Merging upstream version 1.7~pre1.
Signed-off-by: Daniel Baumann <daniel@debian.org>
This commit is contained in:
parent
74a7bf2703
commit
7f9346d39e
GPG key ID:
FBB4F0E80A80222F
21 changed files with 1365 additions and 955 deletions
413
encoder.c
413
encoder.c
|
|
@ -1,5 +1,5 @@
|
|||
/* Clzip - LZMA lossless data compressor
|
||||
Copyright (C) 2010-2014 Antonio Diaz Diaz.
|
||||
Copyright (C) 2010-2015 Antonio Diaz Diaz.
|
||||
|
||||
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
|
||||
|
|
@ -24,138 +24,29 @@
|
|||
#include <string.h>
|
||||
|
||||
#include "lzip.h"
|
||||
#include "encoder_base.h"
|
||||
#include "encoder.h"
|
||||
|
||||
|
||||
Dis_slots dis_slots;
|
||||
Prob_prices prob_prices;
|
||||
|
||||
|
||||
bool Mf_read_block( struct Matchfinder * const mf )
|
||||
int LZe_get_match_pairs( struct LZ_encoder * const e, struct Pair * pairs )
|
||||
{
|
||||
if( !mf->at_stream_end && mf->stream_pos < mf->buffer_size )
|
||||
{
|
||||
const int size = mf->buffer_size - mf->stream_pos;
|
||||
const int rd = readblock( mf->infd, mf->buffer + mf->stream_pos, size );
|
||||
mf->stream_pos += rd;
|
||||
if( rd != size && errno )
|
||||
{ show_error( "Read error", errno, false ); cleanup_and_fail( 1 ); }
|
||||
if( rd < size )
|
||||
{ mf->at_stream_end = true; mf->pos_limit = mf->buffer_size; }
|
||||
}
|
||||
return mf->pos < mf->stream_pos;
|
||||
}
|
||||
|
||||
|
||||
void Mf_normalize_pos( struct Matchfinder * const mf )
|
||||
{
|
||||
if( mf->pos > mf->stream_pos )
|
||||
internal_error( "pos > stream_pos in Mf_normalize_pos." );
|
||||
if( !mf->at_stream_end )
|
||||
{
|
||||
int i;
|
||||
const int offset = mf->pos - mf->dictionary_size - before_size;
|
||||
const int size = mf->stream_pos - offset;
|
||||
memmove( mf->buffer, mf->buffer + offset, size );
|
||||
mf->partial_data_pos += offset;
|
||||
mf->pos -= offset;
|
||||
mf->stream_pos -= offset;
|
||||
for( i = 0; i < mf->num_prev_positions; ++i )
|
||||
mf->prev_positions[i] -= min( mf->prev_positions[i], offset );
|
||||
for( i = 0; i < 2 * ( mf->dictionary_size + 1 ); ++i )
|
||||
mf->prev_pos_tree[i] -= min( mf->prev_pos_tree[i], offset );
|
||||
Mf_read_block( mf );
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
bool Mf_init( struct Matchfinder * const mf, const int dict_size,
|
||||
const int match_len_limit, const int ifd )
|
||||
{
|
||||
const int buffer_size_limit = ( 2 * dict_size ) + before_size + after_size;
|
||||
unsigned size;
|
||||
int i;
|
||||
|
||||
mf->partial_data_pos = 0;
|
||||
mf->match_len_limit = match_len_limit;
|
||||
mf->pos = 0;
|
||||
mf->cyclic_pos = 0;
|
||||
mf->stream_pos = 0;
|
||||
mf->cycles = ( match_len_limit < max_match_len ) ?
|
||||
16 + ( match_len_limit / 2 ) : 256;
|
||||
mf->infd = ifd;
|
||||
mf->at_stream_end = false;
|
||||
|
||||
mf->buffer_size = max( 65536, dict_size );
|
||||
mf->buffer = (uint8_t *)malloc( mf->buffer_size );
|
||||
if( !mf->buffer ) return false;
|
||||
if( Mf_read_block( mf ) && !mf->at_stream_end &&
|
||||
mf->buffer_size < buffer_size_limit )
|
||||
{
|
||||
uint8_t * tmp;
|
||||
mf->buffer_size = buffer_size_limit;
|
||||
tmp = (uint8_t *)realloc( mf->buffer, mf->buffer_size );
|
||||
if( !tmp ) { free( mf->buffer ); return false; }
|
||||
mf->buffer = tmp;
|
||||
Mf_read_block( mf );
|
||||
}
|
||||
if( mf->at_stream_end && mf->stream_pos < dict_size )
|
||||
mf->dictionary_size = max( min_dictionary_size, mf->stream_pos );
|
||||
else
|
||||
mf->dictionary_size = dict_size;
|
||||
mf->pos_limit = mf->buffer_size;
|
||||
if( !mf->at_stream_end ) mf->pos_limit -= after_size;
|
||||
size = 1 << max( 16, real_bits( mf->dictionary_size - 1 ) - 2 );
|
||||
if( mf->dictionary_size > 1 << 26 ) /* 64 MiB */
|
||||
size >>= 1;
|
||||
mf->key4_mask = size - 1;
|
||||
size += num_prev_positions2;
|
||||
size += num_prev_positions3;
|
||||
|
||||
mf->num_prev_positions = size;
|
||||
size += ( 2 * ( mf->dictionary_size + 1 ) );
|
||||
if( size * sizeof (int32_t) <= size ) mf->prev_positions = 0;
|
||||
else mf->prev_positions = (int32_t *)malloc( size * sizeof (int32_t) );
|
||||
if( !mf->prev_positions ) { free( mf->buffer ); return false; }
|
||||
mf->prev_pos_tree = mf->prev_positions + mf->num_prev_positions;
|
||||
for( i = 0; i < mf->num_prev_positions; ++i ) mf->prev_positions[i] = 0;
|
||||
return true;
|
||||
}
|
||||
|
||||
|
||||
void Mf_reset( struct Matchfinder * const mf )
|
||||
{
|
||||
int i;
|
||||
if( mf->stream_pos > mf->pos )
|
||||
memmove( mf->buffer, mf->buffer + mf->pos, mf->stream_pos - mf->pos );
|
||||
mf->partial_data_pos = 0;
|
||||
mf->stream_pos -= mf->pos;
|
||||
mf->pos = 0;
|
||||
mf->cyclic_pos = 0;
|
||||
for( i = 0; i < mf->num_prev_positions; ++i ) mf->prev_positions[i] = 0;
|
||||
Mf_read_block( mf );
|
||||
}
|
||||
|
||||
|
||||
int Mf_get_match_pairs( struct Matchfinder * const mf, struct Pair * pairs )
|
||||
{
|
||||
int32_t * ptr0 = mf->prev_pos_tree + ( mf->cyclic_pos << 1 );
|
||||
int32_t * ptr0 = e->eb.mb.pos_array + ( e->eb.mb.cyclic_pos << 1 );
|
||||
int32_t * ptr1 = ptr0 + 1;
|
||||
int32_t * newptr;
|
||||
int len = 0, len0 = 0, len1 = 0;
|
||||
int maxlen = 0;
|
||||
int num_pairs = 0;
|
||||
const int pos1 = mf->pos + 1;
|
||||
const int min_pos =
|
||||
( mf->pos > mf->dictionary_size ) ? mf->pos - mf->dictionary_size : 0;
|
||||
const uint8_t * const data = mf->buffer + mf->pos;
|
||||
const int pos1 = e->eb.mb.pos + 1;
|
||||
const int min_pos = ( e->eb.mb.pos > e->eb.mb.dictionary_size ) ?
|
||||
e->eb.mb.pos - e->eb.mb.dictionary_size : 0;
|
||||
const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
|
||||
int count, delta, key2, key3, key4, newpos;
|
||||
unsigned tmp;
|
||||
int len_limit = mf->match_len_limit;
|
||||
int len_limit = e->match_len_limit;
|
||||
|
||||
if( len_limit > Mf_available_bytes( mf ) )
|
||||
if( len_limit > Mb_available_bytes( &e->eb.mb ) )
|
||||
{
|
||||
len_limit = Mf_available_bytes( mf );
|
||||
len_limit = Mb_available_bytes( &e->eb.mb );
|
||||
if( len_limit < 4 ) return 0;
|
||||
}
|
||||
|
||||
|
|
@ -164,23 +55,23 @@ int Mf_get_match_pairs( struct Matchfinder * const mf, struct Pair * pairs )
|
|||
tmp ^= (unsigned)data[2] << 8;
|
||||
key3 = num_prev_positions2 + ( tmp & ( num_prev_positions3 - 1 ) );
|
||||
key4 = num_prev_positions2 + num_prev_positions3 +
|
||||
( ( tmp ^ ( crc32[data[3]] << 5 ) ) & mf->key4_mask );
|
||||
( ( tmp ^ ( crc32[data[3]] << 5 ) ) & e->eb.mb.key4_mask );
|
||||
|
||||
if( pairs )
|
||||
{
|
||||
int np2 = mf->prev_positions[key2];
|
||||
int np3 = mf->prev_positions[key3];
|
||||
if( np2 > min_pos && mf->buffer[np2-1] == data[0] )
|
||||
int np2 = e->eb.mb.prev_positions[key2];
|
||||
int np3 = e->eb.mb.prev_positions[key3];
|
||||
if( np2 > min_pos && e->eb.mb.buffer[np2-1] == data[0] )
|
||||
{
|
||||
pairs[0].dis = mf->pos - np2;
|
||||
pairs[0].dis = e->eb.mb.pos - np2;
|
||||
pairs[0].len = maxlen = 2;
|
||||
num_pairs = 1;
|
||||
}
|
||||
if( np2 != np3 && np3 > min_pos && mf->buffer[np3-1] == data[0] )
|
||||
if( np2 != np3 && np3 > min_pos && e->eb.mb.buffer[np3-1] == data[0] )
|
||||
{
|
||||
maxlen = 3;
|
||||
np2 = np3;
|
||||
pairs[num_pairs].dis = mf->pos - np2;
|
||||
pairs[num_pairs].dis = e->eb.mb.pos - np2;
|
||||
++num_pairs;
|
||||
}
|
||||
if( num_pairs > 0 )
|
||||
|
|
@ -194,19 +85,19 @@ int Mf_get_match_pairs( struct Matchfinder * const mf, struct Pair * pairs )
|
|||
if( maxlen < 3 ) maxlen = 3;
|
||||
}
|
||||
|
||||
mf->prev_positions[key2] = pos1;
|
||||
mf->prev_positions[key3] = pos1;
|
||||
newpos = mf->prev_positions[key4];
|
||||
mf->prev_positions[key4] = pos1;
|
||||
e->eb.mb.prev_positions[key2] = pos1;
|
||||
e->eb.mb.prev_positions[key3] = pos1;
|
||||
newpos = e->eb.mb.prev_positions[key4];
|
||||
e->eb.mb.prev_positions[key4] = pos1;
|
||||
|
||||
for( count = mf->cycles; ; )
|
||||
for( count = e->cycles; ; )
|
||||
{
|
||||
if( newpos <= min_pos || --count < 0 ) { *ptr0 = *ptr1 = 0; break; }
|
||||
|
||||
delta = pos1 - newpos;
|
||||
newptr = mf->prev_pos_tree +
|
||||
( ( mf->cyclic_pos - delta +
|
||||
( (mf->cyclic_pos >= delta) ? 0 : mf->dictionary_size + 1 ) ) << 1 );
|
||||
newptr = e->eb.mb.pos_array +
|
||||
( ( e->eb.mb.cyclic_pos - delta +
|
||||
( (e->eb.mb.cyclic_pos >= delta) ? 0 : e->eb.mb.dictionary_size + 1 ) ) << 1 );
|
||||
if( data[len-delta] == data[len] )
|
||||
{
|
||||
while( ++len < len_limit && data[len-delta] == data[len] ) {}
|
||||
|
|
@ -242,39 +133,6 @@ int Mf_get_match_pairs( struct Matchfinder * const mf, struct Pair * pairs )
|
|||
}
|
||||
|
||||
|
||||
void Re_flush_data( struct Range_encoder * const renc )
|
||||
{
|
||||
if( renc->pos > 0 )
|
||||
{
|
||||
if( renc->outfd >= 0 &&
|
||||
writeblock( renc->outfd, renc->buffer, renc->pos ) != renc->pos )
|
||||
{ show_error( "Write error", errno, false ); cleanup_and_fail( 1 ); }
|
||||
renc->partial_member_pos += renc->pos;
|
||||
renc->pos = 0;
|
||||
if( verbosity >= 2 ) show_progress( 0, 0, 0, 0 );
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
/* End Of Stream mark => (dis == 0xFFFFFFFFU, len == min_match_len) */
|
||||
static void LZe_full_flush( struct LZ_encoder * const e, const State state )
|
||||
{
|
||||
int i;
|
||||
const int pos_state = Mf_data_position( e->matchfinder ) & pos_state_mask;
|
||||
File_trailer trailer;
|
||||
Re_encode_bit( &e->renc, &e->bm_match[state][pos_state], 1 );
|
||||
Re_encode_bit( &e->renc, &e->bm_rep[state], 0 );
|
||||
LZe_encode_pair( e, 0xFFFFFFFFU, min_match_len, pos_state );
|
||||
Re_flush( &e->renc );
|
||||
Ft_set_data_crc( trailer, LZe_crc( e ) );
|
||||
Ft_set_data_size( trailer, Mf_data_position( e->matchfinder ) );
|
||||
Ft_set_member_size( trailer, Re_member_position( &e->renc ) + Ft_size );
|
||||
for( i = 0; i < Ft_size; ++i )
|
||||
Re_put_byte( &e->renc, trailer[i] );
|
||||
Re_flush_data( &e->renc );
|
||||
}
|
||||
|
||||
|
||||
static void LZe_update_distance_prices( struct LZ_encoder * const e )
|
||||
{
|
||||
int dis, len_state;
|
||||
|
|
@ -283,7 +141,7 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
|
|||
const int dis_slot = dis_slots[dis];
|
||||
const int direct_bits = ( dis_slot >> 1 ) - 1;
|
||||
const int base = ( 2 | ( dis_slot & 1 ) ) << direct_bits;
|
||||
const int price = price_symbol_reversed( e->bm_dis + base - dis_slot - 1,
|
||||
const int price = price_symbol_reversed( e->eb.bm_dis + base - dis_slot - 1,
|
||||
dis - base, direct_bits );
|
||||
for( len_state = 0; len_state < len_states; ++len_state )
|
||||
e->dis_prices[len_state][dis] = price;
|
||||
|
|
@ -293,7 +151,7 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
|
|||
{
|
||||
int * const dsp = e->dis_slot_prices[len_state];
|
||||
int * const dp = e->dis_prices[len_state];
|
||||
const Bit_model * const bmds = e->bm_dis_slot[len_state];
|
||||
const Bit_model * const bmds = e->eb.bm_dis_slot[len_state];
|
||||
int slot = 0;
|
||||
for( ; slot < end_dis_model; ++slot )
|
||||
dsp[slot] = price_symbol( bmds, slot, dis_slot_bits );
|
||||
|
|
@ -309,39 +167,7 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
|
|||
}
|
||||
|
||||
|
||||
bool LZe_init( struct LZ_encoder * const e, struct Matchfinder * const mf,
|
||||
const File_header header, const int outfd )
|
||||
{
|
||||
int i;
|
||||
e->pending_num_pairs = 0;
|
||||
e->crc = 0xFFFFFFFFU;
|
||||
|
||||
Bm_array_init( e->bm_literal[0], (1 << literal_context_bits) * 0x300 );
|
||||
Bm_array_init( e->bm_match[0], states * pos_states );
|
||||
Bm_array_init( e->bm_rep, states );
|
||||
Bm_array_init( e->bm_rep0, states );
|
||||
Bm_array_init( e->bm_rep1, states );
|
||||
Bm_array_init( e->bm_rep2, states );
|
||||
Bm_array_init( e->bm_len[0], states * pos_states );
|
||||
Bm_array_init( e->bm_dis_slot[0], len_states * (1 << dis_slot_bits) );
|
||||
Bm_array_init( e->bm_dis, modeled_distances - end_dis_model );
|
||||
Bm_array_init( e->bm_align, dis_align_size );
|
||||
|
||||
e->matchfinder = mf;
|
||||
if( !Re_init( &e->renc, outfd ) ) return false;
|
||||
Lm_init( &e->match_len_model );
|
||||
Lm_init( &e->rep_len_model );
|
||||
Lp_init( &e->match_len_prices, &e->match_len_model, mf->match_len_limit );
|
||||
Lp_init( &e->rep_len_prices, &e->rep_len_model, mf->match_len_limit );
|
||||
e->num_dis_slots = 2 * real_bits( mf->dictionary_size - 1 );
|
||||
|
||||
for( i = 0; i < Fh_size; ++i )
|
||||
Re_put_byte( &e->renc, header[i] );
|
||||
return true;
|
||||
}
|
||||
|
||||
|
||||
/* Return value == number of bytes advanced (ahead).
|
||||
/* Returns the number of bytes advanced (ahead).
|
||||
trials[0]..trials[ahead-1] contain the steps to encode.
|
||||
( trials[0].dis == -1 ) means literal.
|
||||
A match/rep longer or equal than match_len_limit finishes the sequence.
|
||||
|
|
@ -365,45 +191,44 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
|
||||
for( i = 0; i < num_rep_distances; ++i )
|
||||
{
|
||||
replens[i] =
|
||||
Mf_true_match_len( e->matchfinder, 0, reps[i] + 1, max_match_len );
|
||||
replens[i] = Mb_true_match_len( &e->eb.mb, 0, reps[i] + 1, max_match_len );
|
||||
if( replens[i] > replens[rep_index] ) rep_index = i;
|
||||
}
|
||||
if( replens[rep_index] >= e->matchfinder->match_len_limit )
|
||||
if( replens[rep_index] >= e->match_len_limit )
|
||||
{
|
||||
e->trials[0].dis = rep_index;
|
||||
e->trials[0].price = replens[rep_index];
|
||||
LZe_move_pos( e, replens[rep_index] );
|
||||
LZe_move_and_update( e, replens[rep_index] );
|
||||
return replens[rep_index];
|
||||
}
|
||||
|
||||
if( main_len >= e->matchfinder->match_len_limit )
|
||||
if( main_len >= e->match_len_limit )
|
||||
{
|
||||
e->trials[0].dis = e->pairs[num_pairs-1].dis + num_rep_distances;
|
||||
e->trials[0].price = main_len;
|
||||
LZe_move_pos( e, main_len );
|
||||
LZe_move_and_update( e, main_len );
|
||||
return main_len;
|
||||
}
|
||||
|
||||
{
|
||||
const int pos_state = Mf_data_position( e->matchfinder ) & pos_state_mask;
|
||||
const int match_price = price1( e->bm_match[state][pos_state] );
|
||||
const int rep_match_price = match_price + price1( e->bm_rep[state] );
|
||||
const uint8_t prev_byte = Mf_peek( e->matchfinder, 1 );
|
||||
const uint8_t cur_byte = Mf_peek( e->matchfinder, 0 );
|
||||
const uint8_t match_byte = Mf_peek( e->matchfinder, reps[0] + 1 );
|
||||
const int pos_state = Mb_data_position( &e->eb.mb ) & pos_state_mask;
|
||||
const int match_price = price1( e->eb.bm_match[state][pos_state] );
|
||||
const int rep_match_price = match_price + price1( e->eb.bm_rep[state] );
|
||||
const uint8_t prev_byte = Mb_peek( &e->eb.mb, 1 );
|
||||
const uint8_t cur_byte = Mb_peek( &e->eb.mb, 0 );
|
||||
const uint8_t match_byte = Mb_peek( &e->eb.mb, reps[0] + 1 );
|
||||
|
||||
e->trials[0].state = state;
|
||||
e->trials[1].dis = -1; /* literal */
|
||||
e->trials[1].price = price0( e->bm_match[state][pos_state] );
|
||||
e->trials[1].price = price0( e->eb.bm_match[state][pos_state] );
|
||||
if( St_is_char( state ) )
|
||||
e->trials[1].price += LZe_price_literal( e, prev_byte, cur_byte );
|
||||
e->trials[1].price += LZeb_price_literal( &e->eb, prev_byte, cur_byte );
|
||||
else
|
||||
e->trials[1].price += LZe_price_matched( e, prev_byte, cur_byte, match_byte );
|
||||
e->trials[1].price += LZeb_price_matched( &e->eb, prev_byte, cur_byte, match_byte );
|
||||
|
||||
if( match_byte == cur_byte )
|
||||
Tr_update( &e->trials[1], rep_match_price +
|
||||
LZe_price_shortrep( e, state, pos_state ), 0, 0 );
|
||||
LZeb_price_shortrep( &e->eb, state, pos_state ), 0, 0 );
|
||||
|
||||
num_trials = max( main_len, replens[rep_index] );
|
||||
|
||||
|
|
@ -411,7 +236,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
{
|
||||
e->trials[0].dis = e->trials[1].dis;
|
||||
e->trials[0].price = 1;
|
||||
Mf_move_pos( e->matchfinder );
|
||||
Mb_move_pos( &e->eb.mb );
|
||||
return 1;
|
||||
}
|
||||
|
||||
|
|
@ -427,8 +252,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
{
|
||||
int price;
|
||||
if( replens[rep] < min_match_len ) continue;
|
||||
price = rep_match_price + LZe_price_rep( e, rep, state, pos_state );
|
||||
|
||||
price = rep_match_price + LZeb_price_rep( &e->eb, rep, state, pos_state );
|
||||
for( len = min_match_len; len <= replens[rep]; ++len )
|
||||
Tr_update( &e->trials[len], price +
|
||||
Lp_price( &e->rep_len_prices, len, pos_state ), rep, 0 );
|
||||
|
|
@ -436,7 +260,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
|
||||
if( main_len > replens[0] )
|
||||
{
|
||||
const int normal_match_price = match_price + price0( e->bm_rep[state] );
|
||||
const int normal_match_price = match_price + price0( e->eb.bm_rep[state] );
|
||||
i = 0, len = max( replens[0] + 1, min_match_len );
|
||||
while( len > e->pairs[i].len ) ++i;
|
||||
while( true )
|
||||
|
|
@ -453,13 +277,13 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
while( true ) /* price optimization loop */
|
||||
{
|
||||
struct Trial *cur_trial, *next_trial;
|
||||
int newlen, pos_state, available_bytes, len_limit;
|
||||
int newlen, pos_state, triable_bytes, len_limit;
|
||||
int start_len = min_match_len;
|
||||
int next_price, match_price, rep_match_price;
|
||||
State cur_state;
|
||||
uint8_t prev_byte, cur_byte, match_byte;
|
||||
|
||||
Mf_move_pos( e->matchfinder );
|
||||
Mb_move_pos( &e->eb.mb );
|
||||
if( ++cur >= num_trials ) /* no more initialized trials */
|
||||
{
|
||||
LZe_backward( e, cur );
|
||||
|
|
@ -468,7 +292,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
|
||||
num_pairs = LZe_read_match_distances( e );
|
||||
newlen = ( num_pairs > 0 ) ? e->pairs[num_pairs-1].len : 0;
|
||||
if( newlen >= e->matchfinder->match_len_limit )
|
||||
if( newlen >= e->match_len_limit )
|
||||
{
|
||||
e->pending_num_pairs = num_pairs;
|
||||
LZe_backward( e, cur );
|
||||
|
|
@ -515,31 +339,31 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
mtf_reps( dis, cur_trial->reps );
|
||||
}
|
||||
|
||||
pos_state = Mf_data_position( e->matchfinder ) & pos_state_mask;
|
||||
prev_byte = Mf_peek( e->matchfinder, 1 );
|
||||
cur_byte = Mf_peek( e->matchfinder, 0 );
|
||||
match_byte = Mf_peek( e->matchfinder, cur_trial->reps[0] + 1 );
|
||||
pos_state = Mb_data_position( &e->eb.mb ) & pos_state_mask;
|
||||
prev_byte = Mb_peek( &e->eb.mb, 1 );
|
||||
cur_byte = Mb_peek( &e->eb.mb, 0 );
|
||||
match_byte = Mb_peek( &e->eb.mb, cur_trial->reps[0] + 1 );
|
||||
|
||||
next_price = cur_trial->price +
|
||||
price0( e->bm_match[cur_state][pos_state] );
|
||||
price0( e->eb.bm_match[cur_state][pos_state] );
|
||||
if( St_is_char( cur_state ) )
|
||||
next_price += LZe_price_literal( e, prev_byte, cur_byte );
|
||||
next_price += LZeb_price_literal( &e->eb, prev_byte, cur_byte );
|
||||
else
|
||||
next_price += LZe_price_matched( e, prev_byte, cur_byte, match_byte );
|
||||
next_price += LZeb_price_matched( &e->eb, prev_byte, cur_byte, match_byte );
|
||||
|
||||
/* try last updates to next trial */
|
||||
next_trial = &e->trials[cur+1];
|
||||
|
||||
Tr_update( next_trial, next_price, -1, cur ); /* literal */
|
||||
|
||||
match_price = cur_trial->price + price1( e->bm_match[cur_state][pos_state] );
|
||||
rep_match_price = match_price + price1( e->bm_rep[cur_state] );
|
||||
match_price = cur_trial->price + price1( e->eb.bm_match[cur_state][pos_state] );
|
||||
rep_match_price = match_price + price1( e->eb.bm_rep[cur_state] );
|
||||
|
||||
if( match_byte == cur_byte && next_trial->dis != 0 &&
|
||||
next_trial->prev_index2 == single_step_trial )
|
||||
{
|
||||
const int price = rep_match_price +
|
||||
LZe_price_shortrep( e, cur_state, pos_state );
|
||||
LZeb_price_shortrep( &e->eb, cur_state, pos_state );
|
||||
if( price <= next_trial->price )
|
||||
{
|
||||
next_trial->price = price;
|
||||
|
|
@ -548,19 +372,18 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
}
|
||||
}
|
||||
|
||||
available_bytes = min( Mf_available_bytes( e->matchfinder ),
|
||||
max_num_trials - 1 - cur );
|
||||
if( available_bytes < min_match_len ) continue;
|
||||
triable_bytes =
|
||||
min( Mb_available_bytes( &e->eb.mb ), max_num_trials - 1 - cur );
|
||||
if( triable_bytes < min_match_len ) continue;
|
||||
|
||||
len_limit = min( e->matchfinder->match_len_limit, available_bytes );
|
||||
len_limit = min( e->match_len_limit, triable_bytes );
|
||||
|
||||
/* try literal + rep0 */
|
||||
if( match_byte != cur_byte && next_trial->prev_index != cur )
|
||||
{
|
||||
const uint8_t * const data = Mf_ptr_to_current_pos( e->matchfinder );
|
||||
const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
|
||||
const int dis = cur_trial->reps[0] + 1;
|
||||
const int limit = min( e->matchfinder->match_len_limit + 1,
|
||||
available_bytes );
|
||||
const int limit = min( e->match_len_limit + 1, triable_bytes );
|
||||
len = 1;
|
||||
while( len < limit && data[len-dis] == data[len] ) ++len;
|
||||
if( --len >= min_match_len )
|
||||
|
|
@ -568,8 +391,8 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
const int pos_state2 = ( pos_state + 1 ) & pos_state_mask;
|
||||
const State state2 = St_set_char( cur_state );
|
||||
const int price = next_price +
|
||||
price1( e->bm_match[state2][pos_state2] ) +
|
||||
price1( e->bm_rep[state2] ) +
|
||||
price1( e->eb.bm_match[state2][pos_state2] ) +
|
||||
price1( e->eb.bm_rep[state2] ) +
|
||||
LZe_price_rep0_len( e, len, state2, pos_state2 );
|
||||
while( num_trials < cur + 1 + len )
|
||||
e->trials[++num_trials].price = infinite_price;
|
||||
|
|
@ -580,7 +403,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
/* try rep distances */
|
||||
for( rep = 0; rep < num_rep_distances; ++rep )
|
||||
{
|
||||
const uint8_t * const data = Mf_ptr_to_current_pos( e->matchfinder );
|
||||
const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
|
||||
int price;
|
||||
const int dis = cur_trial->reps[rep] + 1;
|
||||
|
||||
|
|
@ -589,7 +412,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
if( data[len-dis] != data[len] ) break;
|
||||
while( num_trials < cur + len )
|
||||
e->trials[++num_trials].price = infinite_price;
|
||||
price = rep_match_price + LZe_price_rep( e, rep, cur_state, pos_state );
|
||||
price = rep_match_price + LZeb_price_rep( &e->eb, rep, cur_state, pos_state );
|
||||
for( i = min_match_len; i <= len; ++i )
|
||||
Tr_update( &e->trials[cur+i], price +
|
||||
Lp_price( &e->rep_len_prices, i, pos_state ), rep, cur );
|
||||
|
|
@ -598,9 +421,9 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
|
||||
/* try rep + literal + rep0 */
|
||||
{
|
||||
int len2 = len + 1, pos_state2;
|
||||
const int limit = min( e->matchfinder->match_len_limit + len2,
|
||||
available_bytes );
|
||||
int len2 = len + 1;
|
||||
const int limit = min( e->match_len_limit + len2, triable_bytes );
|
||||
int pos_state2;
|
||||
State state2;
|
||||
while( len2 < limit && data[len2-dis] == data[len2] ) ++len2;
|
||||
len2 -= len + 1;
|
||||
|
|
@ -609,12 +432,12 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
pos_state2 = ( pos_state + len ) & pos_state_mask;
|
||||
state2 = St_set_rep( cur_state );
|
||||
price += Lp_price( &e->rep_len_prices, len, pos_state ) +
|
||||
price0( e->bm_match[state2][pos_state2] ) +
|
||||
LZe_price_matched( e, data[len-1], data[len], data[len-dis] );
|
||||
price0( e->eb.bm_match[state2][pos_state2] ) +
|
||||
LZeb_price_matched( &e->eb, data[len-1], data[len], data[len-dis] );
|
||||
pos_state2 = ( pos_state2 + 1 ) & pos_state_mask;
|
||||
state2 = St_set_char( state2 );
|
||||
price += price1( e->bm_match[state2][pos_state2] ) +
|
||||
price1( e->bm_rep[state2] ) +
|
||||
price += price1( e->eb.bm_match[state2][pos_state2] ) +
|
||||
price1( e->eb.bm_rep[state2] ) +
|
||||
LZe_price_rep0_len( e, len2, state2, pos_state2 );
|
||||
while( num_trials < cur + len + 1 + len2 )
|
||||
e->trials[++num_trials].price = infinite_price;
|
||||
|
|
@ -627,7 +450,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
{
|
||||
int dis;
|
||||
const int normal_match_price = match_price +
|
||||
price0( e->bm_rep[cur_state] );
|
||||
price0( e->eb.bm_rep[cur_state] );
|
||||
|
||||
while( num_trials < cur + newlen )
|
||||
e->trials[++num_trials].price = infinite_price;
|
||||
|
|
@ -644,23 +467,22 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
|
|||
/* try match + literal + rep0 */
|
||||
if( len == e->pairs[i].len )
|
||||
{
|
||||
const uint8_t * const data = Mf_ptr_to_current_pos( e->matchfinder );
|
||||
const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
|
||||
const int dis2 = dis + 1;
|
||||
int len2 = len + 1;
|
||||
const int limit = min( e->matchfinder->match_len_limit + len2,
|
||||
available_bytes );
|
||||
const int limit = min( e->match_len_limit + len2, triable_bytes );
|
||||
while( len2 < limit && data[len2-dis2] == data[len2] ) ++len2;
|
||||
len2 -= len + 1;
|
||||
if( len2 >= min_match_len )
|
||||
{
|
||||
int pos_state2 = ( pos_state + len ) & pos_state_mask;
|
||||
State state2 = St_set_match( cur_state );
|
||||
price += price0( e->bm_match[state2][pos_state2] ) +
|
||||
LZe_price_matched( e, data[len-1], data[len], data[len-dis2] );
|
||||
price += price0( e->eb.bm_match[state2][pos_state2] ) +
|
||||
LZeb_price_matched( &e->eb, data[len-1], data[len], data[len-dis2] );
|
||||
pos_state2 = ( pos_state2 + 1 ) & pos_state_mask;
|
||||
state2 = St_set_char( state2 );
|
||||
price += price1( e->bm_match[state2][pos_state2] ) +
|
||||
price1( e->bm_rep[state2] ) +
|
||||
price += price1( e->eb.bm_match[state2][pos_state2] ) +
|
||||
price1( e->eb.bm_rep[state2] ) +
|
||||
LZe_price_rep0_len( e, len2, state2, pos_state2 );
|
||||
|
||||
while( num_trials < cur + len + 1 + len2 )
|
||||
|
|
@ -682,10 +504,10 @@ bool LZe_encode_member( struct LZ_encoder * const e,
|
|||
{
|
||||
const unsigned long long member_size_limit =
|
||||
member_size - Ft_size - max_marker_size;
|
||||
const bool best = ( e->matchfinder->match_len_limit > 12 );
|
||||
const bool best = ( e->match_len_limit > 12 );
|
||||
const int dis_price_count = best ? 1 : 512;
|
||||
const int align_price_count = best ? 1 : dis_align_size;
|
||||
const int price_count = ( e->matchfinder->match_len_limit > 36 ) ? 1013 : 4093;
|
||||
const int price_count = ( e->match_len_limit > 36 ) ? 1013 : 4093;
|
||||
int price_counter = 0;
|
||||
int dis_price_counter = 0;
|
||||
int align_price_counter = 0;
|
||||
|
|
@ -694,22 +516,22 @@ bool LZe_encode_member( struct LZ_encoder * const e,
|
|||
State state = 0;
|
||||
for( i = 0; i < num_rep_distances; ++i ) reps[i] = 0;
|
||||
|
||||
if( Mf_data_position( e->matchfinder ) != 0 ||
|
||||
Re_member_position( &e->renc ) != Fh_size )
|
||||
if( Mb_data_position( &e->eb.mb ) != 0 ||
|
||||
Re_member_position( &e->eb.renc ) != Fh_size )
|
||||
return false; /* can be called only once */
|
||||
|
||||
if( !Mf_finished( e->matchfinder ) ) /* encode first byte */
|
||||
if( !Mb_data_finished( &e->eb.mb ) ) /* encode first byte */
|
||||
{
|
||||
const uint8_t prev_byte = 0;
|
||||
const uint8_t cur_byte = Mf_peek( e->matchfinder, 0 );
|
||||
Re_encode_bit( &e->renc, &e->bm_match[state][0], 0 );
|
||||
LZe_encode_literal( e, prev_byte, cur_byte );
|
||||
CRC32_update_byte( &e->crc, cur_byte );
|
||||
Mf_get_match_pairs( e->matchfinder, 0 );
|
||||
Mf_move_pos( e->matchfinder );
|
||||
const uint8_t cur_byte = Mb_peek( &e->eb.mb, 0 );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_match[state][0], 0 );
|
||||
LZeb_encode_literal( &e->eb, prev_byte, cur_byte );
|
||||
CRC32_update_byte( &e->eb.crc, cur_byte );
|
||||
LZe_get_match_pairs( e, 0 );
|
||||
Mb_move_pos( &e->eb.mb );
|
||||
}
|
||||
|
||||
while( !Mf_finished( e->matchfinder ) )
|
||||
while( !Mb_data_finished( &e->eb.mb ) )
|
||||
{
|
||||
if( price_counter <= 0 && e->pending_num_pairs == 0 )
|
||||
{
|
||||
|
|
@ -720,7 +542,7 @@ bool LZe_encode_member( struct LZ_encoder * const e,
|
|||
{
|
||||
align_price_counter = align_price_count;
|
||||
for( i = 0; i < dis_align_size; ++i )
|
||||
e->align_prices[i] = price_symbol_reversed( e->bm_align, i, dis_align_bits );
|
||||
e->align_prices[i] = price_symbol_reversed( e->eb.bm_align, i, dis_align_bits );
|
||||
}
|
||||
Lp_update_prices( &e->match_len_prices );
|
||||
Lp_update_prices( &e->rep_len_prices );
|
||||
|
|
@ -733,56 +555,55 @@ bool LZe_encode_member( struct LZ_encoder * const e,
|
|||
for( i = 0; ahead > 0; )
|
||||
{
|
||||
const int pos_state =
|
||||
( Mf_data_position( e->matchfinder ) - ahead ) & pos_state_mask;
|
||||
( Mb_data_position( &e->eb.mb ) - ahead ) & pos_state_mask;
|
||||
const int dis = e->trials[i].dis;
|
||||
const int len = e->trials[i].price;
|
||||
|
||||
bool bit = ( dis < 0 );
|
||||
Re_encode_bit( &e->renc, &e->bm_match[state][pos_state], !bit );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_match[state][pos_state], !bit );
|
||||
if( bit ) /* literal byte */
|
||||
{
|
||||
const uint8_t prev_byte = Mf_peek( e->matchfinder, ahead + 1 );
|
||||
const uint8_t cur_byte = Mf_peek( e->matchfinder, ahead );
|
||||
CRC32_update_byte( &e->crc, cur_byte );
|
||||
const uint8_t prev_byte = Mb_peek( &e->eb.mb, ahead + 1 );
|
||||
const uint8_t cur_byte = Mb_peek( &e->eb.mb, ahead );
|
||||
CRC32_update_byte( &e->eb.crc, cur_byte );
|
||||
if( St_is_char( state ) )
|
||||
LZe_encode_literal( e, prev_byte, cur_byte );
|
||||
LZeb_encode_literal( &e->eb, prev_byte, cur_byte );
|
||||
else
|
||||
{
|
||||
const uint8_t match_byte =
|
||||
Mf_peek( e->matchfinder, ahead + reps[0] + 1 );
|
||||
LZe_encode_matched( e, prev_byte, cur_byte, match_byte );
|
||||
const uint8_t match_byte = Mb_peek( &e->eb.mb, ahead + reps[0] + 1 );
|
||||
LZeb_encode_matched( &e->eb, prev_byte, cur_byte, match_byte );
|
||||
}
|
||||
state = St_set_char( state );
|
||||
}
|
||||
else /* match or repeated match */
|
||||
{
|
||||
CRC32_update_buf( &e->crc, Mf_ptr_to_current_pos( e->matchfinder ) - ahead, len );
|
||||
CRC32_update_buf( &e->eb.crc, Mb_ptr_to_current_pos( &e->eb.mb ) - ahead, len );
|
||||
mtf_reps( dis, reps );
|
||||
bit = ( dis < num_rep_distances );
|
||||
Re_encode_bit( &e->renc, &e->bm_rep[state], bit );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_rep[state], bit );
|
||||
if( bit ) /* repeated match */
|
||||
{
|
||||
bit = ( dis == 0 );
|
||||
Re_encode_bit( &e->renc, &e->bm_rep0[state], !bit );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_rep0[state], !bit );
|
||||
if( bit )
|
||||
Re_encode_bit( &e->renc, &e->bm_len[state][pos_state], len > 1 );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_len[state][pos_state], len > 1 );
|
||||
else
|
||||
{
|
||||
Re_encode_bit( &e->renc, &e->bm_rep1[state], dis > 1 );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_rep1[state], dis > 1 );
|
||||
if( dis > 1 )
|
||||
Re_encode_bit( &e->renc, &e->bm_rep2[state], dis > 2 );
|
||||
Re_encode_bit( &e->eb.renc, &e->eb.bm_rep2[state], dis > 2 );
|
||||
}
|
||||
if( len == 1 ) state = St_set_short_rep( state );
|
||||
else
|
||||
{
|
||||
Re_encode_len( &e->renc, &e->rep_len_model, len, pos_state );
|
||||
Re_encode_len( &e->eb.renc, &e->eb.rep_len_model, len, pos_state );
|
||||
Lp_decrement_counter( &e->rep_len_prices, pos_state );
|
||||
state = St_set_rep( state );
|
||||
}
|
||||
}
|
||||
else /* match */
|
||||
{
|
||||
LZe_encode_pair( e, dis - num_rep_distances, len, pos_state );
|
||||
LZeb_encode_pair( &e->eb, dis - num_rep_distances, len, pos_state );
|
||||
if( get_slot( dis - num_rep_distances ) >= end_dis_model )
|
||||
--align_price_counter;
|
||||
--dis_price_counter;
|
||||
|
|
@ -791,14 +612,14 @@ bool LZe_encode_member( struct LZ_encoder * const e,
|
|||
}
|
||||
}
|
||||
ahead -= len; i += len;
|
||||
if( Re_member_position( &e->renc ) >= member_size_limit )
|
||||
if( Re_member_position( &e->eb.renc ) >= member_size_limit )
|
||||
{
|
||||
if( !Mf_dec_pos( e->matchfinder, ahead ) ) return false;
|
||||
LZe_full_flush( e, state );
|
||||
if( !Mb_dec_pos( &e->eb.mb, ahead ) ) return false;
|
||||
LZeb_full_flush( &e->eb, state );
|
||||
return true;
|
||||
}
|
||||
}
|
||||
}
|
||||
LZe_full_flush( e, state );
|
||||
LZeb_full_flush( &e->eb, state );
|
||||
return true;
|
||||
}
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue