Algorithm 版 (精华区)
发信人: lofe ()感激生活(), 信区: Algorithm
标 题: Comp -- model-1a.c
发信站: 哈工大紫丁香 (Sun Sep 3 16:02:19 2000), 转信
/*
* Listing 16 -- model-1a.c
*
* This module is an order-1 fixed-context modeling unit that can
* be using in conjunction with comp-1.c and expand-1.c to compress
* and expand files. It is a very simple implementation of an order-1
* model, using the same techniques for storing counts as were used
* in model-1.c. This means that it uses a lot of memory, around
* 140 Kbytes, and that it spends a lot of time updating the table.
* Since it can loop up context tables with a simple index on the
* context character, it is still pretty fast.
*
* Building the compression and expansion programs with this model
* requires moving up to compact model.
*
* Building the compressor:
*
* Turbo C: tcc -w -mc comp-1.c model-1a.c bitio.c coder.c
* QuickC: qcl /AC /W3 comp-1.c model-1a.c bitio.c coder.c
* Zortech: ztc -mc comp-1.c model-1a.c bitio.c coder.c
* *NIX: cc -o comp-1 comp-1.c model-1a.c bitio.c coder.c
*
* Building the decompressor:
*
* Turbo C: tcc -w -mc expand-1.c model-1a.c bitio.c coder.c
* QuickC: qcl /AC /W3 expand-1.c model-1a.c bitio.c coder.c
* Zortech: ztc -mc expand-1.c model-1a.c bitio.c coder.c
* *NIX: cc -o expand-1 expand-1.c model-1a.c bitio.c coder.c
*/
#include <stdio.h>
#include <stdlib.h>
#include "coder.h"
#include "model.h"
/*
* *totals[] is an array of pointers to context tables. The EOF
* character doesn't get a context table, since we stop encoding
* as soon as that character appears. Each context table is
* an array of ints with indices ranging from -1 to 255.
*/
short int *totals[ 256 ];
/*
* context is the last character encoded or decoded. It is
* used to index to the appropriate context table. We start the
* model with an arbitray context of 0;
*/
int context = 0;
/*
* To initialize the model, I create all 256 context tables, and
* set all the counts in the table to 1. By default, the model
* starts up in context 0, as if the last byte in was '\0'. Since
* each context table is supposed to be indexed from -1 to 255,
* I increment the pointer to the table in totals[], so that the
* array can be safely indexed with -1.
*/
void initialize_model()
{
int i;
short int j;
int array_size;
array_size = sizeof( short int * ) * ( 257 + 1 );
for ( i = 0 ; i < 256 ; i++ )
{
totals[ i ] = (short int *) malloc( array_size ) ;
if ( totals[ i ] == NULL )
{
printf( "Error allocating table space!\n" );
exit( 1 );
}
totals[ i ]++;
for ( j = -1 ; j <= 256 ; j++ )
totals[ i ][ j ] = j + 1;
}
}
/*
* When the table is updated, every count above "symbol" needs to
* be incremented, which is somewhat expensive. If the counts
* have become to large, the table needs to be rescaled. While
* rescaling, we have to make sure that none of the counts drop
* below 1. After the update is complete, the context is changed
* to be the symbol that was just updated.
*/
void update_model( int symbol )
{
int i;
for ( i = symbol+1 ; i <= 256; i++ )
totals[ context ][ i ]++;
if ( totals[ context ][ 256 ] == MAXIMUM_SCALE )
{
for ( i = 0 ; i <= 256 ; i++ )
{
totals[ context ][ i ] /= 2;
if ( totals[ context ][ i ] <= totals[ context ][ i-1 ])
totals[ context ][ i ] = totals[ context ][ i-1 ] + 1;
}
}
context = symbol;
}
/*
* Since the context table can be directly indexed with the
* symbol, getting the low and high counts for the particular
* symbol is nice and easy.
*/
int convert_int_to_symbol( int c, SYMBOL *s )
{
s->scale = totals[ context ][ 256 ];
s->low_count = totals[ context ][ c ];
s->high_count = totals[ context ][ c + 1 ];
return( 0 );
}
/*
* The symbols scale is always in the same place, which is nice.
*/
void get_symbol_scale( SYMBOL *s )
{
s->scale = totals[ context ][ 256 ];
}
/*
* To find the symbol whose low and high values straddle count
* requires walking through the table until a match is found.
* This is a lengthy operation, and helps to keep decoding
* slower than encoding.
*/
int convert_symbol_to_int( int count, SYMBOL *s )
{
int c;
for ( c = 256; count < totals[ context ][ c ] ; c-- )
;
s->high_count = totals[ context ][ c + 1 ];
s->low_count = totals[ context ][ c ];
return( c );
}
--
_____________________________
Every problem has a solution.
optooff@mail.hust.edu.cn
www.netease.com/~hansen
※ 修改:.haojs 于 Sep 3 15:59:42 修改本文.[FROM: bbs.hit.edu.cn]
--
※ 转寄:.武汉白云黄鹤站 bbs.whnet.edu.cn.[FROM: bbs.hit.edu.cn]
--
☆ 来源:.哈工大紫丁香 bbs.hit.edu.cn.[FROM: haojs.bbs@bbs.whnet.]
※ 修改:·lofe 於 09月03日16:06:45 修改本文·[FROM: 202.118.226.15]
Powered by KBS BBS 2.0 (http://dev.kcn.cn)
页面执行时间:2.453毫秒