From 9444ef01eb53021af38c1b532a09c00cce53b456 Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Fri, 6 Nov 2015 13:52:00 +0100 Subject: Merging upstream version 1.7~pre1. Signed-off-by: Daniel Baumann --- encoder_base.h | 476 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 476 insertions(+) create mode 100644 encoder_base.h (limited to 'encoder_base.h') diff --git a/encoder_base.h b/encoder_base.h new file mode 100644 index 0000000..a72442f --- /dev/null +++ b/encoder_base.h @@ -0,0 +1,476 @@ +/* Clzip - LZMA lossless data compressor + 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 + 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, see . +*/ + +enum { price_shift_bits = 6, + price_step_bits = 2, + price_step = 1 << price_step_bits }; + +typedef uint8_t Dis_slots[1<<10]; + +extern Dis_slots dis_slots; + +static inline void Dis_slots_init( void ) + { + int i, size, slot; + for( slot = 0; slot < 4; ++slot ) dis_slots[slot] = slot; + for( i = 4, size = 2, slot = 4; slot < 20; slot += 2 ) + { + memset( &dis_slots[i], slot, size ); + memset( &dis_slots[i+size], slot + 1, size ); + size <<= 1; + i += size; + } + } + +static inline uint8_t get_slot( const unsigned dis ) + { + if( dis < (1 << 10) ) return dis_slots[dis]; + if( dis < (1 << 19) ) return dis_slots[dis>> 9] + 18; + if( dis < (1 << 28) ) return dis_slots[dis>>18] + 36; + return dis_slots[dis>>27] + 54; + } + + +typedef short Prob_prices[bit_model_total >> price_step_bits]; + +extern Prob_prices prob_prices; + +static inline void Prob_prices_init( void ) + { + int i, j; + for( i = 0; i < bit_model_total >> price_step_bits; ++i ) + { + unsigned val = ( i * price_step ) + ( price_step / 2 ); + int bits = 0; /* base 2 logarithm of val */ + for( j = 0; j < price_shift_bits; ++j ) + { + val = val * val; + bits <<= 1; + while( val >= 1 << 16 ) { val >>= 1; ++bits; } + } + bits += 15; /* remaining bits in val */ + prob_prices[i] = ( bit_model_total_bits << price_shift_bits ) - bits; + } + } + +static inline int get_price( const int probability ) + { return prob_prices[probability >> price_step_bits]; } + + +static inline int price0( const Bit_model probability ) + { return get_price( probability ); } + +static inline int price1( const Bit_model probability ) + { return get_price( bit_model_total - probability ); } + +static inline int price_bit( const Bit_model bm, const int bit ) + { if( bit ) return price1( bm ); else return price0( bm ); } + + +static inline int price_symbol( const Bit_model bm[], int symbol, + const int num_bits ) + { + int price = 0; + symbol |= ( 1 << num_bits ); + while( symbol > 1 ) + { + const int bit = symbol & 1; + symbol >>= 1; + price += price_bit( bm[symbol], bit ); + } + return price; + } + + +static inline int price_symbol_reversed( const Bit_model bm[], int symbol, + const int num_bits ) + { + int price = 0; + int model = 1; + int i; + for( i = num_bits; i > 0; --i ) + { + const int bit = symbol & 1; + price += price_bit( bm[model], bit ); + model = ( model << 1 ) | bit; + symbol >>= 1; + } + return price; + } + + +static inline int price_matched( const Bit_model bm[], int symbol, + int match_byte ) + { + int price = 0; + int mask = 0x100; + symbol |= mask; + + do { + int match_bit, bit; + match_byte <<= 1; + match_bit = match_byte & mask; + symbol <<= 1; + bit = symbol & 0x100; + price += price_bit( bm[match_bit+(symbol>>9)+mask], bit ); + mask &= ~(match_byte ^ symbol); /* if( match_bit != bit ) mask = 0; */ + } + while( symbol < 0x10000 ); + return price; + } + + +struct Matchfinder_base + { + unsigned long long partial_data_pos; + uint8_t * buffer; /* input buffer */ + int32_t * prev_positions; /* 1 + last seen position of key. else 0 */ + int32_t * pos_array; /* may be tree or chain */ + int before_size; /* bytes to keep in buffer before dictionary */ + int buffer_size; + int dictionary_size; /* bytes to keep in buffer before pos */ + int pos; /* current pos in buffer */ + int cyclic_pos; /* cycles through [0, dictionary_size] */ + int stream_pos; /* first byte not yet read from file */ + int pos_limit; /* when reached, a new block must be read */ + int key4_mask; + int num_prev_positions; /* size of prev_positions */ + int pos_array_size; + int infd; /* input file descriptor */ + bool at_stream_end; /* stream_pos shows real end of file */ + }; + +bool Mb_read_block( struct Matchfinder_base * const mb ); +void Mb_normalize_pos( struct Matchfinder_base * const mb ); + +bool Mb_init( struct Matchfinder_base * const mb, + const int before, const int dict_size, + const int after_size, const int dict_factor, + const int num_prev_positions23, + const int pos_array_factor, const int ifd ); + +static inline void Mb_free( struct Matchfinder_base * const mb ) + { free( mb->prev_positions ); free( mb->buffer ); } + +static inline uint8_t Mb_peek( const struct Matchfinder_base * const mb, + const int distance ) + { return mb->buffer[mb->pos-distance]; } + +static inline int Mb_available_bytes( const struct Matchfinder_base * const mb ) + { return mb->stream_pos - mb->pos; } + +static inline unsigned long long +Mb_data_position( const struct Matchfinder_base * const mb ) + { return mb->partial_data_pos + mb->pos; } + +static inline bool Mb_data_finished( const struct Matchfinder_base * const mb ) + { return mb->at_stream_end && mb->pos >= mb->stream_pos; } + +static inline const uint8_t * +Mb_ptr_to_current_pos( const struct Matchfinder_base * const mb ) + { return mb->buffer + mb->pos; } + +static inline int Mb_true_match_len( const struct Matchfinder_base * const mb, + const int index, const int distance, + int len_limit ) + { + const uint8_t * const data = mb->buffer + mb->pos + index; + int i = 0; + if( index + len_limit > Mb_available_bytes( mb ) ) + len_limit = Mb_available_bytes( mb ) - index; + while( i < len_limit && data[i-distance] == data[i] ) ++i; + return i; + } + +static inline void Mb_move_pos( struct Matchfinder_base * const mb ) + { + if( ++mb->cyclic_pos > mb->dictionary_size ) mb->cyclic_pos = 0; + if( ++mb->pos >= mb->pos_limit ) Mb_normalize_pos( mb ); + } + +void Mb_reset( struct Matchfinder_base * const mb ); + + +enum { re_buffer_size = 65536 }; + +struct Range_encoder + { + uint64_t low; + unsigned long long partial_member_pos; + uint8_t * buffer; /* output buffer */ + int pos; /* current pos in buffer */ + uint32_t range; + unsigned ff_count; + int outfd; /* output file descriptor */ + uint8_t cache; + File_header header; + }; + +void Re_flush_data( struct Range_encoder * const renc ); + +static inline void Re_put_byte( struct Range_encoder * const renc, + const uint8_t b ) + { + renc->buffer[renc->pos] = b; + if( ++renc->pos >= re_buffer_size ) Re_flush_data( renc ); + } + +static inline void Re_shift_low( struct Range_encoder * const renc ) + { + const bool carry = ( renc->low > 0xFFFFFFFFU ); + if( carry || renc->low < 0xFF000000U ) + { + Re_put_byte( renc, renc->cache + carry ); + for( ; renc->ff_count > 0; --renc->ff_count ) + Re_put_byte( renc, 0xFF + carry ); + renc->cache = renc->low >> 24; + } + else ++renc->ff_count; + renc->low = ( renc->low & 0x00FFFFFFU ) << 8; + } + +static inline void Re_reset( struct Range_encoder * const renc ) + { + int i; + renc->low = 0; + renc->partial_member_pos = 0; + renc->pos = 0; + renc->range = 0xFFFFFFFFU; + renc->ff_count = 0; + renc->cache = 0; + for( i = 0; i < Fh_size; ++i ) + Re_put_byte( renc, renc->header[i] ); + } + +static inline bool Re_init( struct Range_encoder * const renc, + const unsigned dictionary_size, const int ofd ) + { + renc->buffer = (uint8_t *)malloc( re_buffer_size ); + if( !renc->buffer ) return false; + renc->outfd = ofd; + Fh_set_magic( renc->header ); + Fh_set_dictionary_size( renc->header, dictionary_size ); + Re_reset( renc ); + return true; + } + +static inline void Re_free( struct Range_encoder * const renc ) + { free( renc->buffer ); } + +static inline unsigned long long +Re_member_position( const struct Range_encoder * const renc ) + { return renc->partial_member_pos + renc->pos + renc->ff_count; } + +static inline void Re_flush( struct Range_encoder * const renc ) + { int i; for( i = 0; i < 5; ++i ) Re_shift_low( renc ); } + +static inline void Re_encode( struct Range_encoder * const renc, + const int symbol, const int num_bits ) + { + int i; + for( i = num_bits - 1; i >= 0; --i ) + { + renc->range >>= 1; + if( (symbol >> i) & 1 ) renc->low += renc->range; + if( renc->range <= 0x00FFFFFFU ) + { renc->range <<= 8; Re_shift_low( renc ); } + } + } + +static inline void Re_encode_bit( struct Range_encoder * const renc, + Bit_model * const probability, const int bit ) + { + const uint32_t bound = ( renc->range >> bit_model_total_bits ) * *probability; + if( !bit ) + { + renc->range = bound; + *probability += (bit_model_total - *probability) >> bit_model_move_bits; + } + else + { + renc->low += bound; + renc->range -= bound; + *probability -= *probability >> bit_model_move_bits; + } + if( renc->range <= 0x00FFFFFFU ) + { renc->range <<= 8; Re_shift_low( renc ); } + } + +static inline void Re_encode_tree( struct Range_encoder * const renc, + Bit_model bm[], const int symbol, const int num_bits ) + { + int mask = ( 1 << ( num_bits - 1 ) ); + int model = 1; + int i; + for( i = num_bits; i > 0; --i, mask >>= 1 ) + { + const int bit = ( symbol & mask ); + Re_encode_bit( renc, &bm[model], bit ); + model <<= 1; + if( bit ) model |= 1; + } + } + +static inline void Re_encode_tree_reversed( struct Range_encoder * const renc, + Bit_model bm[], int symbol, const int num_bits ) + { + int model = 1; + int i; + for( i = num_bits; i > 0; --i ) + { + const int bit = symbol & 1; + Re_encode_bit( renc, &bm[model], bit ); + model = ( model << 1 ) | bit; + symbol >>= 1; + } + } + +static inline void Re_encode_matched( struct Range_encoder * const renc, + Bit_model bm[], int symbol, + int match_byte ) + { + int mask = 0x100; + symbol |= mask; + + do { + int match_bit, bit; + match_byte <<= 1; + match_bit = match_byte & mask; + symbol <<= 1; + bit = symbol & 0x100; + Re_encode_bit( renc, &bm[match_bit+(symbol>>9)+mask], bit ); + mask &= ~(match_byte ^ symbol); /* if( match_bit != bit ) mask = 0; */ + } + while( symbol < 0x10000 ); + } + +static inline void Re_encode_len( struct Range_encoder * const renc, + struct Len_model * const lm, + int symbol, const int pos_state ) + { + bool bit = ( ( symbol -= min_match_len ) >= len_low_symbols ); + Re_encode_bit( renc, &lm->choice1, bit ); + if( !bit ) + Re_encode_tree( renc, lm->bm_low[pos_state], symbol, len_low_bits ); + else + { + bit = ( symbol >= len_low_symbols + len_mid_symbols ); + Re_encode_bit( renc, &lm->choice2, bit ); + if( !bit ) + Re_encode_tree( renc, lm->bm_mid[pos_state], + symbol - len_low_symbols, len_mid_bits ); + else + Re_encode_tree( renc, lm->bm_high, + symbol - len_low_symbols - len_mid_symbols, len_high_bits ); + } + } + + +enum { max_marker_size = 16, + num_rep_distances = 4 }; /* must be 4 */ + +struct LZ_encoder_base + { + struct Matchfinder_base mb; + uint32_t crc; + + Bit_model bm_literal[1<mb, before, dict_size, after_size, dict_factor, + num_prev_positions23, pos_array_factor, ifd ) ) return false; + if( !Re_init( &eb->renc, eb->mb.dictionary_size, outfd ) ) return false; + LZeb_reset( eb ); + return true; + } + +static inline void LZeb_free( struct LZ_encoder_base * const eb ) + { Re_free( &eb->renc ); Mb_free( &eb->mb ); } + +static inline unsigned LZeb_crc( const struct LZ_encoder_base * const eb ) + { return eb->crc ^ 0xFFFFFFFFU; } + +static inline int LZeb_price_literal( const struct LZ_encoder_base * const eb, + uint8_t prev_byte, uint8_t symbol ) + { return price_symbol( eb->bm_literal[get_lit_state(prev_byte)], symbol, 8 ); } + +static inline int LZeb_price_matched( const struct LZ_encoder_base * const eb, + uint8_t prev_byte, uint8_t symbol, + uint8_t match_byte ) + { return price_matched( eb->bm_literal[get_lit_state(prev_byte)], symbol, + match_byte ); } + +static inline void LZeb_encode_literal( struct LZ_encoder_base * const eb, + uint8_t prev_byte, uint8_t symbol ) + { Re_encode_tree( &eb->renc, + eb->bm_literal[get_lit_state(prev_byte)], symbol, 8 ); } + +static inline void LZeb_encode_matched( struct LZ_encoder_base * const eb, + uint8_t prev_byte, uint8_t symbol, + uint8_t match_byte ) + { Re_encode_matched( &eb->renc, eb->bm_literal[get_lit_state(prev_byte)], + symbol, match_byte ); } + +static inline void LZeb_encode_pair( struct LZ_encoder_base * const eb, + const unsigned dis, const int len, + const int pos_state ) + { + const int dis_slot = get_slot( dis ); + Re_encode_len( &eb->renc, &eb->match_len_model, len, pos_state ); + Re_encode_tree( &eb->renc, eb->bm_dis_slot[get_len_state(len)], dis_slot, + dis_slot_bits ); + + if( dis_slot >= start_dis_model ) + { + const int direct_bits = ( dis_slot >> 1 ) - 1; + const unsigned base = ( 2 | ( dis_slot & 1 ) ) << direct_bits; + const unsigned direct_dis = dis - base; + + if( dis_slot < end_dis_model ) + Re_encode_tree_reversed( &eb->renc, eb->bm_dis + base - dis_slot - 1, + direct_dis, direct_bits ); + else + { + Re_encode( &eb->renc, direct_dis >> dis_align_bits, + direct_bits - dis_align_bits ); + Re_encode_tree_reversed( &eb->renc, eb->bm_align, direct_dis, dis_align_bits ); + } + } + } + +void LZeb_full_flush( struct LZ_encoder_base * const eb, const State state ); -- cgit v1.2.3