Scott Baker | e7144bc | 2019-10-01 14:16:47 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2016-present, 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 | #include "zstd_compress_internal.h" |
| 12 | #include "zstd_lazy.h" |
| 13 | |
| 14 | |
| 15 | /*-************************************* |
| 16 | * Binary Tree search |
| 17 | ***************************************/ |
| 18 | |
| 19 | void ZSTD_updateDUBT( |
| 20 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 21 | const BYTE* ip, const BYTE* iend, |
| 22 | U32 mls) |
| 23 | { |
| 24 | U32* const hashTable = ms->hashTable; |
| 25 | U32 const hashLog = cParams->hashLog; |
| 26 | |
| 27 | U32* const bt = ms->chainTable; |
| 28 | U32 const btLog = cParams->chainLog - 1; |
| 29 | U32 const btMask = (1 << btLog) - 1; |
| 30 | |
| 31 | const BYTE* const base = ms->window.base; |
| 32 | U32 const target = (U32)(ip - base); |
| 33 | U32 idx = ms->nextToUpdate; |
| 34 | |
| 35 | if (idx != target) |
| 36 | DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)", |
| 37 | idx, target, ms->window.dictLimit); |
| 38 | assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */ |
| 39 | (void)iend; |
| 40 | |
| 41 | assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */ |
| 42 | for ( ; idx < target ; idx++) { |
| 43 | size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */ |
| 44 | U32 const matchIndex = hashTable[h]; |
| 45 | |
| 46 | U32* const nextCandidatePtr = bt + 2*(idx&btMask); |
| 47 | U32* const sortMarkPtr = nextCandidatePtr + 1; |
| 48 | |
| 49 | DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx); |
| 50 | hashTable[h] = idx; /* Update Hash Table */ |
| 51 | *nextCandidatePtr = matchIndex; /* update BT like a chain */ |
| 52 | *sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK; |
| 53 | } |
| 54 | ms->nextToUpdate = target; |
| 55 | } |
| 56 | |
| 57 | |
| 58 | /** ZSTD_insertDUBT1() : |
| 59 | * sort one already inserted but unsorted position |
| 60 | * assumption : current >= btlow == (current - btmask) |
| 61 | * doesn't fail */ |
| 62 | static void ZSTD_insertDUBT1( |
| 63 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 64 | U32 current, const BYTE* inputEnd, |
| 65 | U32 nbCompares, U32 btLow, int extDict) |
| 66 | { |
| 67 | U32* const bt = ms->chainTable; |
| 68 | U32 const btLog = cParams->chainLog - 1; |
| 69 | U32 const btMask = (1 << btLog) - 1; |
| 70 | size_t commonLengthSmaller=0, commonLengthLarger=0; |
| 71 | const BYTE* const base = ms->window.base; |
| 72 | const BYTE* const dictBase = ms->window.dictBase; |
| 73 | const U32 dictLimit = ms->window.dictLimit; |
| 74 | const BYTE* const ip = (current>=dictLimit) ? base + current : dictBase + current; |
| 75 | const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit; |
| 76 | const BYTE* const dictEnd = dictBase + dictLimit; |
| 77 | const BYTE* const prefixStart = base + dictLimit; |
| 78 | const BYTE* match; |
| 79 | U32* smallerPtr = bt + 2*(current&btMask); |
| 80 | U32* largerPtr = smallerPtr + 1; |
| 81 | U32 matchIndex = *smallerPtr; |
| 82 | U32 dummy32; /* to be nullified at the end */ |
| 83 | U32 const windowLow = ms->window.lowLimit; |
| 84 | |
| 85 | DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)", |
| 86 | current, dictLimit, windowLow); |
| 87 | assert(current >= btLow); |
| 88 | assert(ip < iend); /* condition for ZSTD_count */ |
| 89 | |
| 90 | while (nbCompares-- && (matchIndex > windowLow)) { |
| 91 | U32* const nextPtr = bt + 2*(matchIndex & btMask); |
| 92 | size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */ |
| 93 | assert(matchIndex < current); |
| 94 | |
| 95 | if ( (!extDict) |
| 96 | || (matchIndex+matchLength >= dictLimit) /* both in current segment*/ |
| 97 | || (current < dictLimit) /* both in extDict */) { |
| 98 | const BYTE* const mBase = !extDict || ((matchIndex+matchLength) >= dictLimit) ? base : dictBase; |
| 99 | assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */ |
| 100 | || (current < dictLimit) ); |
| 101 | match = mBase + matchIndex; |
| 102 | matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend); |
| 103 | } else { |
| 104 | match = dictBase + matchIndex; |
| 105 | matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart); |
| 106 | if (matchIndex+matchLength >= dictLimit) |
| 107 | match = base + matchIndex; /* to prepare for next usage of match[matchLength] */ |
| 108 | } |
| 109 | |
| 110 | DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ", |
| 111 | current, matchIndex, (U32)matchLength); |
| 112 | |
| 113 | if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */ |
| 114 | break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */ |
| 115 | } |
| 116 | |
| 117 | if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */ |
| 118 | /* match is smaller than current */ |
| 119 | *smallerPtr = matchIndex; /* update smaller idx */ |
| 120 | commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */ |
| 121 | if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */ |
| 122 | DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u", |
| 123 | matchIndex, btLow, nextPtr[1]); |
| 124 | smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */ |
| 125 | matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */ |
| 126 | } else { |
| 127 | /* match is larger than current */ |
| 128 | *largerPtr = matchIndex; |
| 129 | commonLengthLarger = matchLength; |
| 130 | if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */ |
| 131 | DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u", |
| 132 | matchIndex, btLow, nextPtr[0]); |
| 133 | largerPtr = nextPtr; |
| 134 | matchIndex = nextPtr[0]; |
| 135 | } } |
| 136 | |
| 137 | *smallerPtr = *largerPtr = 0; |
| 138 | } |
| 139 | |
| 140 | |
| 141 | static size_t ZSTD_DUBT_findBestMatch ( |
| 142 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 143 | const BYTE* const ip, const BYTE* const iend, |
| 144 | size_t* offsetPtr, |
| 145 | U32 const mls, |
| 146 | U32 const extDict) |
| 147 | { |
| 148 | U32* const hashTable = ms->hashTable; |
| 149 | U32 const hashLog = cParams->hashLog; |
| 150 | size_t const h = ZSTD_hashPtr(ip, hashLog, mls); |
| 151 | U32 matchIndex = hashTable[h]; |
| 152 | |
| 153 | const BYTE* const base = ms->window.base; |
| 154 | U32 const current = (U32)(ip-base); |
| 155 | U32 const windowLow = ms->window.lowLimit; |
| 156 | |
| 157 | U32* const bt = ms->chainTable; |
| 158 | U32 const btLog = cParams->chainLog - 1; |
| 159 | U32 const btMask = (1 << btLog) - 1; |
| 160 | U32 const btLow = (btMask >= current) ? 0 : current - btMask; |
| 161 | U32 const unsortLimit = MAX(btLow, windowLow); |
| 162 | |
| 163 | U32* nextCandidate = bt + 2*(matchIndex&btMask); |
| 164 | U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1; |
| 165 | U32 nbCompares = 1U << cParams->searchLog; |
| 166 | U32 nbCandidates = nbCompares; |
| 167 | U32 previousCandidate = 0; |
| 168 | |
| 169 | DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", current); |
| 170 | assert(ip <= iend-8); /* required for h calculation */ |
| 171 | |
| 172 | /* reach end of unsorted candidates list */ |
| 173 | while ( (matchIndex > unsortLimit) |
| 174 | && (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK) |
| 175 | && (nbCandidates > 1) ) { |
| 176 | DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted", |
| 177 | matchIndex); |
| 178 | *unsortedMark = previousCandidate; |
| 179 | previousCandidate = matchIndex; |
| 180 | matchIndex = *nextCandidate; |
| 181 | nextCandidate = bt + 2*(matchIndex&btMask); |
| 182 | unsortedMark = bt + 2*(matchIndex&btMask) + 1; |
| 183 | nbCandidates --; |
| 184 | } |
| 185 | |
| 186 | if ( (matchIndex > unsortLimit) |
| 187 | && (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) { |
| 188 | DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u", |
| 189 | matchIndex); |
| 190 | *nextCandidate = *unsortedMark = 0; /* nullify next candidate if it's still unsorted (note : simplification, detrimental to compression ratio, beneficial for speed) */ |
| 191 | } |
| 192 | |
| 193 | /* batch sort stacked candidates */ |
| 194 | matchIndex = previousCandidate; |
| 195 | while (matchIndex) { /* will end on matchIndex == 0 */ |
| 196 | U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1; |
| 197 | U32 const nextCandidateIdx = *nextCandidateIdxPtr; |
| 198 | ZSTD_insertDUBT1(ms, cParams, matchIndex, iend, |
| 199 | nbCandidates, unsortLimit, extDict); |
| 200 | matchIndex = nextCandidateIdx; |
| 201 | nbCandidates++; |
| 202 | } |
| 203 | |
| 204 | /* find longest match */ |
| 205 | { size_t commonLengthSmaller=0, commonLengthLarger=0; |
| 206 | const BYTE* const dictBase = ms->window.dictBase; |
| 207 | const U32 dictLimit = ms->window.dictLimit; |
| 208 | const BYTE* const dictEnd = dictBase + dictLimit; |
| 209 | const BYTE* const prefixStart = base + dictLimit; |
| 210 | U32* smallerPtr = bt + 2*(current&btMask); |
| 211 | U32* largerPtr = bt + 2*(current&btMask) + 1; |
| 212 | U32 matchEndIdx = current+8+1; |
| 213 | U32 dummy32; /* to be nullified at the end */ |
| 214 | size_t bestLength = 0; |
| 215 | |
| 216 | matchIndex = hashTable[h]; |
| 217 | hashTable[h] = current; /* Update Hash Table */ |
| 218 | |
| 219 | while (nbCompares-- && (matchIndex > windowLow)) { |
| 220 | U32* const nextPtr = bt + 2*(matchIndex & btMask); |
| 221 | size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */ |
| 222 | const BYTE* match; |
| 223 | |
| 224 | if ((!extDict) || (matchIndex+matchLength >= dictLimit)) { |
| 225 | match = base + matchIndex; |
| 226 | matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend); |
| 227 | } else { |
| 228 | match = dictBase + matchIndex; |
| 229 | matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart); |
| 230 | if (matchIndex+matchLength >= dictLimit) |
| 231 | match = base + matchIndex; /* to prepare for next usage of match[matchLength] */ |
| 232 | } |
| 233 | |
| 234 | if (matchLength > bestLength) { |
| 235 | if (matchLength > matchEndIdx - matchIndex) |
| 236 | matchEndIdx = matchIndex + (U32)matchLength; |
| 237 | if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) |
| 238 | bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex; |
| 239 | if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */ |
| 240 | break; /* drop, to guarantee consistency (miss a little bit of compression) */ |
| 241 | } |
| 242 | } |
| 243 | |
| 244 | if (match[matchLength] < ip[matchLength]) { |
| 245 | /* match is smaller than current */ |
| 246 | *smallerPtr = matchIndex; /* update smaller idx */ |
| 247 | commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */ |
| 248 | if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */ |
| 249 | smallerPtr = nextPtr+1; /* new "smaller" => larger of match */ |
| 250 | matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */ |
| 251 | } else { |
| 252 | /* match is larger than current */ |
| 253 | *largerPtr = matchIndex; |
| 254 | commonLengthLarger = matchLength; |
| 255 | if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */ |
| 256 | largerPtr = nextPtr; |
| 257 | matchIndex = nextPtr[0]; |
| 258 | } } |
| 259 | |
| 260 | *smallerPtr = *largerPtr = 0; |
| 261 | |
| 262 | assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */ |
| 263 | ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */ |
| 264 | if (bestLength >= MINMATCH) { |
| 265 | U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex; |
| 266 | DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)", |
| 267 | current, (U32)bestLength, (U32)*offsetPtr, mIndex); |
| 268 | } |
| 269 | return bestLength; |
| 270 | } |
| 271 | } |
| 272 | |
| 273 | |
| 274 | /** ZSTD_BtFindBestMatch() : Tree updater, providing best match */ |
| 275 | static size_t ZSTD_BtFindBestMatch ( |
| 276 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 277 | const BYTE* const ip, const BYTE* const iLimit, |
| 278 | size_t* offsetPtr, |
| 279 | const U32 mls /* template */) |
| 280 | { |
| 281 | DEBUGLOG(7, "ZSTD_BtFindBestMatch"); |
| 282 | if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */ |
| 283 | ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls); |
| 284 | return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 0); |
| 285 | } |
| 286 | |
| 287 | |
| 288 | static size_t ZSTD_BtFindBestMatch_selectMLS ( |
| 289 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 290 | const BYTE* ip, const BYTE* const iLimit, |
| 291 | size_t* offsetPtr) |
| 292 | { |
| 293 | switch(cParams->searchLength) |
| 294 | { |
| 295 | default : /* includes case 3 */ |
| 296 | case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4); |
| 297 | case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5); |
| 298 | case 7 : |
| 299 | case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6); |
| 300 | } |
| 301 | } |
| 302 | |
| 303 | |
| 304 | /** Tree updater, providing best match */ |
| 305 | static size_t ZSTD_BtFindBestMatch_extDict ( |
| 306 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 307 | const BYTE* const ip, const BYTE* const iLimit, |
| 308 | size_t* offsetPtr, |
| 309 | const U32 mls) |
| 310 | { |
| 311 | DEBUGLOG(7, "ZSTD_BtFindBestMatch_extDict"); |
| 312 | if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */ |
| 313 | ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls); |
| 314 | return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 1); |
| 315 | } |
| 316 | |
| 317 | |
| 318 | static size_t ZSTD_BtFindBestMatch_selectMLS_extDict ( |
| 319 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 320 | const BYTE* ip, const BYTE* const iLimit, |
| 321 | size_t* offsetPtr) |
| 322 | { |
| 323 | switch(cParams->searchLength) |
| 324 | { |
| 325 | default : /* includes case 3 */ |
| 326 | case 4 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 4); |
| 327 | case 5 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 5); |
| 328 | case 7 : |
| 329 | case 6 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 6); |
| 330 | } |
| 331 | } |
| 332 | |
| 333 | |
| 334 | |
| 335 | /* ********************************* |
| 336 | * Hash Chain |
| 337 | ***********************************/ |
| 338 | #define NEXT_IN_CHAIN(d, mask) chainTable[(d) & mask] |
| 339 | |
| 340 | /* Update chains up to ip (excluded) |
| 341 | Assumption : always within prefix (i.e. not within extDict) */ |
| 342 | static U32 ZSTD_insertAndFindFirstIndex_internal( |
| 343 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 344 | const BYTE* ip, U32 const mls) |
| 345 | { |
| 346 | U32* const hashTable = ms->hashTable; |
| 347 | const U32 hashLog = cParams->hashLog; |
| 348 | U32* const chainTable = ms->chainTable; |
| 349 | const U32 chainMask = (1 << cParams->chainLog) - 1; |
| 350 | const BYTE* const base = ms->window.base; |
| 351 | const U32 target = (U32)(ip - base); |
| 352 | U32 idx = ms->nextToUpdate; |
| 353 | |
| 354 | while(idx < target) { /* catch up */ |
| 355 | size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls); |
| 356 | NEXT_IN_CHAIN(idx, chainMask) = hashTable[h]; |
| 357 | hashTable[h] = idx; |
| 358 | idx++; |
| 359 | } |
| 360 | |
| 361 | ms->nextToUpdate = target; |
| 362 | return hashTable[ZSTD_hashPtr(ip, hashLog, mls)]; |
| 363 | } |
| 364 | |
| 365 | U32 ZSTD_insertAndFindFirstIndex( |
| 366 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 367 | const BYTE* ip) |
| 368 | { |
| 369 | return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, cParams->searchLength); |
| 370 | } |
| 371 | |
| 372 | |
| 373 | /* inlining is important to hardwire a hot branch (template emulation) */ |
| 374 | FORCE_INLINE_TEMPLATE |
| 375 | size_t ZSTD_HcFindBestMatch_generic ( |
| 376 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 377 | const BYTE* const ip, const BYTE* const iLimit, |
| 378 | size_t* offsetPtr, |
| 379 | const U32 mls, const U32 extDict) |
| 380 | { |
| 381 | U32* const chainTable = ms->chainTable; |
| 382 | const U32 chainSize = (1 << cParams->chainLog); |
| 383 | const U32 chainMask = chainSize-1; |
| 384 | const BYTE* const base = ms->window.base; |
| 385 | const BYTE* const dictBase = ms->window.dictBase; |
| 386 | const U32 dictLimit = ms->window.dictLimit; |
| 387 | const BYTE* const prefixStart = base + dictLimit; |
| 388 | const BYTE* const dictEnd = dictBase + dictLimit; |
| 389 | const U32 lowLimit = ms->window.lowLimit; |
| 390 | const U32 current = (U32)(ip-base); |
| 391 | const U32 minChain = current > chainSize ? current - chainSize : 0; |
| 392 | U32 nbAttempts = 1U << cParams->searchLog; |
| 393 | size_t ml=4-1; |
| 394 | |
| 395 | /* HC4 match finder */ |
| 396 | U32 matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls); |
| 397 | |
| 398 | for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) { |
| 399 | size_t currentMl=0; |
| 400 | if ((!extDict) || matchIndex >= dictLimit) { |
| 401 | const BYTE* const match = base + matchIndex; |
| 402 | if (match[ml] == ip[ml]) /* potentially better */ |
| 403 | currentMl = ZSTD_count(ip, match, iLimit); |
| 404 | } else { |
| 405 | const BYTE* const match = dictBase + matchIndex; |
| 406 | assert(match+4 <= dictEnd); |
| 407 | if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */ |
| 408 | currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4; |
| 409 | } |
| 410 | |
| 411 | /* save best solution */ |
| 412 | if (currentMl > ml) { |
| 413 | ml = currentMl; |
| 414 | *offsetPtr = current - matchIndex + ZSTD_REP_MOVE; |
| 415 | if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */ |
| 416 | } |
| 417 | |
| 418 | if (matchIndex <= minChain) break; |
| 419 | matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask); |
| 420 | } |
| 421 | |
| 422 | return ml; |
| 423 | } |
| 424 | |
| 425 | |
| 426 | FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS ( |
| 427 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 428 | const BYTE* ip, const BYTE* const iLimit, |
| 429 | size_t* offsetPtr) |
| 430 | { |
| 431 | switch(cParams->searchLength) |
| 432 | { |
| 433 | default : /* includes case 3 */ |
| 434 | case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 0); |
| 435 | case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 0); |
| 436 | case 7 : |
| 437 | case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 0); |
| 438 | } |
| 439 | } |
| 440 | |
| 441 | |
| 442 | FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS ( |
| 443 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 444 | const BYTE* ip, const BYTE* const iLimit, |
| 445 | size_t* const offsetPtr) |
| 446 | { |
| 447 | switch(cParams->searchLength) |
| 448 | { |
| 449 | default : /* includes case 3 */ |
| 450 | case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 1); |
| 451 | case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 1); |
| 452 | case 7 : |
| 453 | case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 1); |
| 454 | } |
| 455 | } |
| 456 | |
| 457 | |
| 458 | /* ******************************* |
| 459 | * Common parser - lazy strategy |
| 460 | *********************************/ |
| 461 | FORCE_INLINE_TEMPLATE |
| 462 | size_t ZSTD_compressBlock_lazy_generic( |
| 463 | ZSTD_matchState_t* ms, seqStore_t* seqStore, |
| 464 | U32 rep[ZSTD_REP_NUM], |
| 465 | ZSTD_compressionParameters const* cParams, |
| 466 | const void* src, size_t srcSize, |
| 467 | const U32 searchMethod, const U32 depth) |
| 468 | { |
| 469 | const BYTE* const istart = (const BYTE*)src; |
| 470 | const BYTE* ip = istart; |
| 471 | const BYTE* anchor = istart; |
| 472 | const BYTE* const iend = istart + srcSize; |
| 473 | const BYTE* const ilimit = iend - 8; |
| 474 | const BYTE* const base = ms->window.base + ms->window.dictLimit; |
| 475 | |
| 476 | typedef size_t (*searchMax_f)( |
| 477 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 478 | const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr); |
| 479 | searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS; |
| 480 | U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0; |
| 481 | |
| 482 | /* init */ |
| 483 | ip += (ip==base); |
| 484 | ms->nextToUpdate3 = ms->nextToUpdate; |
| 485 | { U32 const maxRep = (U32)(ip-base); |
| 486 | if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0; |
| 487 | if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0; |
| 488 | } |
| 489 | |
| 490 | /* Match Loop */ |
| 491 | while (ip < ilimit) { |
| 492 | size_t matchLength=0; |
| 493 | size_t offset=0; |
| 494 | const BYTE* start=ip+1; |
| 495 | |
| 496 | /* check repCode */ |
| 497 | if ((offset_1>0) & (MEM_read32(ip+1) == MEM_read32(ip+1 - offset_1))) { |
| 498 | /* repcode : we take it */ |
| 499 | matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4; |
| 500 | if (depth==0) goto _storeSequence; |
| 501 | } |
| 502 | |
| 503 | /* first search (depth 0) */ |
| 504 | { size_t offsetFound = 99999999; |
| 505 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound); |
| 506 | if (ml2 > matchLength) |
| 507 | matchLength = ml2, start = ip, offset=offsetFound; |
| 508 | } |
| 509 | |
| 510 | if (matchLength < 4) { |
| 511 | ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */ |
| 512 | continue; |
| 513 | } |
| 514 | |
| 515 | /* let's try to find a better solution */ |
| 516 | if (depth>=1) |
| 517 | while (ip<ilimit) { |
| 518 | ip ++; |
| 519 | if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) { |
| 520 | size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4; |
| 521 | int const gain2 = (int)(mlRep * 3); |
| 522 | int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1); |
| 523 | if ((mlRep >= 4) && (gain2 > gain1)) |
| 524 | matchLength = mlRep, offset = 0, start = ip; |
| 525 | } |
| 526 | { size_t offset2=99999999; |
| 527 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2); |
| 528 | int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */ |
| 529 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4); |
| 530 | if ((ml2 >= 4) && (gain2 > gain1)) { |
| 531 | matchLength = ml2, offset = offset2, start = ip; |
| 532 | continue; /* search a better one */ |
| 533 | } } |
| 534 | |
| 535 | /* let's find an even better one */ |
| 536 | if ((depth==2) && (ip<ilimit)) { |
| 537 | ip ++; |
| 538 | if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) { |
| 539 | size_t const ml2 = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4; |
| 540 | int const gain2 = (int)(ml2 * 4); |
| 541 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1); |
| 542 | if ((ml2 >= 4) && (gain2 > gain1)) |
| 543 | matchLength = ml2, offset = 0, start = ip; |
| 544 | } |
| 545 | { size_t offset2=99999999; |
| 546 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2); |
| 547 | int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */ |
| 548 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7); |
| 549 | if ((ml2 >= 4) && (gain2 > gain1)) { |
| 550 | matchLength = ml2, offset = offset2, start = ip; |
| 551 | continue; |
| 552 | } } } |
| 553 | break; /* nothing found : store previous solution */ |
| 554 | } |
| 555 | |
| 556 | /* NOTE: |
| 557 | * start[-offset+ZSTD_REP_MOVE-1] is undefined behavior. |
| 558 | * (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which |
| 559 | * overflows the pointer, which is undefined behavior. |
| 560 | */ |
| 561 | /* catch up */ |
| 562 | if (offset) { |
| 563 | while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > base)) |
| 564 | && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */ |
| 565 | { start--; matchLength++; } |
| 566 | offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE); |
| 567 | } |
| 568 | /* store sequence */ |
| 569 | _storeSequence: |
| 570 | { size_t const litLength = start - anchor; |
| 571 | ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH); |
| 572 | anchor = ip = start + matchLength; |
| 573 | } |
| 574 | |
| 575 | /* check immediate repcode */ |
| 576 | while ( ((ip <= ilimit) & (offset_2>0)) |
| 577 | && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) { |
| 578 | /* store sequence */ |
| 579 | matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4; |
| 580 | offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */ |
| 581 | ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH); |
| 582 | ip += matchLength; |
| 583 | anchor = ip; |
| 584 | continue; /* faster when present ... (?) */ |
| 585 | } } |
| 586 | |
| 587 | /* Save reps for next block */ |
| 588 | rep[0] = offset_1 ? offset_1 : savedOffset; |
| 589 | rep[1] = offset_2 ? offset_2 : savedOffset; |
| 590 | |
| 591 | /* Return the last literals size */ |
| 592 | return iend - anchor; |
| 593 | } |
| 594 | |
| 595 | |
| 596 | size_t ZSTD_compressBlock_btlazy2( |
| 597 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 598 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 599 | { |
| 600 | return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2); |
| 601 | } |
| 602 | |
| 603 | size_t ZSTD_compressBlock_lazy2( |
| 604 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 605 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 606 | { |
| 607 | return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2); |
| 608 | } |
| 609 | |
| 610 | size_t ZSTD_compressBlock_lazy( |
| 611 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 612 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 613 | { |
| 614 | return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1); |
| 615 | } |
| 616 | |
| 617 | size_t ZSTD_compressBlock_greedy( |
| 618 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 619 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 620 | { |
| 621 | return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0); |
| 622 | } |
| 623 | |
| 624 | |
| 625 | FORCE_INLINE_TEMPLATE |
| 626 | size_t ZSTD_compressBlock_lazy_extDict_generic( |
| 627 | ZSTD_matchState_t* ms, seqStore_t* seqStore, |
| 628 | U32 rep[ZSTD_REP_NUM], |
| 629 | ZSTD_compressionParameters const* cParams, |
| 630 | const void* src, size_t srcSize, |
| 631 | const U32 searchMethod, const U32 depth) |
| 632 | { |
| 633 | const BYTE* const istart = (const BYTE*)src; |
| 634 | const BYTE* ip = istart; |
| 635 | const BYTE* anchor = istart; |
| 636 | const BYTE* const iend = istart + srcSize; |
| 637 | const BYTE* const ilimit = iend - 8; |
| 638 | const BYTE* const base = ms->window.base; |
| 639 | const U32 dictLimit = ms->window.dictLimit; |
| 640 | const U32 lowestIndex = ms->window.lowLimit; |
| 641 | const BYTE* const prefixStart = base + dictLimit; |
| 642 | const BYTE* const dictBase = ms->window.dictBase; |
| 643 | const BYTE* const dictEnd = dictBase + dictLimit; |
| 644 | const BYTE* const dictStart = dictBase + lowestIndex; |
| 645 | |
| 646 | typedef size_t (*searchMax_f)( |
| 647 | ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams, |
| 648 | const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr); |
| 649 | searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS; |
| 650 | |
| 651 | U32 offset_1 = rep[0], offset_2 = rep[1]; |
| 652 | |
| 653 | /* init */ |
| 654 | ms->nextToUpdate3 = ms->nextToUpdate; |
| 655 | ip += (ip == prefixStart); |
| 656 | |
| 657 | /* Match Loop */ |
| 658 | while (ip < ilimit) { |
| 659 | size_t matchLength=0; |
| 660 | size_t offset=0; |
| 661 | const BYTE* start=ip+1; |
| 662 | U32 current = (U32)(ip-base); |
| 663 | |
| 664 | /* check repCode */ |
| 665 | { const U32 repIndex = (U32)(current+1 - offset_1); |
| 666 | const BYTE* const repBase = repIndex < dictLimit ? dictBase : base; |
| 667 | const BYTE* const repMatch = repBase + repIndex; |
| 668 | if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex)) /* intentional overflow */ |
| 669 | if (MEM_read32(ip+1) == MEM_read32(repMatch)) { |
| 670 | /* repcode detected we should take it */ |
| 671 | const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend; |
| 672 | matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4; |
| 673 | if (depth==0) goto _storeSequence; |
| 674 | } } |
| 675 | |
| 676 | /* first search (depth 0) */ |
| 677 | { size_t offsetFound = 99999999; |
| 678 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound); |
| 679 | if (ml2 > matchLength) |
| 680 | matchLength = ml2, start = ip, offset=offsetFound; |
| 681 | } |
| 682 | |
| 683 | if (matchLength < 4) { |
| 684 | ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */ |
| 685 | continue; |
| 686 | } |
| 687 | |
| 688 | /* let's try to find a better solution */ |
| 689 | if (depth>=1) |
| 690 | while (ip<ilimit) { |
| 691 | ip ++; |
| 692 | current++; |
| 693 | /* check repCode */ |
| 694 | if (offset) { |
| 695 | const U32 repIndex = (U32)(current - offset_1); |
| 696 | const BYTE* const repBase = repIndex < dictLimit ? dictBase : base; |
| 697 | const BYTE* const repMatch = repBase + repIndex; |
| 698 | if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex)) /* intentional overflow */ |
| 699 | if (MEM_read32(ip) == MEM_read32(repMatch)) { |
| 700 | /* repcode detected */ |
| 701 | const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend; |
| 702 | size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4; |
| 703 | int const gain2 = (int)(repLength * 3); |
| 704 | int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1); |
| 705 | if ((repLength >= 4) && (gain2 > gain1)) |
| 706 | matchLength = repLength, offset = 0, start = ip; |
| 707 | } } |
| 708 | |
| 709 | /* search match, depth 1 */ |
| 710 | { size_t offset2=99999999; |
| 711 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2); |
| 712 | int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */ |
| 713 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4); |
| 714 | if ((ml2 >= 4) && (gain2 > gain1)) { |
| 715 | matchLength = ml2, offset = offset2, start = ip; |
| 716 | continue; /* search a better one */ |
| 717 | } } |
| 718 | |
| 719 | /* let's find an even better one */ |
| 720 | if ((depth==2) && (ip<ilimit)) { |
| 721 | ip ++; |
| 722 | current++; |
| 723 | /* check repCode */ |
| 724 | if (offset) { |
| 725 | const U32 repIndex = (U32)(current - offset_1); |
| 726 | const BYTE* const repBase = repIndex < dictLimit ? dictBase : base; |
| 727 | const BYTE* const repMatch = repBase + repIndex; |
| 728 | if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex)) /* intentional overflow */ |
| 729 | if (MEM_read32(ip) == MEM_read32(repMatch)) { |
| 730 | /* repcode detected */ |
| 731 | const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend; |
| 732 | size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4; |
| 733 | int const gain2 = (int)(repLength * 4); |
| 734 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1); |
| 735 | if ((repLength >= 4) && (gain2 > gain1)) |
| 736 | matchLength = repLength, offset = 0, start = ip; |
| 737 | } } |
| 738 | |
| 739 | /* search match, depth 2 */ |
| 740 | { size_t offset2=99999999; |
| 741 | size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2); |
| 742 | int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */ |
| 743 | int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7); |
| 744 | if ((ml2 >= 4) && (gain2 > gain1)) { |
| 745 | matchLength = ml2, offset = offset2, start = ip; |
| 746 | continue; |
| 747 | } } } |
| 748 | break; /* nothing found : store previous solution */ |
| 749 | } |
| 750 | |
| 751 | /* catch up */ |
| 752 | if (offset) { |
| 753 | U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE)); |
| 754 | const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex; |
| 755 | const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart; |
| 756 | while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */ |
| 757 | offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE); |
| 758 | } |
| 759 | |
| 760 | /* store sequence */ |
| 761 | _storeSequence: |
| 762 | { size_t const litLength = start - anchor; |
| 763 | ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH); |
| 764 | anchor = ip = start + matchLength; |
| 765 | } |
| 766 | |
| 767 | /* check immediate repcode */ |
| 768 | while (ip <= ilimit) { |
| 769 | const U32 repIndex = (U32)((ip-base) - offset_2); |
| 770 | const BYTE* const repBase = repIndex < dictLimit ? dictBase : base; |
| 771 | const BYTE* const repMatch = repBase + repIndex; |
| 772 | if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex)) /* intentional overflow */ |
| 773 | if (MEM_read32(ip) == MEM_read32(repMatch)) { |
| 774 | /* repcode detected we should take it */ |
| 775 | const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend; |
| 776 | matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4; |
| 777 | offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */ |
| 778 | ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH); |
| 779 | ip += matchLength; |
| 780 | anchor = ip; |
| 781 | continue; /* faster when present ... (?) */ |
| 782 | } |
| 783 | break; |
| 784 | } } |
| 785 | |
| 786 | /* Save reps for next block */ |
| 787 | rep[0] = offset_1; |
| 788 | rep[1] = offset_2; |
| 789 | |
| 790 | /* Return the last literals size */ |
| 791 | return iend - anchor; |
| 792 | } |
| 793 | |
| 794 | |
| 795 | size_t ZSTD_compressBlock_greedy_extDict( |
| 796 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 797 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 798 | { |
| 799 | return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0); |
| 800 | } |
| 801 | |
| 802 | size_t ZSTD_compressBlock_lazy_extDict( |
| 803 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 804 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 805 | |
| 806 | { |
| 807 | return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1); |
| 808 | } |
| 809 | |
| 810 | size_t ZSTD_compressBlock_lazy2_extDict( |
| 811 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 812 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 813 | |
| 814 | { |
| 815 | return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2); |
| 816 | } |
| 817 | |
| 818 | size_t ZSTD_compressBlock_btlazy2_extDict( |
| 819 | ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], |
| 820 | ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize) |
| 821 | |
| 822 | { |
| 823 | return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2); |
| 824 | } |