diff options
Diffstat (limited to 'Utilities/cmzstd/lib/compress/zstd_ldm.c')
-rw-r--r-- | Utilities/cmzstd/lib/compress/zstd_ldm.c | 44 |
1 files changed, 33 insertions, 11 deletions
diff --git a/Utilities/cmzstd/lib/compress/zstd_ldm.c b/Utilities/cmzstd/lib/compress/zstd_ldm.c index 58eb2ff..8c47948 100644 --- a/Utilities/cmzstd/lib/compress/zstd_ldm.c +++ b/Utilities/cmzstd/lib/compress/zstd_ldm.c @@ -1,15 +1,16 @@ /* - * Copyright (c) 2016-present, Yann Collet, Facebook, Inc. + * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc. * All rights reserved. * * This source code is licensed under both the BSD-style license (found in the * LICENSE file in the root directory of this source tree) and the GPLv2 (found * in the COPYING file in the root directory of this source tree). + * You may select, at your option, one of the above-listed licenses. */ #include "zstd_ldm.h" -#include "debug.h" +#include "../common/debug.h" #include "zstd_fast.h" /* ZSTD_fillHashTable() */ #include "zstd_double_fast.h" /* ZSTD_fillDoubleHashTable() */ @@ -49,9 +50,9 @@ size_t ZSTD_ldm_getTableSize(ldmParams_t params) { size_t const ldmHSize = ((size_t)1) << params.hashLog; size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog); - size_t const ldmBucketSize = - ((size_t)1) << (params.hashLog - ldmBucketSizeLog); - size_t const totalSize = ldmBucketSize + ldmHSize * sizeof(ldmEntry_t); + size_t const ldmBucketSize = ((size_t)1) << (params.hashLog - ldmBucketSizeLog); + size_t const totalSize = ZSTD_cwksp_alloc_size(ldmBucketSize) + + ZSTD_cwksp_alloc_size(ldmHSize * sizeof(ldmEntry_t)); return params.enableLdm ? totalSize : 0; } @@ -223,6 +224,20 @@ static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state, return rollingHash; } +void ZSTD_ldm_fillHashTable( + ldmState_t* state, const BYTE* ip, + const BYTE* iend, ldmParams_t const* params) +{ + DEBUGLOG(5, "ZSTD_ldm_fillHashTable"); + if ((size_t)(iend - ip) >= params->minMatchLength) { + U64 startingHash = ZSTD_rollingHash_compute(ip, params->minMatchLength); + ZSTD_ldm_fillLdmHashTable( + state, startingHash, ip, iend - params->minMatchLength, state->window.base, + params->hashLog - params->bucketSizeLog, + *params); + } +} + /** ZSTD_ldm_limitTableUpdate() : * @@ -429,7 +444,7 @@ size_t ZSTD_ldm_generateSequences( */ assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize); /* The input could be very large (in zstdmt), so it must be broken up into - * chunks to enforce the maximmum distance and handle overflow correction. + * chunks to enforce the maximum distance and handle overflow correction. */ assert(sequences->pos <= sequences->size); assert(sequences->size <= sequences->capacity); @@ -447,8 +462,10 @@ size_t ZSTD_ldm_generateSequences( if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) { U32 const ldmHSize = 1U << params->hashLog; U32 const correction = ZSTD_window_correctOverflow( - &ldmState->window, /* cycleLog */ 0, maxDist, src); + &ldmState->window, /* cycleLog */ 0, maxDist, chunkStart); ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction); + /* invalidate dictionaries on overflow correction */ + ldmState->loadedDictEnd = 0; } /* 2. We enforce the maximum offset allowed. * @@ -457,8 +474,14 @@ size_t ZSTD_ldm_generateSequences( * TODO: * Test the chunk size. * * Try invalidation after the sequence generation and test the * the offset against maxDist directly. + * + * NOTE: Because of dictionaries + sequence splitting we MUST make sure + * that any offset used is valid at the END of the sequence, since it may + * be split into two sequences. This condition holds when using + * ZSTD_window_enforceMaxDist(), but if we move to checking offsets + * against maxDist directly, we'll have to carefully handle that case. */ - ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, NULL, NULL); + ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, &ldmState->loadedDictEnd, NULL); /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */ newLeftoverSize = ZSTD_ldm_generateSequences_internal( ldmState, sequences, params, chunkStart, chunkSize); @@ -566,14 +589,13 @@ size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore, if (sequence.offset == 0) break; - assert(sequence.offset <= (1U << cParams->windowLog)); assert(ip + sequence.litLength + sequence.matchLength <= iend); /* Fill tables for block compressor */ ZSTD_ldm_limitTableUpdate(ms, ip); ZSTD_ldm_fillFastTables(ms, ip); /* Run the block compressor */ - DEBUGLOG(5, "calling block compressor on segment of size %u", sequence.litLength); + DEBUGLOG(5, "pos %u : calling block compressor on segment of size %u", (unsigned)(ip-istart), sequence.litLength); { size_t const newLitLength = blockCompressor(ms, seqStore, rep, ip, sequence.litLength); @@ -583,7 +605,7 @@ size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore, rep[i] = rep[i-1]; rep[0] = sequence.offset; /* Store the sequence */ - ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, + ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, iend, sequence.offset + ZSTD_REP_MOVE, sequence.matchLength - MINMATCH); ip += sequence.matchLength; |