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 const U32 prefixLowestIndex = (endIndex - lowestValid > maxDistance) ? endIndex - maxDistance : lowestValid; 69 const BYTE* const prefixLowest = base + prefixLowestIndex; 70 const BYTE* const iend = istart + srcSize; 71 const BYTE* const ilimit = iend - HASH_READ_SIZE; 72 U32 offset_1=rep[0], offset_2=rep[1]; 73 U32 offsetSaved = 0; 74 75 const ZSTD_matchState_t* const dms = ms->dictMatchState; 76 const ZSTD_compressionParameters* const dictCParams = 77 dictMode == ZSTD_dictMatchState ? 78 &dms->cParams : NULL; 79 const U32* const dictHashLong = dictMode == ZSTD_dictMatchState ? 80 dms->hashTable : NULL; 81 const U32* const dictHashSmall = dictMode == ZSTD_dictMatchState ? 82 dms->chainTable : NULL; 83 const U32 dictStartIndex = dictMode == ZSTD_dictMatchState ? 84 dms->window.dictLimit : 0; 85 const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ? 86 dms->window.base : NULL; 87 const BYTE* const dictStart = dictMode == ZSTD_dictMatchState ? 88 dictBase + dictStartIndex : NULL; 89 const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ? 90 dms->window.nextSrc : NULL; 91 const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ? 92 prefixLowestIndex - (U32)(dictEnd - dictBase) : 93 0; 94 const U32 dictHBitsL = dictMode == ZSTD_dictMatchState ? 95 dictCParams->hashLog : hBitsL; 96 const U32 dictHBitsS = dictMode == ZSTD_dictMatchState ? 97 dictCParams->chainLog : hBitsS; 98 const U32 dictAndPrefixLength = (U32)(ip - prefixLowest + dictEnd - dictStart); 99 100 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_generic"); 101 102 assert(dictMode == ZSTD_noDict || dictMode == ZSTD_dictMatchState); 103 104 /* if a dictionary is attached, it must be within window range */ 105 if (dictMode == ZSTD_dictMatchState) { 106 assert(lowestValid + maxDistance >= endIndex); 107 } 108 109 /* init */ 110 ip += (dictAndPrefixLength == 0); 111 if (dictMode == ZSTD_noDict) { 112 U32 const maxRep = (U32)(ip - prefixLowest); 113 if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0; 114 if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0; 115 } 116 if (dictMode == ZSTD_dictMatchState) { 117 /* dictMatchState repCode checks don't currently handle repCode == 0 118 * disabling. */ 119 assert(offset_1 <= dictAndPrefixLength); 120 assert(offset_2 <= dictAndPrefixLength); 121 } 122 123 /* Main Search Loop */ 124 while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */ 125 size_t mLength; 126 U32 offset; 127 size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8); 128 size_t const h = ZSTD_hashPtr(ip, hBitsS, mls); 129 size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8); 130 size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls); 131 U32 const current = (U32)(ip-base); 132 U32 const matchIndexL = hashLong[h2]; 133 U32 matchIndexS = hashSmall[h]; 134 const BYTE* matchLong = base + matchIndexL; 135 const BYTE* match = base + matchIndexS; 136 const U32 repIndex = current + 1 - offset_1; 137 const BYTE* repMatch = (dictMode == ZSTD_dictMatchState 138 && repIndex < prefixLowestIndex) ? 139 dictBase + (repIndex - dictIndexDelta) : 140 base + repIndex; 141 hashLong[h2] = hashSmall[h] = current; /* update hash tables */ 142 143 /* check dictMatchState repcode */ 144 if (dictMode == ZSTD_dictMatchState 145 && ((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */) 146 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) { 147 const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend; 148 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4; 149 ip++; 150 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH); 151 goto _match_stored; 152 } 153 154 /* check noDict repcode */ 155 if ( dictMode == ZSTD_noDict 156 && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) { 157 mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4; 158 ip++; 159 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH); 160 goto _match_stored; 161 } 162 163 if (matchIndexL > prefixLowestIndex) { 164 /* check prefix long match */ 165 if (MEM_read64(matchLong) == MEM_read64(ip)) { 166 mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8; 167 offset = (U32)(ip-matchLong); 168 while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */ 169 goto _match_found; 170 } 171 } else if (dictMode == ZSTD_dictMatchState) { 172 /* check dictMatchState long match */ 173 U32 const dictMatchIndexL = dictHashLong[dictHL]; 174 const BYTE* dictMatchL = dictBase + dictMatchIndexL; 175 assert(dictMatchL < dictEnd); 176 177 if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) { 178 mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8; 179 offset = (U32)(current - dictMatchIndexL - dictIndexDelta); 180 while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */ 181 goto _match_found; 182 } } 183 184 if (matchIndexS > prefixLowestIndex) { 185 /* check prefix short match */ 186 if (MEM_read32(match) == MEM_read32(ip)) { 187 goto _search_next_long; 188 } 189 } else if (dictMode == ZSTD_dictMatchState) { 190 /* check dictMatchState short match */ 191 U32 const dictMatchIndexS = dictHashSmall[dictHS]; 192 match = dictBase + dictMatchIndexS; 193 matchIndexS = dictMatchIndexS + dictIndexDelta; 194 195 if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) { 196 goto _search_next_long; 197 } } 198 199 ip += ((ip-anchor) >> kSearchStrength) + 1; 200 continue; 201 202 _search_next_long: 203 204 { size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8); 205 size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8); 206 U32 const matchIndexL3 = hashLong[hl3]; 207 const BYTE* matchL3 = base + matchIndexL3; 208 hashLong[hl3] = current + 1; 209 210 /* check prefix long +1 match */ 211 if (matchIndexL3 > prefixLowestIndex) { 212 if (MEM_read64(matchL3) == MEM_read64(ip+1)) { 213 mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8; 214 ip++; 215 offset = (U32)(ip-matchL3); 216 while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */ 217 goto _match_found; 218 } 219 } else if (dictMode == ZSTD_dictMatchState) { 220 /* check dict long +1 match */ 221 U32 const dictMatchIndexL3 = dictHashLong[dictHLNext]; 222 const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3; 223 assert(dictMatchL3 < dictEnd); 224 if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) { 225 mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8; 226 ip++; 227 offset = (U32)(current + 1 - dictMatchIndexL3 - dictIndexDelta); 228 while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */ 229 goto _match_found; 230 } } } 231 232 /* if no long +1 match, explore the short match we found */ 233 if (dictMode == ZSTD_dictMatchState && matchIndexS < prefixLowestIndex) { 234 mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4; 235 offset = (U32)(current - matchIndexS); 236 while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 237 } else { 238 mLength = ZSTD_count(ip+4, match+4, iend) + 4; 239 offset = (U32)(ip - match); 240 while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 241 } 242 243 /* fall-through */ 244 245 _match_found: 246 offset_2 = offset_1; 247 offset_1 = offset; 248 249 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 250 251 _match_stored: 252 /* match found */ 253 ip += mLength; 254 anchor = ip; 255 256 if (ip <= ilimit) { 257 /* Complementary insertion */ 258 /* done after iLimit test, as candidates could be > iend-8 */ 259 { U32 const indexToInsert = current+2; 260 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert; 261 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base); 262 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert; 263 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base); 264 } 265 266 /* check immediate repcode */ 267 if (dictMode == ZSTD_dictMatchState) { 268 while (ip <= ilimit) { 269 U32 const current2 = (U32)(ip-base); 270 U32 const repIndex2 = current2 - offset_2; 271 const BYTE* repMatch2 = dictMode == ZSTD_dictMatchState 272 && repIndex2 < prefixLowestIndex ? 273 dictBase - dictIndexDelta + repIndex2 : 274 base + repIndex2; 275 if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */) 276 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) { 277 const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend; 278 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4; 279 U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */ 280 ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH); 281 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2; 282 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2; 283 ip += repLength2; 284 anchor = ip; 285 continue; 286 } 287 break; 288 } } 289 290 if (dictMode == ZSTD_noDict) { 291 while ( (ip <= ilimit) 292 && ( (offset_2>0) 293 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) { 294 /* store sequence */ 295 size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4; 296 U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */ 297 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base); 298 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base); 299 ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH); 300 ip += rLength; 301 anchor = ip; 302 continue; /* faster when present ... (?) */ 303 } } } 304 } /* while (ip < ilimit) */ 305 306 /* save reps for next block */ 307 rep[0] = offset_1 ? offset_1 : offsetSaved; 308 rep[1] = offset_2 ? offset_2 : offsetSaved; 309 310 /* Return the last literals size */ 311 return (size_t)(iend - anchor); 312 } 313 314 315 size_t ZSTD_compressBlock_doubleFast( 316 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 317 void const* src, size_t srcSize) 318 { 319 const U32 mls = ms->cParams.minMatch; 320 switch(mls) 321 { 322 default: /* includes case 3 */ 323 case 4 : 324 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_noDict); 325 case 5 : 326 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_noDict); 327 case 6 : 328 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_noDict); 329 case 7 : 330 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_noDict); 331 } 332 } 333 334 335 size_t ZSTD_compressBlock_doubleFast_dictMatchState( 336 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 337 void const* src, size_t srcSize) 338 { 339 const U32 mls = ms->cParams.minMatch; 340 switch(mls) 341 { 342 default: /* includes case 3 */ 343 case 4 : 344 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_dictMatchState); 345 case 5 : 346 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_dictMatchState); 347 case 6 : 348 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_dictMatchState); 349 case 7 : 350 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_dictMatchState); 351 } 352 } 353 354 355 static size_t ZSTD_compressBlock_doubleFast_extDict_generic( 356 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 357 void const* src, size_t srcSize, 358 U32 const mls /* template */) 359 { 360 ZSTD_compressionParameters const* cParams = &ms->cParams; 361 U32* const hashLong = ms->hashTable; 362 U32 const hBitsL = cParams->hashLog; 363 U32* const hashSmall = ms->chainTable; 364 U32 const hBitsS = cParams->chainLog; 365 const BYTE* const istart = (const BYTE*)src; 366 const BYTE* ip = istart; 367 const BYTE* anchor = istart; 368 const BYTE* const iend = istart + srcSize; 369 const BYTE* const ilimit = iend - 8; 370 const BYTE* const base = ms->window.base; 371 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize); 372 const U32 maxDistance = 1U << cParams->windowLog; 373 const U32 lowestValid = ms->window.lowLimit; 374 const U32 lowLimit = (endIndex - lowestValid > maxDistance) ? endIndex - maxDistance : lowestValid; 375 const U32 dictStartIndex = lowLimit; 376 const U32 dictLimit = ms->window.dictLimit; 377 const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit; 378 const BYTE* const prefixStart = base + prefixStartIndex; 379 const BYTE* const dictBase = ms->window.dictBase; 380 const BYTE* const dictStart = dictBase + dictStartIndex; 381 const BYTE* const dictEnd = dictBase + prefixStartIndex; 382 U32 offset_1=rep[0], offset_2=rep[1]; 383 384 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize); 385 386 /* if extDict is invalidated due to maxDistance, switch to "regular" variant */ 387 if (prefixStartIndex == dictStartIndex) 388 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, mls, ZSTD_noDict); 389 390 /* Search Loop */ 391 while (ip < ilimit) { /* < instead of <=, because (ip+1) */ 392 const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls); 393 const U32 matchIndex = hashSmall[hSmall]; 394 const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base; 395 const BYTE* match = matchBase + matchIndex; 396 397 const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8); 398 const U32 matchLongIndex = hashLong[hLong]; 399 const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base; 400 const BYTE* matchLong = matchLongBase + matchLongIndex; 401 402 const U32 current = (U32)(ip-base); 403 const U32 repIndex = current + 1 - offset_1; /* offset_1 expected <= current +1 */ 404 const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base; 405 const BYTE* const repMatch = repBase + repIndex; 406 size_t mLength; 407 hashSmall[hSmall] = hashLong[hLong] = current; /* update hash table */ 408 409 if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */ 410 & (repIndex > dictStartIndex)) 411 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) { 412 const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend; 413 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4; 414 ip++; 415 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH); 416 } else { 417 if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) { 418 const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend; 419 const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart; 420 U32 offset; 421 mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8; 422 offset = current - matchLongIndex; 423 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */ 424 offset_2 = offset_1; 425 offset_1 = offset; 426 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 427 428 } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) { 429 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8); 430 U32 const matchIndex3 = hashLong[h3]; 431 const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base; 432 const BYTE* match3 = match3Base + matchIndex3; 433 U32 offset; 434 hashLong[h3] = current + 1; 435 if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) { 436 const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend; 437 const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart; 438 mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8; 439 ip++; 440 offset = current+1 - matchIndex3; 441 while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */ 442 } else { 443 const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend; 444 const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart; 445 mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4; 446 offset = current - matchIndex; 447 while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */ 448 } 449 offset_2 = offset_1; 450 offset_1 = offset; 451 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH); 452 453 } else { 454 ip += ((ip-anchor) >> kSearchStrength) + 1; 455 continue; 456 } } 457 458 /* move to next sequence start */ 459 ip += mLength; 460 anchor = ip; 461 462 if (ip <= ilimit) { 463 /* Complementary insertion */ 464 /* done after iLimit test, as candidates could be > iend-8 */ 465 { U32 const indexToInsert = current+2; 466 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert; 467 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base); 468 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert; 469 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base); 470 } 471 472 /* check immediate repcode */ 473 while (ip <= ilimit) { 474 U32 const current2 = (U32)(ip-base); 475 U32 const repIndex2 = current2 - offset_2; 476 const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2; 477 if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */ 478 & (repIndex2 > dictStartIndex)) 479 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) { 480 const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend; 481 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4; 482 U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */ 483 ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH); 484 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2; 485 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2; 486 ip += repLength2; 487 anchor = ip; 488 continue; 489 } 490 break; 491 } } } 492 493 /* save reps for next block */ 494 rep[0] = offset_1; 495 rep[1] = offset_2; 496 497 /* Return the last literals size */ 498 return (size_t)(iend - anchor); 499 } 500 501 502 size_t ZSTD_compressBlock_doubleFast_extDict( 503 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], 504 void const* src, size_t srcSize) 505 { 506 U32 const mls = ms->cParams.minMatch; 507 switch(mls) 508 { 509 default: /* includes case 3 */ 510 case 4 : 511 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 4); 512 case 5 : 513 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 5); 514 case 6 : 515 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 6); 516 case 7 : 517 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 7); 518 } 519 } 520