Lasse Collin | 24fa040 | 2011-01-12 17:01:22 -0800 | [diff] [blame] | 1 | /* |
| 2 | * LZMA2 definitions |
| 3 | * |
| 4 | * Authors: Lasse Collin <lasse.collin@tukaani.org> |
| 5 | * Igor Pavlov <http://7-zip.org/> |
| 6 | * |
| 7 | * This file has been put into the public domain. |
| 8 | * You can do whatever you want with this file. |
| 9 | */ |
| 10 | |
| 11 | #ifndef XZ_LZMA2_H |
| 12 | #define XZ_LZMA2_H |
| 13 | |
| 14 | /* Range coder constants */ |
| 15 | #define RC_SHIFT_BITS 8 |
| 16 | #define RC_TOP_BITS 24 |
| 17 | #define RC_TOP_VALUE (1 << RC_TOP_BITS) |
| 18 | #define RC_BIT_MODEL_TOTAL_BITS 11 |
| 19 | #define RC_BIT_MODEL_TOTAL (1 << RC_BIT_MODEL_TOTAL_BITS) |
| 20 | #define RC_MOVE_BITS 5 |
| 21 | |
| 22 | /* |
| 23 | * Maximum number of position states. A position state is the lowest pb |
| 24 | * number of bits of the current uncompressed offset. In some places there |
| 25 | * are different sets of probabilities for different position states. |
| 26 | */ |
| 27 | #define POS_STATES_MAX (1 << 4) |
| 28 | |
| 29 | /* |
| 30 | * This enum is used to track which LZMA symbols have occurred most recently |
| 31 | * and in which order. This information is used to predict the next symbol. |
| 32 | * |
| 33 | * Symbols: |
| 34 | * - Literal: One 8-bit byte |
| 35 | * - Match: Repeat a chunk of data at some distance |
| 36 | * - Long repeat: Multi-byte match at a recently seen distance |
| 37 | * - Short repeat: One-byte repeat at a recently seen distance |
| 38 | * |
| 39 | * The symbol names are in from STATE_oldest_older_previous. REP means |
| 40 | * either short or long repeated match, and NONLIT means any non-literal. |
| 41 | */ |
| 42 | enum lzma_state { |
| 43 | STATE_LIT_LIT, |
| 44 | STATE_MATCH_LIT_LIT, |
| 45 | STATE_REP_LIT_LIT, |
| 46 | STATE_SHORTREP_LIT_LIT, |
| 47 | STATE_MATCH_LIT, |
| 48 | STATE_REP_LIT, |
| 49 | STATE_SHORTREP_LIT, |
| 50 | STATE_LIT_MATCH, |
| 51 | STATE_LIT_LONGREP, |
| 52 | STATE_LIT_SHORTREP, |
| 53 | STATE_NONLIT_MATCH, |
| 54 | STATE_NONLIT_REP |
| 55 | }; |
| 56 | |
| 57 | /* Total number of states */ |
| 58 | #define STATES 12 |
| 59 | |
| 60 | /* The lowest 7 states indicate that the previous state was a literal. */ |
| 61 | #define LIT_STATES 7 |
| 62 | |
| 63 | /* Indicate that the latest symbol was a literal. */ |
| 64 | static inline void lzma_state_literal(enum lzma_state *state) |
| 65 | { |
| 66 | if (*state <= STATE_SHORTREP_LIT_LIT) |
| 67 | *state = STATE_LIT_LIT; |
| 68 | else if (*state <= STATE_LIT_SHORTREP) |
| 69 | *state -= 3; |
| 70 | else |
| 71 | *state -= 6; |
| 72 | } |
| 73 | |
| 74 | /* Indicate that the latest symbol was a match. */ |
| 75 | static inline void lzma_state_match(enum lzma_state *state) |
| 76 | { |
| 77 | *state = *state < LIT_STATES ? STATE_LIT_MATCH : STATE_NONLIT_MATCH; |
| 78 | } |
| 79 | |
| 80 | /* Indicate that the latest state was a long repeated match. */ |
| 81 | static inline void lzma_state_long_rep(enum lzma_state *state) |
| 82 | { |
| 83 | *state = *state < LIT_STATES ? STATE_LIT_LONGREP : STATE_NONLIT_REP; |
| 84 | } |
| 85 | |
| 86 | /* Indicate that the latest symbol was a short match. */ |
| 87 | static inline void lzma_state_short_rep(enum lzma_state *state) |
| 88 | { |
| 89 | *state = *state < LIT_STATES ? STATE_LIT_SHORTREP : STATE_NONLIT_REP; |
| 90 | } |
| 91 | |
| 92 | /* Test if the previous symbol was a literal. */ |
| 93 | static inline bool lzma_state_is_literal(enum lzma_state state) |
| 94 | { |
| 95 | return state < LIT_STATES; |
| 96 | } |
| 97 | |
| 98 | /* Each literal coder is divided in three sections: |
| 99 | * - 0x001-0x0FF: Without match byte |
| 100 | * - 0x101-0x1FF: With match byte; match bit is 0 |
| 101 | * - 0x201-0x2FF: With match byte; match bit is 1 |
| 102 | * |
| 103 | * Match byte is used when the previous LZMA symbol was something else than |
| 104 | * a literal (that is, it was some kind of match). |
| 105 | */ |
| 106 | #define LITERAL_CODER_SIZE 0x300 |
| 107 | |
| 108 | /* Maximum number of literal coders */ |
| 109 | #define LITERAL_CODERS_MAX (1 << 4) |
| 110 | |
| 111 | /* Minimum length of a match is two bytes. */ |
| 112 | #define MATCH_LEN_MIN 2 |
| 113 | |
| 114 | /* Match length is encoded with 4, 5, or 10 bits. |
| 115 | * |
| 116 | * Length Bits |
| 117 | * 2-9 4 = Choice=0 + 3 bits |
| 118 | * 10-17 5 = Choice=1 + Choice2=0 + 3 bits |
| 119 | * 18-273 10 = Choice=1 + Choice2=1 + 8 bits |
| 120 | */ |
| 121 | #define LEN_LOW_BITS 3 |
| 122 | #define LEN_LOW_SYMBOLS (1 << LEN_LOW_BITS) |
| 123 | #define LEN_MID_BITS 3 |
| 124 | #define LEN_MID_SYMBOLS (1 << LEN_MID_BITS) |
| 125 | #define LEN_HIGH_BITS 8 |
| 126 | #define LEN_HIGH_SYMBOLS (1 << LEN_HIGH_BITS) |
| 127 | #define LEN_SYMBOLS (LEN_LOW_SYMBOLS + LEN_MID_SYMBOLS + LEN_HIGH_SYMBOLS) |
| 128 | |
| 129 | /* |
| 130 | * Maximum length of a match is 273 which is a result of the encoding |
| 131 | * described above. |
| 132 | */ |
| 133 | #define MATCH_LEN_MAX (MATCH_LEN_MIN + LEN_SYMBOLS - 1) |
| 134 | |
| 135 | /* |
| 136 | * Different sets of probabilities are used for match distances that have |
| 137 | * very short match length: Lengths of 2, 3, and 4 bytes have a separate |
| 138 | * set of probabilities for each length. The matches with longer length |
| 139 | * use a shared set of probabilities. |
| 140 | */ |
| 141 | #define DIST_STATES 4 |
| 142 | |
| 143 | /* |
| 144 | * Get the index of the appropriate probability array for decoding |
| 145 | * the distance slot. |
| 146 | */ |
| 147 | static inline uint32_t lzma_get_dist_state(uint32_t len) |
| 148 | { |
| 149 | return len < DIST_STATES + MATCH_LEN_MIN |
| 150 | ? len - MATCH_LEN_MIN : DIST_STATES - 1; |
| 151 | } |
| 152 | |
| 153 | /* |
| 154 | * The highest two bits of a 32-bit match distance are encoded using six bits. |
| 155 | * This six-bit value is called a distance slot. This way encoding a 32-bit |
| 156 | * value takes 6-36 bits, larger values taking more bits. |
| 157 | */ |
| 158 | #define DIST_SLOT_BITS 6 |
| 159 | #define DIST_SLOTS (1 << DIST_SLOT_BITS) |
| 160 | |
| 161 | /* Match distances up to 127 are fully encoded using probabilities. Since |
| 162 | * the highest two bits (distance slot) are always encoded using six bits, |
| 163 | * the distances 0-3 don't need any additional bits to encode, since the |
| 164 | * distance slot itself is the same as the actual distance. DIST_MODEL_START |
| 165 | * indicates the first distance slot where at least one additional bit is |
| 166 | * needed. |
| 167 | */ |
| 168 | #define DIST_MODEL_START 4 |
| 169 | |
| 170 | /* |
| 171 | * Match distances greater than 127 are encoded in three pieces: |
| 172 | * - distance slot: the highest two bits |
| 173 | * - direct bits: 2-26 bits below the highest two bits |
| 174 | * - alignment bits: four lowest bits |
| 175 | * |
| 176 | * Direct bits don't use any probabilities. |
| 177 | * |
| 178 | * The distance slot value of 14 is for distances 128-191. |
| 179 | */ |
| 180 | #define DIST_MODEL_END 14 |
| 181 | |
| 182 | /* Distance slots that indicate a distance <= 127. */ |
| 183 | #define FULL_DISTANCES_BITS (DIST_MODEL_END / 2) |
| 184 | #define FULL_DISTANCES (1 << FULL_DISTANCES_BITS) |
| 185 | |
| 186 | /* |
| 187 | * For match distances greater than 127, only the highest two bits and the |
| 188 | * lowest four bits (alignment) is encoded using probabilities. |
| 189 | */ |
| 190 | #define ALIGN_BITS 4 |
| 191 | #define ALIGN_SIZE (1 << ALIGN_BITS) |
| 192 | #define ALIGN_MASK (ALIGN_SIZE - 1) |
| 193 | |
| 194 | /* Total number of all probability variables */ |
| 195 | #define PROBS_TOTAL (1846 + LITERAL_CODERS_MAX * LITERAL_CODER_SIZE) |
| 196 | |
| 197 | /* |
| 198 | * LZMA remembers the four most recent match distances. Reusing these |
| 199 | * distances tends to take less space than re-encoding the actual |
| 200 | * distance value. |
| 201 | */ |
| 202 | #define REPS 4 |
| 203 | |
| 204 | #endif |