1 /////////////////////////////////////////////////////////////////////////////// 2 // 3 /// \file lzma_common.h 4 /// \brief Private definitions common to LZMA encoder and decoder 5 /// 6 // Authors: Igor Pavlov 7 // Lasse Collin 8 // 9 // This file has been put into the public domain. 10 // You can do whatever you want with this file. 11 // 12 /////////////////////////////////////////////////////////////////////////////// 13 14 #ifndef LZMA_LZMA_COMMON_H 15 #define LZMA_LZMA_COMMON_H 16 17 #include "common.h" 18 #include "range_common.h" 19 20 21 /////////////////// 22 // Miscellaneous // 23 /////////////////// 24 25 /// Maximum number of position states. A position state is the lowest pos bits 26 /// number of bits of the current uncompressed offset. In some places there 27 /// are different sets of probabilities for different pos states. 28 #define POS_STATES_MAX (1 << LZMA_PB_MAX) 29 30 31 /// Validates lc, lp, and pb. 32 static inline bool 33 is_lclppb_valid(const lzma_options_lzma *options) 34 { 35 return options->lc <= LZMA_LCLP_MAX && options->lp <= LZMA_LCLP_MAX 36 && options->lc + options->lp <= LZMA_LCLP_MAX 37 && options->pb <= LZMA_PB_MAX; 38 } 39 40 41 /////////// 42 // State // 43 /////////// 44 45 /// This enum is used to track which events have occurred most recently and 46 /// in which order. This information is used to predict the next event. 47 /// 48 /// Events: 49 /// - Literal: One 8-bit byte 50 /// - Match: Repeat a chunk of data at some distance 51 /// - Long repeat: Multi-byte match at a recently seen distance 52 /// - Short repeat: One-byte repeat at a recently seen distance 53 /// 54 /// The event names are in from STATE_oldest_older_previous. REP means 55 /// either short or long repeated match, and NONLIT means any non-literal. 56 typedef enum { 57 STATE_LIT_LIT, 58 STATE_MATCH_LIT_LIT, 59 STATE_REP_LIT_LIT, 60 STATE_SHORTREP_LIT_LIT, 61 STATE_MATCH_LIT, 62 STATE_REP_LIT, 63 STATE_SHORTREP_LIT, 64 STATE_LIT_MATCH, 65 STATE_LIT_LONGREP, 66 STATE_LIT_SHORTREP, 67 STATE_NONLIT_MATCH, 68 STATE_NONLIT_REP, 69 } lzma_lzma_state; 70 71 72 /// Total number of states 73 #define STATES 12 74 75 /// The lowest 7 states indicate that the previous state was a literal. 76 #define LIT_STATES 7 77 78 79 /// Indicate that the latest state was a literal. 80 #define update_literal(state) \ 81 state = ((state) <= STATE_SHORTREP_LIT_LIT \ 82 ? STATE_LIT_LIT \ 83 : ((state) <= STATE_LIT_SHORTREP \ 84 ? (state) - 3 \ 85 : (state) - 6)) 86 87 /// Indicate that the latest state was a match. 88 #define update_match(state) \ 89 state = ((state) < LIT_STATES ? STATE_LIT_MATCH : STATE_NONLIT_MATCH) 90 91 /// Indicate that the latest state was a long repeated match. 92 #define update_long_rep(state) \ 93 state = ((state) < LIT_STATES ? STATE_LIT_LONGREP : STATE_NONLIT_REP) 94 95 /// Indicate that the latest state was a short match. 96 #define update_short_rep(state) \ 97 state = ((state) < LIT_STATES ? STATE_LIT_SHORTREP : STATE_NONLIT_REP) 98 99 /// Test if the previous state was a literal. 100 #define is_literal_state(state) \ 101 ((state) < LIT_STATES) 102 103 104 ///////////// 105 // Literal // 106 ///////////// 107 108 /// Each literal coder is divided in three sections: 109 /// - 0x001-0x0FF: Without match byte 110 /// - 0x101-0x1FF: With match byte; match bit is 0 111 /// - 0x201-0x2FF: With match byte; match bit is 1 112 /// 113 /// Match byte is used when the previous LZMA symbol was something else than 114 /// a literal (that is, it was some kind of match). 115 #define LITERAL_CODER_SIZE 0x300 116 117 /// Maximum number of literal coders 118 #define LITERAL_CODERS_MAX (1 << LZMA_LCLP_MAX) 119 120 /// Locate the literal coder for the next literal byte. The choice depends on 121 /// - the lowest literal_pos_bits bits of the position of the current 122 /// byte; and 123 /// - the highest literal_context_bits bits of the previous byte. 124 #define literal_subcoder(probs, lc, lp_mask, pos, prev_byte) \ 125 ((probs)[(((pos) & lp_mask) << lc) + ((prev_byte) >> (8 - lc))]) 126 127 128 static inline void 129 literal_init(probability (*probs)[LITERAL_CODER_SIZE], 130 uint32_t lc, uint32_t lp) 131 { 132 assert(lc + lp <= LZMA_LCLP_MAX); 133 134 const uint32_t coders = 1U << (lc + lp); 135 136 for (uint32_t i = 0; i < coders; ++i) 137 for (uint32_t j = 0; j < LITERAL_CODER_SIZE; ++j) 138 bit_reset(probs[i][j]); 139 140 return; 141 } 142 143 144 ////////////////// 145 // Match length // 146 ////////////////// 147 148 // Minimum length of a match is two bytes. 149 #define MATCH_LEN_MIN 2 150 151 // Match length is encoded with 4, 5, or 10 bits. 152 // 153 // Length Bits 154 // 2-9 4 = Choice=0 + 3 bits 155 // 10-17 5 = Choice=1 + Choice2=0 + 3 bits 156 // 18-273 10 = Choice=1 + Choice2=1 + 8 bits 157 #define LEN_LOW_BITS 3 158 #define LEN_LOW_SYMBOLS (1 << LEN_LOW_BITS) 159 #define LEN_MID_BITS 3 160 #define LEN_MID_SYMBOLS (1 << LEN_MID_BITS) 161 #define LEN_HIGH_BITS 8 162 #define LEN_HIGH_SYMBOLS (1 << LEN_HIGH_BITS) 163 #define LEN_SYMBOLS (LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS + LEN_HIGH_SYMBOLS) 164 165 // Maximum length of a match is 273 which is a result of the encoding 166 // described above. 167 #define MATCH_LEN_MAX (MATCH_LEN_MIN + LEN_SYMBOLS - 1) 168 169 170 //////////////////// 171 // Match distance // 172 //////////////////// 173 174 // Different set of probabilities is used for match distances that have very 175 // short match length: Lengths of 2, 3, and 4 bytes have a separate set of 176 // probabilities for each length. The matches with longer length use a shared 177 // set of probabilities. 178 #define LEN_TO_POS_STATES 4 179 180 // Macro to get the index of the appropriate probability array. 181 #define get_len_to_pos_state(len) \ 182 ((len) < LEN_TO_POS_STATES + MATCH_LEN_MIN \ 183 ? (len) - MATCH_LEN_MIN \ 184 : LEN_TO_POS_STATES - 1) 185 186 // The highest two bits of a match distance (pos slot) are encoded using six 187 // bits. See fastpos.h for more explanation. 188 #define POS_SLOT_BITS 6 189 #define POS_SLOTS (1 << POS_SLOT_BITS) 190 191 // Match distances up to 127 are fully encoded using probabilities. Since 192 // the highest two bits (pos slot) are always encoded using six bits, the 193 // distances 0-3 don't need any additional bits to encode, since the pos 194 // slot itself is the same as the actual distance. START_POS_MODEL_INDEX 195 // indicates the first pos slot where at least one additional bit is needed. 196 #define START_POS_MODEL_INDEX 4 197 198 // Match distances greater than 127 are encoded in three pieces: 199 // - pos slot: the highest two bits 200 // - direct bits: 2-26 bits below the highest two bits 201 // - alignment bits: four lowest bits 202 // 203 // Direct bits don't use any probabilities. 204 // 205 // The pos slot value of 14 is for distances 128-191 (see the table in 206 // fastpos.h to understand why). 207 #define END_POS_MODEL_INDEX 14 208 209 // Pos slots that indicate a distance <= 127. 210 #define FULL_DISTANCES_BITS (END_POS_MODEL_INDEX / 2) 211 #define FULL_DISTANCES (1 << FULL_DISTANCES_BITS) 212 213 // For match distances greater than 127, only the highest two bits and the 214 // lowest four bits (alignment) is encoded using probabilities. 215 #define ALIGN_BITS 4 216 #define ALIGN_TABLE_SIZE (1 << ALIGN_BITS) 217 #define ALIGN_MASK (ALIGN_TABLE_SIZE - 1) 218 219 // LZMA remembers the four most recent match distances. Reusing these distances 220 // tends to take less space than re-encoding the actual distance value. 221 #define REP_DISTANCES 4 222 223 #endif 224