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_double_fast.h" 13 14 15 void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms, 16 void const* end, ZSTD_dictTableLoadMethod_e dtlm) 17 { 18 const ZSTD_compressionParameters* const cParams = &ms->cParams; 19 U32* const hashLarge = ms->hashTable; 20 U32 const hBitsL = cParams->hashLog; 21 U32 const mls = cParams->minMatch; 22 U32* const hashSmall = ms->chainTable; 23 U32 const hBitsS = cParams->chainLog; 24 const BYTE* const base = ms->window.base; 25 const BYTE* ip = base + ms->nextToUpdate; 26 const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE; 27 const U32 fastHashFillStep = 3; 28 29 /* Always insert every fastHashFillStep position into the hash tables. 30 * Insert the other positions into the large hash table if their entry 31 * is empty. 32 */ 33 for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) { 34 U32 const current = (U32)(ip - base); 35 U32 i; 36 for (i = 0; i < fastHashFillStep; ++i) { 37 size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls); 38 size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8); 39 if (i == 0) 40 hashSmall[smHash] = current + i; 41 if (i == 0 || hashLarge[lgHash] == 0) 42 hashLarge[lgHash] = current + i; 43 /* Only load extra positions for ZSTD_dtlm_full */ 44 if (dtlm == ZSTD_dtlm_fast) 45 break; 46 } } 47 } 48 49 50 FORCE_INLINE_TEMPLATE 51 size_t ZSTD_compressBlock_doubleFast_generic( 52 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 53 void const* src, size_t srcSize, 54 U32 const mls /* template */, ZSTD_dictMode_e const dictMode) 55 { 56 ZSTD_compressionParameters const* cParams = &ms->cParams; 57 U32* const hashLong = ms->hashTable; 58 const U32 hBitsL = cParams->hashLog; 59 U32* const hashSmall = ms->chainTable; 60 const U32 hBitsS = cParams->chainLog; 61 const BYTE* const base = ms->window.base; 62 const BYTE* const istart = (const BYTE*)src; 63 const BYTE* ip = istart; 64 const BYTE* anchor = istart; 65 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize); 66 const U32 lowestValid = ms->window.dictLimit; 67 const U32 maxDistance = 1U << cParams->windowLog; 68 /* presumes that, if there is a dictionary, it must be using Attach mode */ 69 const U32 prefixLowestIndex = (endIndex - lowestValid > maxDistance) ? endIndex - maxDistance : lowestValid; 70 const BYTE* const prefixLowest = base + prefixLowestIndex; 71 const BYTE* const iend = istart + srcSize; 72 const BYTE* const ilimit = iend - HASH_READ_SIZE; 73 U32 offset_1=rep[0], offset_2=rep[1]; 74 U32 offsetSaved = 0; 75 76 const ZSTD_matchState_t* const dms = ms->dictMatchState; 77 const ZSTD_compressionParameters* const dictCParams = 78 dictMode == ZSTD_dictMatchState ? 79 &dms->cParams : NULL; 80 const U32* const dictHashLong = dictMode == ZSTD_dictMatchState ? 81 dms->hashTable : NULL; 82 const U32* const dictHashSmall = dictMode == ZSTD_dictMatchState ? 83 dms->chainTable : NULL; 84 const U32 dictStartIndex = dictMode == ZSTD_dictMatchState ? 85 dms->window.dictLimit : 0; 86 const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ? 87 dms->window.base : NULL; 88 const BYTE* const dictStart = dictMode == ZSTD_dictMatchState ? 89 dictBase + dictStartIndex : NULL; 90 const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ? 91 dms->window.nextSrc : NULL; 92 const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ? 93 prefixLowestIndex - (U32)(dictEnd - dictBase) : 94 0; 95 const U32 dictHBitsL = dictMode == ZSTD_dictMatchState ? 96 dictCParams->hashLog : hBitsL; 97 const U32 dictHBitsS = dictMode == ZSTD_dictMatchState ? 98 dictCParams->chainLog : hBitsS; 99 const U32 dictAndPrefixLength = (U32)(ip - prefixLowest + dictEnd - dictStart); 100 101 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_generic"); 102 103 assert(dictMode == ZSTD_noDict || dictMode == ZSTD_dictMatchState); 104 105 /* if a dictionary is attached, it must be within window range */ 106 if (dictMode == ZSTD_dictMatchState) { 107 assert(lowestValid + maxDistance >= endIndex); 108 } 109 110 /* init */ 111 ip += (dictAndPrefixLength == 0); 112 if (dictMode == ZSTD_noDict) { 113 U32 const maxRep = (U32)(ip - prefixLowest); 114 if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0; 115 if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0; 116 } 117 if (dictMode == ZSTD_dictMatchState) { 118 /* dictMatchState repCode checks don't currently handle repCode == 0 119 * disabling. */ 120 assert(offset_1 <= dictAndPrefixLength); 121 assert(offset_2 <= dictAndPrefixLength); 122 } 123 124 /* Main Search Loop */ 125 while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */ 126 size_t mLength; 127 U32 offset; 128 size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8); 129 size_t const h = ZSTD_hashPtr(ip, hBitsS, mls); 130 size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8); 131 size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls); 132 U32 const current = (U32)(ip-base); 133 U32 const matchIndexL = hashLong[h2]; 134 U32 matchIndexS = hashSmall[h]; 135 const BYTE* matchLong = base + matchIndexL; 136 const BYTE* match = base + matchIndexS; 137 const U32 repIndex = current + 1 - offset_1; 138 const BYTE* repMatch = (dictMode == ZSTD_dictMatchState 139 && repIndex < prefixLowestIndex) ? 140 dictBase + (repIndex - dictIndexDelta) : 141 base + repIndex; 142 hashLong[h2] = hashSmall[h] = current; /* update hash tables */ 143 144 /* check dictMatchState repcode */ 145 if (dictMode == ZSTD_dictMatchState 146 && ((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */) 147 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) { 148 const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend; 149 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4; 150 ip++; 151 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH); 152 goto _match_stored; 153 } 154 155 /* check noDict repcode */ 156 if ( dictMode == ZSTD_noDict 157 && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) { 158 mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4; 159 ip++; 160 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH); 161 goto _match_stored; 162 } 163 164 if (matchIndexL > prefixLowestIndex) { 165 /* check prefix long match */ 166 if (MEM_read64(matchLong) == MEM_read64(ip)) { 167 mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8; 168 offset = (U32)(ip-matchLong); 169 while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */ 170 goto _match_found; 171 } 172 } else if (dictMode == ZSTD_dictMatchState) { 173 /* check dictMatchState long match */ 174 U32 const dictMatchIndexL = dictHashLong[dictHL]; 175 const BYTE* dictMatchL = dictBase + dictMatchIndexL; 176 assert(dictMatchL < dictEnd); 177 178 if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) { 179 mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8; 180 offset = (U32)(current - dictMatchIndexL - dictIndexDelta); 181 while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */ 182 goto _match_found; 183 } } 184 185 if (matchIndexS > prefixLowestIndex) { 186 /* check prefix short match */ 187 if (MEM_read32(match) == MEM_read32(ip)) { 188 goto _search_next_long; 189 } 190 } else if (dictMode == ZSTD_dictMatchState) { 191 /* check dictMatchState short match */ 192 U32 const dictMatchIndexS = dictHashSmall[dictHS]; 193 match = dictBase + dictMatchIndexS; 194 matchIndexS = dictMatchIndexS + dictIndexDelta; 195 196 if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) { 197 goto _search_next_long; 198 } } 199 200 ip += ((ip-anchor) >> kSearchStrength) + 1; 201 continue; 202 203 _search_next_long: 204 205 { size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8); 206 size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8); 207 U32 const matchIndexL3 = hashLong[hl3]; 208 const BYTE* matchL3 = base + matchIndexL3; 209 hashLong[hl3] = current + 1; 210 211 /* check prefix long +1 match */ 212 if (matchIndexL3 > prefixLowestIndex) { 213 if (MEM_read64(matchL3) == MEM_read64(ip+1)) { 214 mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8; 215 ip++; 216 offset = (U32)(ip-matchL3); 217 while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */ 218 goto _match_found; 219 } 220 } else if (dictMode == ZSTD_dictMatchState) { 221 /* check dict long +1 match */ 222 U32 const dictMatchIndexL3 = dictHashLong[dictHLNext]; 223 const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3; 224 assert(dictMatchL3 < dictEnd); 225 if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) { 226 mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8; 227 ip++; 228 offset = (U32)(current + 1 - dictMatchIndexL3 - dictIndexDelta); 229 while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */ 230 goto _match_found; 231 } } } 232 233 /* if no long +1 match, explore the short match we found */ 234 if (dictMode == ZSTD_dictMatchState && matchIndexS < prefixLowestIndex) { 235 mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4; 236 offset = (U32)(current - matchIndexS); 237 while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 238 } else { 239 mLength = ZSTD_count(ip+4, match+4, iend) + 4; 240 offset = (U32)(ip - match); 241 while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 242 } 243 244 /* fall-through */ 245 246 _match_found: 247 offset_2 = offset_1; 248 offset_1 = offset; 249 250 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 251 252 _match_stored: 253 /* match found */ 254 ip += mLength; 255 anchor = ip; 256 257 if (ip <= ilimit) { 258 /* Complementary insertion */ 259 /* done after iLimit test, as candidates could be > iend-8 */ 260 { U32 const indexToInsert = current+2; 261 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert; 262 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base); 263 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert; 264 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base); 265 } 266 267 /* check immediate repcode */ 268 if (dictMode == ZSTD_dictMatchState) { 269 while (ip <= ilimit) { 270 U32 const current2 = (U32)(ip-base); 271 U32 const repIndex2 = current2 - offset_2; 272 const BYTE* repMatch2 = dictMode == ZSTD_dictMatchState 273 && repIndex2 < prefixLowestIndex ? 274 dictBase - dictIndexDelta + repIndex2 : 275 base + repIndex2; 276 if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */) 277 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) { 278 const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend; 279 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4; 280 U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */ 281 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH); 282 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2; 283 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2; 284 ip += repLength2; 285 anchor = ip; 286 continue; 287 } 288 break; 289 } } 290 291 if (dictMode == ZSTD_noDict) { 292 while ( (ip <= ilimit) 293 && ( (offset_2>0) 294 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) { 295 /* store sequence */ 296 size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4; 297 U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */ 298 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base); 299 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base); 300 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, rLength-MINMATCH); 301 ip += rLength; 302 anchor = ip; 303 continue; /* faster when present ... (?) */ 304 } } } 305 } /* while (ip < ilimit) */ 306 307 /* save reps for next block */ 308 rep[0] = offset_1 ? offset_1 : offsetSaved; 309 rep[1] = offset_2 ? offset_2 : offsetSaved; 310 311 /* Return the last literals size */ 312 return (size_t)(iend - anchor); 313 } 314 315 316 size_t ZSTD_compressBlock_doubleFast( 317 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 318 void const* src, size_t srcSize) 319 { 320 const U32 mls = ms->cParams.minMatch; 321 switch(mls) 322 { 323 default: /* includes case 3 */ 324 case 4 : 325 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_noDict); 326 case 5 : 327 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_noDict); 328 case 6 : 329 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_noDict); 330 case 7 : 331 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_noDict); 332 } 333 } 334 335 336 size_t ZSTD_compressBlock_doubleFast_dictMatchState( 337 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 338 void const* src, size_t srcSize) 339 { 340 const U32 mls = ms->cParams.minMatch; 341 switch(mls) 342 { 343 default: /* includes case 3 */ 344 case 4 : 345 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_dictMatchState); 346 case 5 : 347 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_dictMatchState); 348 case 6 : 349 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_dictMatchState); 350 case 7 : 351 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_dictMatchState); 352 } 353 } 354 355 356 static size_t ZSTD_compressBlock_doubleFast_extDict_generic( 357 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 358 void const* src, size_t srcSize, 359 U32 const mls /* template */) 360 { 361 ZSTD_compressionParameters const* cParams = &ms->cParams; 362 U32* const hashLong = ms->hashTable; 363 U32 const hBitsL = cParams->hashLog; 364 U32* const hashSmall = ms->chainTable; 365 U32 const hBitsS = cParams->chainLog; 366 const BYTE* const istart = (const BYTE*)src; 367 const BYTE* ip = istart; 368 const BYTE* anchor = istart; 369 const BYTE* const iend = istart + srcSize; 370 const BYTE* const ilimit = iend - 8; 371 const BYTE* const base = ms->window.base; 372 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize); 373 const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog); 374 const U32 dictStartIndex = lowLimit; 375 const U32 dictLimit = ms->window.dictLimit; 376 const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit; 377 const BYTE* const prefixStart = base + prefixStartIndex; 378 const BYTE* const dictBase = ms->window.dictBase; 379 const BYTE* const dictStart = dictBase + dictStartIndex; 380 const BYTE* const dictEnd = dictBase + prefixStartIndex; 381 U32 offset_1=rep[0], offset_2=rep[1]; 382 383 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize); 384 385 /* if extDict is invalidated due to maxDistance, switch to "regular" variant */ 386 if (prefixStartIndex == dictStartIndex) 387 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, mls, ZSTD_noDict); 388 389 /* Search Loop */ 390 while (ip < ilimit) { /* < instead of <=, because (ip+1) */ 391 const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls); 392 const U32 matchIndex = hashSmall[hSmall]; 393 const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base; 394 const BYTE* match = matchBase + matchIndex; 395 396 const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8); 397 const U32 matchLongIndex = hashLong[hLong]; 398 const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base; 399 const BYTE* matchLong = matchLongBase + matchLongIndex; 400 401 const U32 current = (U32)(ip-base); 402 const U32 repIndex = current + 1 - offset_1; /* offset_1 expected <= current +1 */ 403 const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base; 404 const BYTE* const repMatch = repBase + repIndex; 405 size_t mLength; 406 hashSmall[hSmall] = hashLong[hLong] = current; /* update hash table */ 407 408 if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */ 409 & (repIndex > dictStartIndex)) 410 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) { 411 const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend; 412 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4; 413 ip++; 414 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH); 415 } else { 416 if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) { 417 const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend; 418 const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart; 419 U32 offset; 420 mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8; 421 offset = current - matchLongIndex; 422 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */ 423 offset_2 = offset_1; 424 offset_1 = offset; 425 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 426 427 } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) { 428 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8); 429 U32 const matchIndex3 = hashLong[h3]; 430 const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base; 431 const BYTE* match3 = match3Base + matchIndex3; 432 U32 offset; 433 hashLong[h3] = current + 1; 434 if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) { 435 const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend; 436 const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart; 437 mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8; 438 ip++; 439 offset = current+1 - matchIndex3; 440 while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */ 441 } else { 442 const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend; 443 const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart; 444 mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4; 445 offset = current - matchIndex; 446 while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 447 } 448 offset_2 = offset_1; 449 offset_1 = offset; 450 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 451 452 } else { 453 ip += ((ip-anchor) >> kSearchStrength) + 1; 454 continue; 455 } } 456 457 /* move to next sequence start */ 458 ip += mLength; 459 anchor = ip; 460 461 if (ip <= ilimit) { 462 /* Complementary insertion */ 463 /* done after iLimit test, as candidates could be > iend-8 */ 464 { U32 const indexToInsert = current+2; 465 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert; 466 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base); 467 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert; 468 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base); 469 } 470 471 /* check immediate repcode */ 472 while (ip <= ilimit) { 473 U32 const current2 = (U32)(ip-base); 474 U32 const repIndex2 = current2 - offset_2; 475 const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2; 476 if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */ 477 & (repIndex2 > dictStartIndex)) 478 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) { 479 const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend; 480 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4; 481 U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */ 482 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH); 483 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2; 484 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2; 485 ip += repLength2; 486 anchor = ip; 487 continue; 488 } 489 break; 490 } } } 491 492 /* save reps for next block */ 493 rep[0] = offset_1; 494 rep[1] = offset_2; 495 496 /* Return the last literals size */ 497 return (size_t)(iend - anchor); 498 } 499 500 501 size_t ZSTD_compressBlock_doubleFast_extDict( 502 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 503 void const* src, size_t srcSize) 504 { 505 U32 const mls = ms->cParams.minMatch; 506 switch(mls) 507 { 508 default: /* includes case 3 */ 509 case 4 : 510 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 4); 511 case 5 : 512 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 5); 513 case 6 : 514 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 6); 515 case 7 : 516 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 7); 517 } 518 } 519