
4529
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1978 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
|
1979 { |
6492 | 1980 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
|
1981 srcW, srcH, dstW, dstH); |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1982 return NULL; |
8f982d17c479
printing error messages if something is wrong instead of just return NULL;
michael
parents:
4492
diff
changeset
|
1983 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1984 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1985 if(!dstFilter) dstFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1986 if(!srcFilter) srcFilter= &dummyFilter; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1987 |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1988 c= memalign(64, sizeof(SwsContext)); |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
1989 memset(c, 0, sizeof(SwsContext)); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1990 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1991 c->srcW= srcW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1992 c->srcH= srcH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1993 c->dstW= dstW; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1994 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
|
1995 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
|
1996 c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1997 c->flags= flags; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1998 c->dstFormat= dstFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
1999 c->srcFormat= srcFormat; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2000 |
4554 | 2001 usesFilter=0; |
2002 if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesFilter=1; | |
2003 if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesFilter=1; | |
2004 if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesFilter=1; | |
2005 if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesFilter=1; | |
2006 if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesFilter=1; | |
2007 if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesFilter=1; | |
2008 if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesFilter=1; | |
2009 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
|
2010 |
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
|
2011 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
|
2012 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
|
2013 |
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
|
2014 // 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
|
2015 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
|
2016 |
6540 | 2017 // drop some chroma lines if the user wants it |
2018 c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT; | |
2019 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
|
2020 |
6540 | 2021 // 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
|
2022 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
|
2023 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
|
2024 |
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
|
2025 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
|
2026 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
|
2027 |
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
|
2028 // 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
|
2029 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
|
2030 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
|
2031 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
|
2032 c->chrDstH= -((-dstH) >> c->chrDstVSubSample); |
4554 | 2033 |
6578 | 2034 if(isBGR(dstFormat)) |
2035 c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_RGB, c->table_rV, c->table_gU, c->table_gV, c->table_bU); | |
2036 if(isRGB(dstFormat)) | |
2037 c->yuvTable= yuv2rgb_c_init(dstFormat & 0xFF, MODE_BGR, c->table_rV, c->table_gU, c->table_gV, c->table_bU); | |
2038 | |
4567 | 2039 /* unscaled special Cases */ |
6503 | 2040 if(unscaled && !usesFilter) |
4554 | 2041 { |
6492 | 2042 /* yv12_to_nv12 */ |
2043 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_NV12) | |
2044 { | |
2045 c->swScale= PlanarToNV12Wrapper; | |
2046 | |
2047 if(flags&SWS_PRINT_INFO) | |
2048 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2049 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2050 return c; | |
2051 } | |
2052 /* yv12_to_yuy2 */ | |
2053 if((srcFormat == IMGFMT_YV12||srcFormat==IMGFMT_I420)&&dstFormat == IMGFMT_YUY2) | |
2054 { | |
6503 | 2055 c->swScale= PlanarToYuy2Wrapper; |
6492 | 2056 |
2057 if(flags&SWS_PRINT_INFO) | |
2058 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2059 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2060 return c; | |
2061 } | |
4554 | 2062 /* 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
|
2063 if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_I420) && isBGR(dstFormat)) |
4554 | 2064 { |
2065 // FIXME multiple yuv2rgb converters wont work that way cuz that thing is full of globals&statics | |
6578 | 2066 //FIXME rgb vs. bgr ? |
4791 | 2067 #ifdef WORDS_BIGENDIAN |
5246 | 2068 if(dstFormat==IMGFMT_BGR32) |
2069 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_BGR); | |
2070 else | |
2071 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); | |
4791 | 2072 #else |
4555 | 2073 yuv2rgb_init( dstFormat&0xFF /* =bpp */, MODE_RGB); |
4791 | 2074 #endif |
4554 | 2075 c->swScale= planarYuvToBgr; |
4555 | 2076 |
2077 if(flags&SWS_PRINT_INFO) | |
6492 | 2078 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4555 | 2079 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2080 return c; | |
2081 } | |
6582
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2082 |
4555 | 2083 /* 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
|
2084 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
|
2085 || (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
|
2086 || (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
|
2087 || (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
|
2088 || (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
|
2089 ) |
4555 | 2090 { |
2091 c->swScale= simpleCopy; | |
2092 | |
4554 | 2093 if(flags&SWS_PRINT_INFO) |
6492 | 2094 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4554 | 2095 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2096 return c; | |
2097 } | |
6582
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2098 |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2099 if( srcFormat==IMGFMT_YVU9 && (dstFormat==IMGFMT_YV12 || dstFormat==IMGFMT_I420) ) |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2100 { |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2101 c->swScale= yvu9toyv12Wrapper; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2102 |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2103 if(flags&SWS_PRINT_INFO) |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2104 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2105 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2106 return c; |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2107 } |
f98313dcd428
yvu9 -> yv12 unscaled converter with linear chroma scaling
michael
parents:
6580
diff
changeset
|
2108 |
4567 | 2109 /* bgr32to24 & rgb32to24*/ |
2110 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR24) | |
2111 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB24)) | |
2112 { | |
2113 c->swScale= bgr32to24Wrapper; | |
2114 | |
2115 if(flags&SWS_PRINT_INFO) | |
6492 | 2116 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2117 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2118 return c; | |
2119 } | |
2120 | |
2121 /* bgr32to16 & rgb32to16*/ | |
2122 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR16) | |
2123 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB16)) | |
2124 { | |
2125 c->swScale= bgr32to16Wrapper; | |
2126 | |
2127 if(flags&SWS_PRINT_INFO) | |
2128 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2129 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2130 return c; | |
2131 } | |
2132 | |
2133 /* bgr32to15 & rgb32to15*/ | |
2134 if((srcFormat==IMGFMT_BGR32 && dstFormat==IMGFMT_BGR15) | |
2135 ||(srcFormat==IMGFMT_RGB32 && dstFormat==IMGFMT_RGB15)) | |
2136 { | |
2137 c->swScale= bgr32to15Wrapper; | |
2138 | |
2139 if(flags&SWS_PRINT_INFO) | |
2140 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 2141 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2142 return c; | |
2143 } | |
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
|
2144 |
4567 | 2145 /* bgr24to32 & rgb24to32*/ |
2146 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR32) | |
2147 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB32)) | |
2148 { | |
2149 c->swScale= bgr24to32Wrapper; | |
2150 | |
2151 if(flags&SWS_PRINT_INFO) | |
6492 | 2152 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2153 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2154 return c; | |
2155 } | |
2156 | |
2157 /* bgr24to16 & rgb24to16*/ | |
2158 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR16) | |
2159 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB16)) | |
2160 { | |
2161 c->swScale= bgr24to16Wrapper; | |
2162 | |
2163 if(flags&SWS_PRINT_INFO) | |
2164 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2165 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2166 return c; | |
2167 } | |
2168 | |
2169 /* bgr24to15 & rgb24to15*/ | |
2170 if((srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_BGR15) | |
2171 ||(srcFormat==IMGFMT_RGB24 && dstFormat==IMGFMT_RGB15)) | |
2172 { | |
2173 c->swScale= bgr24to15Wrapper; | |
2174 | |
2175 if(flags&SWS_PRINT_INFO) | |
2176 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4567 | 2177 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2178 return c; | |
2179 } | |
4581 | 2180 |
2181 /* bgr15to16 */ | |
2182 if(srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR16) | |
2183 { | |
2184 c->swScale= bgr15to16Wrapper; | |
2185 | |
2186 if(flags&SWS_PRINT_INFO) | |
6492 | 2187 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
2188 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2189 return c; | |
2190 } | |
2191 | |
2192 /* bgr15to24 */ | |
2193 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR24) | |
2194 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB24)) | |
2195 { | |
2196 c->swScale= bgr15to24Wrapper; | |
2197 | |
2198 if(flags&SWS_PRINT_INFO) | |
2199 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2200 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2201 return c; | |
2202 } | |
2203 | |
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
|
2204 #if 0 //segfaults |
6492 | 2205 /* bgr15to32 */ |
2206 if((srcFormat==IMGFMT_BGR15 && dstFormat==IMGFMT_BGR32) | |
2207 ||(srcFormat==IMGFMT_RGB15 && dstFormat==IMGFMT_RGB32)) | |
2208 { | |
2209 c->swScale= bgr15to32Wrapper; | |
2210 | |
2211 if(flags&SWS_PRINT_INFO) | |
2212 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2213 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2214 return c; | |
2215 } | |
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
|
2216 #endif |
6492 | 2217 /* bgr16to24 */ |
2218 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR24) | |
2219 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB24)) | |
2220 { | |
2221 c->swScale= bgr16to24Wrapper; | |
2222 | |
2223 if(flags&SWS_PRINT_INFO) | |
2224 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
2225 vo_format_name(srcFormat), vo_format_name(dstFormat)); | |
2226 return c; | |
2227 } | |
2228 | |
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
|
2229 #if 0 //segfaults |
6492 | 2230 /* bgr16to32 */ |
2231 if((srcFormat==IMGFMT_BGR16 && dstFormat==IMGFMT_BGR32) | |
2232 ||(srcFormat==IMGFMT_RGB16 && dstFormat==IMGFMT_RGB32)) | |
2233 { | |
2234 c->swScale= bgr16to32Wrapper; | |
2235 | |
2236 if(flags&SWS_PRINT_INFO) | |
2237 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", | |
4581 | 2238 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2239 return c; | |
2240 } | |
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
|
2241 #endif |
4623 | 2242 /* bgr24toYV12 */ |
2243 if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12) | |
2244 { | |
2245 c->swScale= bgr24toyv12Wrapper; | |
2246 | |
2247 if(flags&SWS_PRINT_INFO) | |
6492 | 2248 MSG_INFO("SwScaler: using unscaled %s -> %s special converter\n", |
4623 | 2249 vo_format_name(srcFormat), vo_format_name(dstFormat)); |
2250 return c; | |
2251 } | |
4554 | 2252 } |
2253 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2254 if(cpuCaps.hasMMX2) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2255 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2256 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
|
2257 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
|
2258 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2259 if(flags&SWS_PRINT_INFO) |
6492 | 2260 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
|
2261 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2262 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2263 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2264 c->canMMX2BeUsed=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2265 |
4467 | 2266 c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW; |
2267 c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH; | |
2268 | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2269 // 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
|
2270 // 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
|
2271 // n-2 is the last chrominance sample available |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2272 // 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
|
2273 // 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
|
2274 // first and last pixel |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2275 if(flags&SWS_FAST_BILINEAR) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2276 { |
4467 | 2277 if(c->canMMX2BeUsed) |
2278 { | |
2279 c->lumXInc+= 20; | |
2280 c->chrXInc+= 20; | |
2281 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2282 //we dont use the x86asm scaler if mmx is available |
4467 | 2283 else if(cpuCaps.hasMMX) |
2284 { | |
2285 c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20; | |
2286 c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20; | |
2287 } | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2288 } |
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 /* precalculate horizontal scaler filter coefficients */ |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2291 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2292 const int filterAlign= cpuCaps.hasMMX ? 4 : 1; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2293 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2294 initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc, |
6580 | 2295 srcW , dstW, filterAlign, 1<<14, |
2296 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2297 srcFilter->lumH, dstFilter->lumH); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2298 initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc, |
6580 | 2299 c->chrSrcW, c->chrDstW, filterAlign, 1<<14, |
2300 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2301 srcFilter->chrH, dstFilter->chrH); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2302 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2303 #ifdef ARCH_X86 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2304 // cant downscale !!! |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2305 if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR)) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2306 { |
5452 | 2307 c->lumMmx2Filter = (int16_t*)memalign(8, (dstW /8+8)*sizeof(int16_t)); |
2308 c->chrMmx2Filter = (int16_t*)memalign(8, (c->chrDstW /4+8)*sizeof(int16_t)); | |
2309 c->lumMmx2FilterPos= (int32_t*)memalign(8, (dstW /2/8+8)*sizeof(int32_t)); | |
2310 c->chrMmx2FilterPos= (int32_t*)memalign(8, (c->chrDstW/2/4+8)*sizeof(int32_t)); | |
2311 | |
2312 initMMX2HScaler( dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8); | |
2313 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
|
2314 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2315 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2316 } // Init Horizontal stuff |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2317 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2318 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2319 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2320 /* precalculate vertical scaler filter coefficients */ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2321 initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc, |
6580 | 2322 srcH , dstH, 1, (1<<12)-4, |
2323 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC) : flags, | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2324 srcFilter->lumV, dstFilter->lumV); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2325 initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc, |
6580 | 2326 c->chrSrcH, c->chrDstH, 1, (1<<12)-4, |
2327 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags, | |
2328 srcFilter->chrV, dstFilter->chrV); | |
4276
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 // 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
|
2331 c->vLumBufSize= c->vLumFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2332 c->vChrBufSize= c->vChrFilterSize; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2333 for(i=0; i<dstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2334 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2335 int chrI= i*c->chrDstH / dstH; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2336 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
|
2337 ((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
|
2338 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
|
2339 if(c->vLumFilterPos[i ] + c->vLumBufSize < nextSlice) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2340 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
|
2341 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
|
2342 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
|
2343 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2344 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2345 // 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
|
2346 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
|
2347 c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*)); |
4419 | 2348 //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
|
2349 for(i=0; i<c->vLumBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2350 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
|
2351 for(i=0; i<c->vChrBufSize; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2352 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
|
2353 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2354 //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
|
2355 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
|
2356 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
|
2357 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2358 ASSERT(c->chrDstH <= dstH) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2359 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2360 // pack filter data for mmx code |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2361 if(cpuCaps.hasMMX) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2362 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2363 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
|
2364 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
|
2365 for(i=0; i<c->vLumFilterSize*dstH; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2366 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
|
2367 c->vLumFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2368 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
|
2369 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
|
2370 c->vChrFilter[i]; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2371 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2372 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2373 if(flags&SWS_PRINT_INFO) |
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 #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
|
2376 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
|
2377 #else |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2378 char *dither= ""; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2379 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2380 if(flags&SWS_FAST_BILINEAR) |
6492 | 2381 MSG_INFO("\nSwScaler: FAST_BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2382 else if(flags&SWS_BILINEAR) |
6492 | 2383 MSG_INFO("\nSwScaler: BILINEAR scaler, "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2384 else if(flags&SWS_BICUBIC) |
6492 | 2385 MSG_INFO("\nSwScaler: BICUBIC scaler, "); |
4467 | 2386 else if(flags&SWS_X) |
6492 | 2387 MSG_INFO("\nSwScaler: Experimental scaler, "); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2388 else if(flags&SWS_POINT) |
6492 | 2389 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
|
2390 else if(flags&SWS_AREA) |
6492 | 2391 MSG_INFO("\nSwScaler: Area Averageing scaler, "); |
6580 | 2392 else if(flags&SWS_BICUBLIN) |
2393 MSG_INFO("\nSwScaler: luma BICUBIC / chroma BILINEAR, "); | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2394 else |
6492 | 2395 MSG_INFO("\nSwScaler: ehh flags invalid?! "); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2396 |
6492 | 2397 if(dstFormat==IMGFMT_BGR15 || dstFormat==IMGFMT_BGR16) |
2398 MSG_INFO("from %s to%s %s ", | |
2399 vo_format_name(srcFormat), dither, vo_format_name(dstFormat)); | |
2400 else | |
2401 MSG_INFO("from %s to %s ", | |
2402 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
|
2403 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2404 if(cpuCaps.hasMMX2) |
6492 | 2405 MSG_INFO("using MMX2\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2406 else if(cpuCaps.has3DNow) |
6492 | 2407 MSG_INFO("using 3DNOW\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2408 else if(cpuCaps.hasMMX) |
6492 | 2409 MSG_INFO("using MMX\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2410 else |
6492 | 2411 MSG_INFO("using C\n"); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2412 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2413 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2414 if((flags & SWS_PRINT_INFO) && verbose) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2415 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2416 if(cpuCaps.hasMMX) |
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 if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR)) |
6492 | 2419 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
|
2420 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2421 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2422 if(c->hLumFilterSize==4) |
6492 | 2423 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
|
2424 else if(c->hLumFilterSize==8) |
6492 | 2425 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
|
2426 else |
6492 | 2427 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
|
2428 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2429 if(c->hChrFilterSize==4) |
6492 | 2430 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
|
2431 else if(c->hChrFilterSize==8) |
6492 | 2432 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
|
2433 else |
6492 | 2434 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
|
2435 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2436 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2437 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2438 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2439 #ifdef ARCH_X86 |
6492 | 2440 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
|
2441 #else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2442 if(flags & SWS_FAST_BILINEAR) |
6492 | 2443 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
|
2444 else |
6492 | 2445 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
|
2446 #endif |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2447 } |
4419 | 2448 if(isPlanarYUV(dstFormat)) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2449 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2450 if(c->vLumFilterSize==1) |
6492 | 2451 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
|
2452 else |
6492 | 2453 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
|
2454 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2455 else |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2456 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2457 if(c->vLumFilterSize==1 && c->vChrFilterSize==2) |
6492 | 2458 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
|
2459 "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
|
2460 else if(c->vLumFilterSize==2 && c->vChrFilterSize==2) |
6492 | 2461 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
|
2462 else |
6492 | 2463 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
|
2464 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2465 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2466 if(dstFormat==IMGFMT_BGR24) |
6492 | 2467 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
|
2468 cpuCaps.hasMMX2 ? "MMX2" : (cpuCaps.hasMMX ? "MMX" : "C")); |
4304 | 2469 else if(dstFormat==IMGFMT_BGR32) |
6492 | 2470 MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2471 else if(dstFormat==IMGFMT_BGR16) |
6492 | 2472 MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", cpuCaps.hasMMX ? "MMX" : "C"); |
4304 | 2473 else if(dstFormat==IMGFMT_BGR15) |
6492 | 2474 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
|
2475 |
6492 | 2476 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
|
2477 } |
4467 | 2478 if((flags & SWS_PRINT_INFO) && verbose>1) |
2479 { | |
6492 | 2480 MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2481 c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc); |
6492 | 2482 MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n", |
4467 | 2483 c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc); |
2484 } | |
4554 | 2485 |
2486 c->swScale= swScale; | |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2487 return c; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2488 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2489 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2490 /** |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2491 * 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
|
2492 * 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
|
2493 */ |
4290
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 SwsVector *getGaussianVec(double variance, double quality){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2496 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
|
2497 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2498 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2499 double middle= (length-1)*0.5; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2500 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2501 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2502 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2503 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2504 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2505 for(i=0; i<length; i++) |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2506 { |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2507 double dist= i-middle; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2508 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
|
2509 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2510 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2511 normalizeVec(vec, 1.0); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2512 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2513 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2514 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2515 |
4297
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2516 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
|
2517 int i; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2518 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
|
2519 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
|
2520 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2521 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
|
2522 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
|
2523 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2524 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
|
2525 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
|
2526 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2527 return vec; |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2528 } |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2529 |
29fef3982238
15/16 bit dithering in C (5% slower, can be disabled by comenting #define DITHER1XBPP out)
michael
parents:
4295
diff
changeset
|
2530 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2531 SwsVector *getIdentityVec(void){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2532 double *coeff= memalign(sizeof(double), sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2533 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2534 coeff[0]= 1.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2535 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2536 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2537 vec->length= 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2538 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2539 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2540 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2541 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2542 void normalizeVec(SwsVector *a, double height){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2543 int i; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2544 double sum=0; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2545 double inv; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2546 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2547 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2548 sum+= a->coeff[i]; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2549 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2550 inv= height/sum; |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2551 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2552 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2553 a->coeff[i]*= height; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2554 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2555 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2556 void scaleVec(SwsVector *a, double scalar){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2557 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2558 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2559 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2560 a->coeff[i]*= scalar; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2561 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2562 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2563 static SwsVector *getConvVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2564 int length= a->length + b->length - 1; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2565 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2566 int i, j; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2567 SwsVector *vec= malloc(sizeof(SwsVector)); |
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 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2570 vec->length= length; |
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 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2573 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2574 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2575 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2576 for(j=0; j<b->length; j++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2577 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2578 coeff[i+j]+= a->coeff[i]*b->coeff[j]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2579 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2580 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2581 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2582 return vec; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2583 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2584 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2585 static SwsVector *sumVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2586 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2587 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2588 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2589 SwsVector *vec= malloc(sizeof(SwsVector)); |
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 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2592 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2593 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2594 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2595 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2596 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
|
2597 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
|
2598 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2599 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2600 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2601 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2602 static SwsVector *diffVec(SwsVector *a, SwsVector *b){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2603 int length= MAX(a->length, b->length); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2604 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2605 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2606 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2607 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2608 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2609 vec->length= length; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2610 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2611 for(i=0; i<length; i++) coeff[i]= 0.0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2612 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2613 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
|
2614 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
|
2615 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2616 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2617 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2618 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2619 /* 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
|
2620 static SwsVector *getShiftedVec(SwsVector *a, int shift){ |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2621 int length= a->length + ABS(shift)*2; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2622 double *coeff= memalign(sizeof(double), length*sizeof(double)); |
4401
8d00348d0d6b
nearest neighbor / sdl emulation ;) scaling (-sws 4)
michael
parents:
4304
diff
changeset
|
2623 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2624 SwsVector *vec= malloc(sizeof(SwsVector)); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2625 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2626 vec->coeff= coeff; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2627 vec->length= length; |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2628 |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2629 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
|
2630 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2631 for(i=0; i<a->length; i++) |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2632 { |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2633 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
|
2634 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2635 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2636 return vec; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2637 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2638 |
4294
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2639 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
|
2640 SwsVector *shifted= getShiftedVec(a, shift); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2641 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2642 a->coeff= shifted->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2643 a->length= shifted->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2644 free(shifted); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2645 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2646 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2647 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
|
2648 SwsVector *sum= sumVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2649 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2650 a->coeff= sum->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2651 a->length= sum->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2652 free(sum); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2653 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2654 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2655 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
|
2656 SwsVector *diff= diffVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2657 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2658 a->coeff= diff->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2659 a->length= diff->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2660 free(diff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2661 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2662 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2663 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
|
2664 SwsVector *conv= getConvVec(a, b); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2665 free(a->coeff); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2666 a->coeff= conv->coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2667 a->length= conv->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2668 free(conv); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2669 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2670 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2671 SwsVector *cloneVec(SwsVector *a){ |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2672 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
|
2673 int i; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2674 SwsVector *vec= malloc(sizeof(SwsVector)); |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2675 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2676 vec->coeff= coeff; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2677 vec->length= a->length; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2678 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2679 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
|
2680 |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2681 return vec; |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2682 } |
21dbbbbd5479
a few filters (should be removed/merged when arpis videofilter stuff is finished)
michael
parents:
4290
diff
changeset
|
2683 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2684 void printVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2685 int i; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2686 double max=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2687 double min=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2688 double range; |
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 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2691 if(a->coeff[i]>max) max= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2692 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2693 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2694 if(a->coeff[i]<min) min= a->coeff[i]; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2695 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2696 range= max - min; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2697 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2698 for(i=0; i<a->length; i++) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2699 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2700 int x= (int)((a->coeff[i]-min)*60.0/range +0.5); |
6492 | 2701 MSG_DBG2("%1.3f ", a->coeff[i]); |
2702 for(;x>0; x--) MSG_DBG2(" "); | |
2703 MSG_DBG2("|\n"); | |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2704 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2705 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2706 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2707 void freeVec(SwsVector *a){ |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2708 if(!a) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2709 if(a->coeff) free(a->coeff); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2710 a->coeff=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2711 a->length=0; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2712 free(a); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2713 } |
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2714 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2715 void freeSwsContext(SwsContext *c){ |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2716 int i; |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2717 if(!c) return; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2718 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2719 if(c->lumPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2720 { |
4419 | 2721 for(i=0; i<c->vLumBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2722 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2723 if(c->lumPixBuf[i]) free(c->lumPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2724 c->lumPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2725 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2726 free(c->lumPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2727 c->lumPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2728 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2729 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2730 if(c->chrPixBuf) |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2731 { |
4419 | 2732 for(i=0; i<c->vChrBufSize; i++) |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2733 { |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2734 if(c->chrPixBuf[i]) free(c->chrPixBuf[i]); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2735 c->chrPixBuf[i]=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2736 } |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2737 free(c->chrPixBuf); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2738 c->chrPixBuf=NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2739 } |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2740 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2741 if(c->vLumFilter) free(c->vLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2742 c->vLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2743 if(c->vChrFilter) free(c->vChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2744 c->vChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2745 if(c->hLumFilter) free(c->hLumFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2746 c->hLumFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2747 if(c->hChrFilter) free(c->hChrFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2748 c->hChrFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2749 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2750 if(c->vLumFilterPos) free(c->vLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2751 c->vLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2752 if(c->vChrFilterPos) free(c->vChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2753 c->vChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2754 if(c->hLumFilterPos) free(c->hLumFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2755 c->hLumFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2756 if(c->hChrFilterPos) free(c->hChrFilterPos); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2757 c->hChrFilterPos = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2758 |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2759 if(c->lumMmxFilter) free(c->lumMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2760 c->lumMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2761 if(c->chrMmxFilter) free(c->chrMmxFilter); |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2762 c->chrMmxFilter = NULL; |
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2763 |
5452 | 2764 if(c->lumMmx2Filter) free(c->lumMmx2Filter); |
2765 c->lumMmx2Filter=NULL; | |
2766 if(c->chrMmx2Filter) free(c->chrMmx2Filter); | |
2767 c->chrMmx2Filter=NULL; | |
2768 if(c->lumMmx2FilterPos) free(c->lumMmx2FilterPos); | |
2769 c->lumMmx2FilterPos=NULL; | |
2770 if(c->chrMmx2FilterPos) free(c->chrMmx2FilterPos); | |
2771 c->chrMmx2FilterPos=NULL; | |
6578 | 2772 if(c->yuvTable) free(c->yuvTable); |
2773 c->yuvTable=NULL; | |
5452 | 2774 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2775 free(c); |
4276
9199d15cb4e0
removed global vars so that multiple swscalers can be used
michael
parents:
4248
diff
changeset
|
2776 } |
4281 | 2777 |
4290
1f8ceb12284d
general convolution filtering of the source picture
michael
parents:
4281
diff
changeset
|
2778 |