small changes for mplayer integration:
[libav.git] / postproc / postprocess.c
CommitLineData
3057fa66
A
1/*
2 Copyright (C) 2001 Michael Niedermayer (michaelni@gmx.at)
3
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.
8
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*/
18
19/*
13e00528 20 C MMX MMX2 3DNow*
3057fa66
A
21isVertDC Ec Ec
22isVertMinMaxOk Ec Ec
13e00528 23doVertLowPass E e e*
3057fa66
A
24doVertDefFilter Ec Ec Ec
25isHorizDC Ec Ec
26isHorizMinMaxOk a
13e00528
A
27doHorizLowPass E a a*
28doHorizDefFilter E ac ac
3057fa66 29deRing
9f45d04d
MN
30Vertical RKAlgo1 E a a*
31Vertical X1 a E E*
32Horizontal X1 a E E*
d5a1a995 33
3057fa66 34
13e00528 35* i dont have a 3dnow CPU -> its untested
3057fa66
A
36E = Exact implementation
37e = allmost exact implementation
38a = alternative / approximate impl
39c = checked against the other implementations (-vo md5)
40*/
41
42/*
43TODO:
9f45d04d 44verify that everything workes as it should (how?)
3057fa66
A
45reduce the time wasted on the mem transfer
46implement dering
13e00528 47implement everything in C at least (done at the moment but ...)
3057fa66
A
48unroll stuff if instructions depend too much on the prior one
49we use 8x8 blocks for the horizontal filters, opendivx seems to use 8x4?
50move YScale thing to the end instead of fixing QP
13e00528 51write a faster and higher quality deblocking filter :)
d5a1a995
MN
52do something about the speed of the horizontal filters
53make the mainloop more flexible (variable number of blocks at once
54 (the if/else stuff per block is slowing things down)
9f45d04d
MN
55compare the quality & speed of all filters
56implement a few simple deinterlacing filters
57split this huge file
3057fa66
A
58...
59
60Notes:
61
62*/
63
13e00528 64/*
9f45d04d 65Changelog: use the CVS log
d5a1a995
MN
660.1.3
67 bugfixes: last 3 lines not brightness/contrast corrected
68 brightness statistics messed up with initial black pic
69 changed initial values of the brightness statistics
70 C++ -> C conversation
71 QP range question solved (very likely 1<=QP<=32 according to arpi)
72 new experimental vertical deblocking filter
73 RK filter has 3dNow support now (untested)
13e00528
A
740.1.2
75 fixed a bug in the horizontal default filter
76 3dnow version of the Horizontal & Vertical Lowpass filters
77 mmx version of the Horizontal Default filter
78 mmx2 & C versions of a simple filter described in a paper from ramkishor & karandikar
79 added mode flags & quality2mode function
800.1.1
81*/
82
3057fa66
A
83
84#include <inttypes.h>
85#include <stdio.h>
d5a1a995 86#include <stdlib.h>
3057fa66 87#include "../config.h"
3057fa66 88//#undef HAVE_MMX2
13e00528 89//#define HAVE_3DNOW
3057fa66 90//#undef HAVE_MMX
13e00528 91#include "postprocess.h"
3057fa66
A
92
93
94static uint64_t packedYOffset= 0x0000000000000000LL;
95static uint64_t packedYScale= 0x0100010001000100LL;
96static uint64_t w05= 0x0005000500050005LL;
97static uint64_t w20= 0x0020002000200020LL;
98static uint64_t w1400= 0x1400140014001400LL;
99static uint64_t bm00000001= 0x00000000000000FFLL;
100static uint64_t bm00010000= 0x000000FF00000000LL;
101static uint64_t bm00001000= 0x00000000FF000000LL;
102static uint64_t bm10000000= 0xFF00000000000000LL;
103static uint64_t bm10000001= 0xFF000000000000FFLL;
104static uint64_t bm11000011= 0xFFFF00000000FFFFLL;
13e00528 105static uint64_t bm00000011= 0x000000000000FFFFLL;
9f45d04d 106static uint64_t bm11111110= 0xFFFFFFFFFFFFFF00LL;
13e00528 107static uint64_t bm11000000= 0xFFFF000000000000LL;
3057fa66
A
108static uint64_t bm00011000= 0x000000FFFF000000LL;
109static uint64_t bm00110011= 0x0000FFFF0000FFFFLL;
110static uint64_t bm11001100= 0xFFFF0000FFFF0000LL;
111static uint64_t b00= 0x0000000000000000LL;
9f45d04d 112static uint64_t b01= 0x0101010101010101LL;
3057fa66
A
113static uint64_t b02= 0x0202020202020202LL;
114static uint64_t b0F= 0x0F0F0F0F0F0F0F0FLL;
115static uint64_t bFF= 0xFFFFFFFFFFFFFFFFLL;
13e00528
A
116static uint64_t b20= 0x2020202020202020LL;
117static uint64_t b80= 0x8080808080808080LL;
3057fa66
A
118static uint64_t b7E= 0x7E7E7E7E7E7E7E7ELL;
119static uint64_t b7C= 0x7C7C7C7C7C7C7C7CLL;
120static uint64_t b3F= 0x3F3F3F3F3F3F3F3FLL;
121static uint64_t temp0=0;
122static uint64_t temp1=0;
123static uint64_t temp2=0;
124static uint64_t temp3=0;
125static uint64_t temp4=0;
126static uint64_t temp5=0;
127static uint64_t pQPb=0;
128static uint8_t tempBlock[16*16];
129
130int hFlatnessThreshold= 56 - 16;
131int vFlatnessThreshold= 56 - 16;
132
133//amount of "black" u r willing to loose to get a brightness corrected picture
134double maxClippedThreshold= 0.01;
135
136int maxAllowedY=255;
137