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