Altivec Patch (Mark III) by (Dieter Shirley <dieters at schemasoft dot com>)

[libav.git] / postproc / postprocess.c

/*
    Copyright (C) 2001-2002 Michael Niedermayer (michaelni@gmx.at)

    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
*/

/*
			C	MMX	MMX2	3DNow
isVertDC		Ec	Ec
isVertMinMaxOk		Ec	Ec
doVertLowPass		E		e	e
doVertDefFilter		Ec	Ec	e	e
isHorizDC		Ec	Ec
isHorizMinMaxOk		a	E
doHorizLowPass		E		e	e
doHorizDefFilter	Ec	Ec	e	e
deRing			E		e	e*
Vertical RKAlgo1	E		a	a
Horizontal RKAlgo1			a	a
Vertical X1#		a		E	E
Horizontal X1#		a		E	E
LinIpolDeinterlace	e		E	E*
CubicIpolDeinterlace	a		e	e*
LinBlendDeinterlace	e		E	E*
MedianDeinterlace#	 	Ec	Ec
TempDeNoiser#		E		e	e

* i dont have a 3dnow CPU -> its untested, but noone said it doesnt work so it seems to work
# more or less selfinvented filters so the exactness isnt too meaningfull
E = Exact implementation
e = allmost exact implementation (slightly different rounding,...)
a = alternative / approximate impl
c = checked against the other implementations (-vo md5)
*/

/*
TODO:
reduce the time wasted on the mem transfer
unroll stuff if instructions depend too much on the prior one
move YScale thing to the end instead of fixing QP
write a faster and higher quality deblocking filter :)
make the mainloop more flexible (variable number of blocks at once
	(the if/else stuff per block is slowing things down)
compare the quality & speed of all filters
split this huge file
optimize c versions
try to unroll inner for(x=0 ... loop to avoid these damn if(x ... checks
...
*/

//Changelog: use the CVS log

#include "../config.h"
#include <inttypes.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
//#undef HAVE_MMX2
//#define HAVE_3DNOW
//#undef HAVE_MMX
//#undef ARCH_X86
//#define DEBUG_BRIGHTNESS
#include "../libvo/fastmemcpy.h"
#include "postprocess.h"
#include "../mangle.h"

#define MIN(a,b) ((a) > (b) ? (b) : (a))
#define MAX(a,b) ((a) < (b) ? (b) : (a))
#define ABS(a) ((a) > 0 ? (a) : (-(a)))
#define SIGN(a) ((a) > 0 ? 1 : -1)

#define GET_MODE_BUFFER_SIZE 500
#define OPTIONS_ARRAY_SIZE 10
#define BLOCK_SIZE 8
#define TEMP_STRIDE 8
//#define NUM_BLOCKS_AT_ONCE 16 //not used yet

#ifdef ARCH_X86
static uint64_t __attribute__((aligned(8))) w05=		0x0005000500050005LL;
static uint64_t __attribute__((aligned(8))) w20=		0x0020002000200020LL;
static uint64_t __attribute__((aligned(8))) b00= 		0x0000000000000000LL;
static uint64_t __attribute__((aligned(8))) b01= 		0x0101010101010101LL;
static uint64_t __attribute__((aligned(8))) b02= 		0x0202020202020202LL;
static uint64_t __attribute__((aligned(8))) b08= 		0x0808080808080808LL;
static uint64_t __attribute__((aligned(8))) b80= 		0x8080808080808080LL;
#endif

static int verbose= 0;

static const int deringThreshold= 20;

struct PPFilter{
	char *shortName;
	char *longName;
	int chromDefault; 	// is chrominance filtering on by default if this filter is manually activated
	int minLumQuality; 	// minimum quality to turn luminance filtering on
	int minChromQuality;	// minimum quality to turn chrominance filtering on
	int mask; 		// Bitmask to turn this filter on
};

typedef struct PPContext{
	uint8_t *tempBlocks; //used for the horizontal code