| 1 | // SPDX-License-Identifier: GPL-2.0+ OR BSD-3-Clause |
| 2 | /* |
| 3 | * Copyright (c) Meta Platforms, Inc. and affiliates. |
| 4 | * All rights reserved. |
| 5 | * |
| 6 | * This source code is licensed under both the BSD-style license (found in the |
| 7 | * LICENSE file in the root directory of this source tree) and the GPLv2 (found |
| 8 | * in the COPYING file in the root directory of this source tree). |
| 9 | * You may select, at your option, one of the above-listed licenses. |
| 10 | */ |
| 11 | |
| 12 | /*-************************************* |
| 13 | * Dependencies |
| 14 | ***************************************/ |
| 15 | #include "zstd_compress_sequences.h" |
| 16 | |
| 17 | /* |
| 18 | * -log2(x / 256) lookup table for x in [0, 256). |
| 19 | * If x == 0: Return 0 |
| 20 | * Else: Return floor(-log2(x / 256) * 256) |
| 21 | */ |
| 22 | static unsigned const kInverseProbabilityLog256[256] = { |
| 23 | 0, 2048, 1792, 1642, 1536, 1453, 1386, 1329, 1280, 1236, 1197, 1162, |
| 24 | 1130, 1100, 1073, 1047, 1024, 1001, 980, 960, 941, 923, 906, 889, |
| 25 | 874, 859, 844, 830, 817, 804, 791, 779, 768, 756, 745, 734, |
| 26 | 724, 714, 704, 694, 685, 676, 667, 658, 650, 642, 633, 626, |
| 27 | 618, 610, 603, 595, 588, 581, 574, 567, 561, 554, 548, 542, |
| 28 | 535, 529, 523, 517, 512, 506, 500, 495, 489, 484, 478, 473, |
| 29 | 468, 463, 458, 453, 448, 443, 438, 434, 429, 424, 420, 415, |
| 30 | 411, 407, 402, 398, 394, 390, 386, 382, 377, 373, 370, 366, |
| 31 | 362, 358, 354, 350, 347, 343, 339, 336, 332, 329, 325, 322, |
| 32 | 318, 315, 311, 308, 305, 302, 298, 295, 292, 289, 286, 282, |
| 33 | 279, 276, 273, 270, 267, 264, 261, 258, 256, 253, 250, 247, |
| 34 | 244, 241, 239, 236, 233, 230, 228, 225, 222, 220, 217, 215, |
| 35 | 212, 209, 207, 204, 202, 199, 197, 194, 192, 190, 187, 185, |
| 36 | 182, 180, 178, 175, 173, 171, 168, 166, 164, 162, 159, 157, |
| 37 | 155, 153, 151, 149, 146, 144, 142, 140, 138, 136, 134, 132, |
| 38 | 130, 128, 126, 123, 121, 119, 117, 115, 114, 112, 110, 108, |
| 39 | 106, 104, 102, 100, 98, 96, 94, 93, 91, 89, 87, 85, |
| 40 | 83, 82, 80, 78, 76, 74, 73, 71, 69, 67, 66, 64, |
| 41 | 62, 61, 59, 57, 55, 54, 52, 50, 49, 47, 46, 44, |
| 42 | 42, 41, 39, 37, 36, 34, 33, 31, 30, 28, 26, 25, |
| 43 | 23, 22, 20, 19, 17, 16, 14, 13, 11, 10, 8, 7, |
| 44 | 5, 4, 2, 1, |
| 45 | }; |
| 46 | |
| 47 | static unsigned ZSTD_getFSEMaxSymbolValue(FSE_CTable const* ctable) { |
| 48 | void const* ptr = ctable; |
| 49 | U16 const* u16ptr = (U16 const*)ptr; |
| 50 | U32 const maxSymbolValue = MEM_read16(memPtr: u16ptr + 1); |
| 51 | return maxSymbolValue; |
| 52 | } |
| 53 | |
| 54 | /* |
| 55 | * Returns true if we should use ncount=-1 else we should |
| 56 | * use ncount=1 for low probability symbols instead. |
| 57 | */ |
| 58 | static unsigned ZSTD_useLowProbCount(size_t const nbSeq) |
| 59 | { |
| 60 | /* Heuristic: This should cover most blocks <= 16K and |
| 61 | * start to fade out after 16K to about 32K depending on |
| 62 | * compressibility. |
| 63 | */ |
| 64 | return nbSeq >= 2048; |
| 65 | } |
| 66 | |
| 67 | /* |
| 68 | * Returns the cost in bytes of encoding the normalized count header. |
| 69 | * Returns an error if any of the helper functions return an error. |
| 70 | */ |
| 71 | static size_t ZSTD_NCountCost(unsigned const* count, unsigned const max, |
| 72 | size_t const nbSeq, unsigned const FSELog) |
| 73 | { |
| 74 | BYTE wksp[FSE_NCOUNTBOUND]; |
| 75 | S16 norm[MaxSeq + 1]; |
| 76 | const U32 tableLog = FSE_optimalTableLog(maxTableLog: FSELog, srcSize: nbSeq, maxSymbolValue: max); |
| 77 | FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq, max, ZSTD_useLowProbCount(nbSeq)), "" ); |
| 78 | return FSE_writeNCount(buffer: wksp, bufferSize: sizeof(wksp), normalizedCounter: norm, maxSymbolValue: max, tableLog); |
| 79 | } |
| 80 | |
| 81 | /* |
| 82 | * Returns the cost in bits of encoding the distribution described by count |
| 83 | * using the entropy bound. |
| 84 | */ |
| 85 | static size_t ZSTD_entropyCost(unsigned const* count, unsigned const max, size_t const total) |
| 86 | { |
| 87 | unsigned cost = 0; |
| 88 | unsigned s; |
| 89 | |
| 90 | assert(total > 0); |
| 91 | for (s = 0; s <= max; ++s) { |
| 92 | unsigned norm = (unsigned)((256 * count[s]) / total); |
| 93 | if (count[s] != 0 && norm == 0) |
| 94 | norm = 1; |
| 95 | assert(count[s] < total); |
| 96 | cost += count[s] * kInverseProbabilityLog256[norm]; |
| 97 | } |
| 98 | return cost >> 8; |
| 99 | } |
| 100 | |
| 101 | /* |
| 102 | * Returns the cost in bits of encoding the distribution in count using ctable. |
| 103 | * Returns an error if ctable cannot represent all the symbols in count. |
| 104 | */ |
| 105 | size_t ZSTD_fseBitCost( |
| 106 | FSE_CTable const* ctable, |
| 107 | unsigned const* count, |
| 108 | unsigned const max) |
| 109 | { |
| 110 | unsigned const kAccuracyLog = 8; |
| 111 | size_t cost = 0; |
| 112 | unsigned s; |
| 113 | FSE_CState_t cstate; |
| 114 | FSE_initCState(statePtr: &cstate, ct: ctable); |
| 115 | if (ZSTD_getFSEMaxSymbolValue(ctable) < max) { |
| 116 | DEBUGLOG(5, "Repeat FSE_CTable has maxSymbolValue %u < %u" , |
| 117 | ZSTD_getFSEMaxSymbolValue(ctable), max); |
| 118 | return ERROR(GENERIC); |
| 119 | } |
| 120 | for (s = 0; s <= max; ++s) { |
| 121 | unsigned const tableLog = cstate.stateLog; |
| 122 | unsigned const badCost = (tableLog + 1) << kAccuracyLog; |
| 123 | unsigned const bitCost = FSE_bitCost(symbolTTPtr: cstate.symbolTT, tableLog, symbolValue: s, accuracyLog: kAccuracyLog); |
| 124 | if (count[s] == 0) |
| 125 | continue; |
| 126 | if (bitCost >= badCost) { |
| 127 | DEBUGLOG(5, "Repeat FSE_CTable has Prob[%u] == 0" , s); |
| 128 | return ERROR(GENERIC); |
| 129 | } |
| 130 | cost += (size_t)count[s] * bitCost; |
| 131 | } |
| 132 | return cost >> kAccuracyLog; |
| 133 | } |
| 134 | |
| 135 | /* |
| 136 | * Returns the cost in bits of encoding the distribution in count using the |
| 137 | * table described by norm. The max symbol support by norm is assumed >= max. |
| 138 | * norm must be valid for every symbol with non-zero probability in count. |
| 139 | */ |
| 140 | size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog, |
| 141 | unsigned const* count, unsigned const max) |
| 142 | { |
| 143 | unsigned const shift = 8 - accuracyLog; |
| 144 | size_t cost = 0; |
| 145 | unsigned s; |
| 146 | assert(accuracyLog <= 8); |
| 147 | for (s = 0; s <= max; ++s) { |
| 148 | unsigned const normAcc = (norm[s] != -1) ? (unsigned)norm[s] : 1; |
| 149 | unsigned const norm256 = normAcc << shift; |
| 150 | assert(norm256 > 0); |
| 151 | assert(norm256 < 256); |
| 152 | cost += count[s] * kInverseProbabilityLog256[norm256]; |
| 153 | } |
| 154 | return cost >> 8; |
| 155 | } |
| 156 | |
| 157 | SymbolEncodingType_e |
| 158 | ZSTD_selectEncodingType( |
| 159 | FSE_repeat* repeatMode, unsigned const* count, unsigned const max, |
| 160 | size_t const mostFrequent, size_t nbSeq, unsigned const FSELog, |
| 161 | FSE_CTable const* prevCTable, |
| 162 | short const* defaultNorm, U32 defaultNormLog, |
| 163 | ZSTD_DefaultPolicy_e const isDefaultAllowed, |
| 164 | ZSTD_strategy const strategy) |
| 165 | { |
| 166 | ZSTD_STATIC_ASSERT(ZSTD_defaultDisallowed == 0 && ZSTD_defaultAllowed != 0); |
| 167 | if (mostFrequent == nbSeq) { |
| 168 | *repeatMode = FSE_repeat_none; |
| 169 | if (isDefaultAllowed && nbSeq <= 2) { |
| 170 | /* Prefer set_basic over set_rle when there are 2 or fewer symbols, |
| 171 | * since RLE uses 1 byte, but set_basic uses 5-6 bits per symbol. |
| 172 | * If basic encoding isn't possible, always choose RLE. |
| 173 | */ |
| 174 | DEBUGLOG(5, "Selected set_basic" ); |
| 175 | return set_basic; |
| 176 | } |
| 177 | DEBUGLOG(5, "Selected set_rle" ); |
| 178 | return set_rle; |
| 179 | } |
| 180 | if (strategy < ZSTD_lazy) { |
| 181 | if (isDefaultAllowed) { |
| 182 | size_t const staticFse_nbSeq_max = 1000; |
| 183 | size_t const mult = 10 - strategy; |
| 184 | size_t const baseLog = 3; |
| 185 | size_t const dynamicFse_nbSeq_min = (((size_t)1 << defaultNormLog) * mult) >> baseLog; /* 28-36 for offset, 56-72 for lengths */ |
| 186 | assert(defaultNormLog >= 5 && defaultNormLog <= 6); /* xx_DEFAULTNORMLOG */ |
| 187 | assert(mult <= 9 && mult >= 7); |
| 188 | if ( (*repeatMode == FSE_repeat_valid) |
| 189 | && (nbSeq < staticFse_nbSeq_max) ) { |
| 190 | DEBUGLOG(5, "Selected set_repeat" ); |
| 191 | return set_repeat; |
| 192 | } |
| 193 | if ( (nbSeq < dynamicFse_nbSeq_min) |
| 194 | || (mostFrequent < (nbSeq >> (defaultNormLog-1))) ) { |
| 195 | DEBUGLOG(5, "Selected set_basic" ); |
| 196 | /* The format allows default tables to be repeated, but it isn't useful. |
| 197 | * When using simple heuristics to select encoding type, we don't want |
| 198 | * to confuse these tables with dictionaries. When running more careful |
| 199 | * analysis, we don't need to waste time checking both repeating tables |
| 200 | * and default tables. |
| 201 | */ |
| 202 | *repeatMode = FSE_repeat_none; |
| 203 | return set_basic; |
| 204 | } |
| 205 | } |
| 206 | } else { |
| 207 | size_t const basicCost = isDefaultAllowed ? ZSTD_crossEntropyCost(norm: defaultNorm, accuracyLog: defaultNormLog, count, max) : ERROR(GENERIC); |
| 208 | size_t const repeatCost = *repeatMode != FSE_repeat_none ? ZSTD_fseBitCost(ctable: prevCTable, count, max) : ERROR(GENERIC); |
| 209 | size_t const NCountCost = ZSTD_NCountCost(count, max, nbSeq, FSELog); |
| 210 | size_t const compressedCost = (NCountCost << 3) + ZSTD_entropyCost(count, max, total: nbSeq); |
| 211 | |
| 212 | if (isDefaultAllowed) { |
| 213 | assert(!ZSTD_isError(basicCost)); |
| 214 | assert(!(*repeatMode == FSE_repeat_valid && ZSTD_isError(repeatCost))); |
| 215 | } |
| 216 | assert(!ZSTD_isError(NCountCost)); |
| 217 | assert(compressedCost < ERROR(maxCode)); |
| 218 | DEBUGLOG(5, "Estimated bit costs: basic=%u\trepeat=%u\tcompressed=%u" , |
| 219 | (unsigned)basicCost, (unsigned)repeatCost, (unsigned)compressedCost); |
| 220 | if (basicCost <= repeatCost && basicCost <= compressedCost) { |
| 221 | DEBUGLOG(5, "Selected set_basic" ); |
| 222 | assert(isDefaultAllowed); |
| 223 | *repeatMode = FSE_repeat_none; |
| 224 | return set_basic; |
| 225 | } |
| 226 | if (repeatCost <= compressedCost) { |
| 227 | DEBUGLOG(5, "Selected set_repeat" ); |
| 228 | assert(!ZSTD_isError(repeatCost)); |
| 229 | return set_repeat; |
| 230 | } |
| 231 | assert(compressedCost < basicCost && compressedCost < repeatCost); |
| 232 | } |
| 233 | DEBUGLOG(5, "Selected set_compressed" ); |
| 234 | *repeatMode = FSE_repeat_check; |
| 235 | return set_compressed; |
| 236 | } |
| 237 | |
| 238 | typedef struct { |
| 239 | S16 norm[MaxSeq + 1]; |
| 240 | U32 wksp[FSE_BUILD_CTABLE_WORKSPACE_SIZE_U32(MaxSeq, MaxFSELog)]; |
| 241 | } ZSTD_BuildCTableWksp; |
| 242 | |
| 243 | size_t |
| 244 | ZSTD_buildCTable(void* dst, size_t dstCapacity, |
| 245 | FSE_CTable* nextCTable, U32 FSELog, SymbolEncodingType_e type, |
| 246 | unsigned* count, U32 max, |
| 247 | const BYTE* codeTable, size_t nbSeq, |
| 248 | const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax, |
| 249 | const FSE_CTable* prevCTable, size_t prevCTableSize, |
| 250 | void* entropyWorkspace, size_t entropyWorkspaceSize) |
| 251 | { |
| 252 | BYTE* op = (BYTE*)dst; |
| 253 | const BYTE* const oend = op + dstCapacity; |
| 254 | DEBUGLOG(6, "ZSTD_buildCTable (dstCapacity=%u)" , (unsigned)dstCapacity); |
| 255 | |
| 256 | switch (type) { |
| 257 | case set_rle: |
| 258 | FORWARD_IF_ERROR(FSE_buildCTable_rle(nextCTable, (BYTE)max), "" ); |
| 259 | RETURN_ERROR_IF(dstCapacity==0, dstSize_tooSmall, "not enough space" ); |
| 260 | *op = codeTable[0]; |
| 261 | return 1; |
| 262 | case set_repeat: |
| 263 | ZSTD_memcpy(nextCTable, prevCTable, prevCTableSize); |
| 264 | return 0; |
| 265 | case set_basic: |
| 266 | FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize), "" ); /* note : could be pre-calculated */ |
| 267 | return 0; |
| 268 | case set_compressed: { |
| 269 | ZSTD_BuildCTableWksp* wksp = (ZSTD_BuildCTableWksp*)entropyWorkspace; |
| 270 | size_t nbSeq_1 = nbSeq; |
| 271 | const U32 tableLog = FSE_optimalTableLog(maxTableLog: FSELog, srcSize: nbSeq, maxSymbolValue: max); |
| 272 | if (count[codeTable[nbSeq-1]] > 1) { |
| 273 | count[codeTable[nbSeq-1]]--; |
| 274 | nbSeq_1--; |
| 275 | } |
| 276 | assert(nbSeq_1 > 1); |
| 277 | assert(entropyWorkspaceSize >= sizeof(ZSTD_BuildCTableWksp)); |
| 278 | (void)entropyWorkspaceSize; |
| 279 | FORWARD_IF_ERROR(FSE_normalizeCount(wksp->norm, tableLog, count, nbSeq_1, max, ZSTD_useLowProbCount(nbSeq_1)), "FSE_normalizeCount failed" ); |
| 280 | assert(oend >= op); |
| 281 | { size_t const NCountSize = FSE_writeNCount(buffer: op, bufferSize: (size_t)(oend - op), normalizedCounter: wksp->norm, maxSymbolValue: max, tableLog); /* overflow protected */ |
| 282 | FORWARD_IF_ERROR(NCountSize, "FSE_writeNCount failed" ); |
| 283 | FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, wksp->norm, max, tableLog, wksp->wksp, sizeof(wksp->wksp)), "FSE_buildCTable_wksp failed" ); |
| 284 | return NCountSize; |
| 285 | } |
| 286 | } |
| 287 | default: assert(0); RETURN_ERROR(GENERIC, "impossible to reach" ); |
| 288 | } |
| 289 | } |
| 290 | |
| 291 | FORCE_INLINE_TEMPLATE size_t |
| 292 | ZSTD_encodeSequences_body( |
| 293 | void* dst, size_t dstCapacity, |
| 294 | FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable, |
| 295 | FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable, |
| 296 | FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable, |
| 297 | SeqDef const* sequences, size_t nbSeq, int longOffsets) |
| 298 | { |
| 299 | BIT_CStream_t blockStream; |
| 300 | FSE_CState_t stateMatchLength; |
| 301 | FSE_CState_t stateOffsetBits; |
| 302 | FSE_CState_t stateLitLength; |
| 303 | |
| 304 | RETURN_ERROR_IF( |
| 305 | ERR_isError(BIT_initCStream(&blockStream, dst, dstCapacity)), |
| 306 | dstSize_tooSmall, "not enough space remaining" ); |
| 307 | DEBUGLOG(6, "available space for bitstream : %i (dstCapacity=%u)" , |
| 308 | (int)(blockStream.endPtr - blockStream.startPtr), |
| 309 | (unsigned)dstCapacity); |
| 310 | |
| 311 | /* first symbols */ |
| 312 | FSE_initCState2(statePtr: &stateMatchLength, ct: CTable_MatchLength, symbol: mlCodeTable[nbSeq-1]); |
| 313 | FSE_initCState2(statePtr: &stateOffsetBits, ct: CTable_OffsetBits, symbol: ofCodeTable[nbSeq-1]); |
| 314 | FSE_initCState2(statePtr: &stateLitLength, ct: CTable_LitLength, symbol: llCodeTable[nbSeq-1]); |
| 315 | BIT_addBits(bitC: &blockStream, value: sequences[nbSeq-1].litLength, nbBits: LL_bits[llCodeTable[nbSeq-1]]); |
| 316 | if (MEM_32bits()) BIT_flushBits(bitC: &blockStream); |
| 317 | BIT_addBits(bitC: &blockStream, value: sequences[nbSeq-1].mlBase, nbBits: ML_bits[mlCodeTable[nbSeq-1]]); |
| 318 | if (MEM_32bits()) BIT_flushBits(bitC: &blockStream); |
| 319 | if (longOffsets) { |
| 320 | U32 const ofBits = ofCodeTable[nbSeq-1]; |
| 321 | unsigned const = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1); |
| 322 | if (extraBits) { |
| 323 | BIT_addBits(bitC: &blockStream, value: sequences[nbSeq-1].offBase, nbBits: extraBits); |
| 324 | BIT_flushBits(bitC: &blockStream); |
| 325 | } |
| 326 | BIT_addBits(bitC: &blockStream, value: sequences[nbSeq-1].offBase >> extraBits, |
| 327 | nbBits: ofBits - extraBits); |
| 328 | } else { |
| 329 | BIT_addBits(bitC: &blockStream, value: sequences[nbSeq-1].offBase, nbBits: ofCodeTable[nbSeq-1]); |
| 330 | } |
| 331 | BIT_flushBits(bitC: &blockStream); |
| 332 | |
| 333 | { size_t n; |
| 334 | for (n=nbSeq-2 ; n<nbSeq ; n--) { /* intentional underflow */ |
| 335 | BYTE const llCode = llCodeTable[n]; |
| 336 | BYTE const ofCode = ofCodeTable[n]; |
| 337 | BYTE const mlCode = mlCodeTable[n]; |
| 338 | U32 const llBits = LL_bits[llCode]; |
| 339 | U32 const ofBits = ofCode; |
| 340 | U32 const mlBits = ML_bits[mlCode]; |
| 341 | DEBUGLOG(6, "encoding: litlen:%2u - matchlen:%2u - offCode:%7u" , |
| 342 | (unsigned)sequences[n].litLength, |
| 343 | (unsigned)sequences[n].mlBase + MINMATCH, |
| 344 | (unsigned)sequences[n].offBase); |
| 345 | /* 32b*/ /* 64b*/ |
| 346 | /* (7)*/ /* (7)*/ |
| 347 | FSE_encodeSymbol(bitC: &blockStream, statePtr: &stateOffsetBits, symbol: ofCode); /* 15 */ /* 15 */ |
| 348 | FSE_encodeSymbol(bitC: &blockStream, statePtr: &stateMatchLength, symbol: mlCode); /* 24 */ /* 24 */ |
| 349 | if (MEM_32bits()) BIT_flushBits(bitC: &blockStream); /* (7)*/ |
| 350 | FSE_encodeSymbol(bitC: &blockStream, statePtr: &stateLitLength, symbol: llCode); /* 16 */ /* 33 */ |
| 351 | if (MEM_32bits() || (ofBits+mlBits+llBits >= 64-7-(LLFSELog+MLFSELog+OffFSELog))) |
| 352 | BIT_flushBits(bitC: &blockStream); /* (7)*/ |
| 353 | BIT_addBits(bitC: &blockStream, value: sequences[n].litLength, nbBits: llBits); |
| 354 | if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(bitC: &blockStream); |
| 355 | BIT_addBits(bitC: &blockStream, value: sequences[n].mlBase, nbBits: mlBits); |
| 356 | if (MEM_32bits() || (ofBits+mlBits+llBits > 56)) BIT_flushBits(bitC: &blockStream); |
| 357 | if (longOffsets) { |
| 358 | unsigned const = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1); |
| 359 | if (extraBits) { |
| 360 | BIT_addBits(bitC: &blockStream, value: sequences[n].offBase, nbBits: extraBits); |
| 361 | BIT_flushBits(bitC: &blockStream); /* (7)*/ |
| 362 | } |
| 363 | BIT_addBits(bitC: &blockStream, value: sequences[n].offBase >> extraBits, |
| 364 | nbBits: ofBits - extraBits); /* 31 */ |
| 365 | } else { |
| 366 | BIT_addBits(bitC: &blockStream, value: sequences[n].offBase, nbBits: ofBits); /* 31 */ |
| 367 | } |
| 368 | BIT_flushBits(bitC: &blockStream); /* (7)*/ |
| 369 | DEBUGLOG(7, "remaining space : %i" , (int)(blockStream.endPtr - blockStream.ptr)); |
| 370 | } } |
| 371 | |
| 372 | DEBUGLOG(6, "ZSTD_encodeSequences: flushing ML state with %u bits" , stateMatchLength.stateLog); |
| 373 | FSE_flushCState(bitC: &blockStream, statePtr: &stateMatchLength); |
| 374 | DEBUGLOG(6, "ZSTD_encodeSequences: flushing Off state with %u bits" , stateOffsetBits.stateLog); |
| 375 | FSE_flushCState(bitC: &blockStream, statePtr: &stateOffsetBits); |
| 376 | DEBUGLOG(6, "ZSTD_encodeSequences: flushing LL state with %u bits" , stateLitLength.stateLog); |
| 377 | FSE_flushCState(bitC: &blockStream, statePtr: &stateLitLength); |
| 378 | |
| 379 | { size_t const streamSize = BIT_closeCStream(bitC: &blockStream); |
| 380 | RETURN_ERROR_IF(streamSize==0, dstSize_tooSmall, "not enough space" ); |
| 381 | return streamSize; |
| 382 | } |
| 383 | } |
| 384 | |
| 385 | static size_t |
| 386 | ZSTD_encodeSequences_default( |
| 387 | void* dst, size_t dstCapacity, |
| 388 | FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable, |
| 389 | FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable, |
| 390 | FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable, |
| 391 | SeqDef const* sequences, size_t nbSeq, int longOffsets) |
| 392 | { |
| 393 | return ZSTD_encodeSequences_body(dst, dstCapacity, |
| 394 | CTable_MatchLength, mlCodeTable, |
| 395 | CTable_OffsetBits, ofCodeTable, |
| 396 | CTable_LitLength, llCodeTable, |
| 397 | sequences, nbSeq, longOffsets); |
| 398 | } |
| 399 | |
| 400 | |
| 401 | #if DYNAMIC_BMI2 |
| 402 | |
| 403 | static BMI2_TARGET_ATTRIBUTE size_t |
| 404 | ZSTD_encodeSequences_bmi2( |
| 405 | void* dst, size_t dstCapacity, |
| 406 | FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable, |
| 407 | FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable, |
| 408 | FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable, |
| 409 | SeqDef const* sequences, size_t nbSeq, int longOffsets) |
| 410 | { |
| 411 | return ZSTD_encodeSequences_body(dst, dstCapacity, |
| 412 | CTable_MatchLength, mlCodeTable, |
| 413 | CTable_OffsetBits, ofCodeTable, |
| 414 | CTable_LitLength, llCodeTable, |
| 415 | sequences, nbSeq, longOffsets); |
| 416 | } |
| 417 | |
| 418 | #endif |
| 419 | |
| 420 | size_t ZSTD_encodeSequences( |
| 421 | void* dst, size_t dstCapacity, |
| 422 | FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable, |
| 423 | FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable, |
| 424 | FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable, |
| 425 | SeqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2) |
| 426 | { |
| 427 | DEBUGLOG(5, "ZSTD_encodeSequences: dstCapacity = %u" , (unsigned)dstCapacity); |
| 428 | #if DYNAMIC_BMI2 |
| 429 | if (bmi2) { |
| 430 | return ZSTD_encodeSequences_bmi2(dst, dstCapacity, |
| 431 | CTable_MatchLength, mlCodeTable, |
| 432 | CTable_OffsetBits, ofCodeTable, |
| 433 | CTable_LitLength, llCodeTable, |
| 434 | sequences, nbSeq, longOffsets); |
| 435 | } |
| 436 | #endif |
| 437 | (void)bmi2; |
| 438 | return ZSTD_encodeSequences_default(dst, dstCapacity, |
| 439 | CTable_MatchLength, mlCodeTable, |
| 440 | CTable_OffsetBits, ofCodeTable, |
| 441 | CTable_LitLength, llCodeTable, |
| 442 | sequences, nbSeq, longOffsets); |
| 443 | } |
| 444 | |