zstd_ldm.h File Reference

#include "zstd_compress_internal.h"
#include "zstd.h"

Include dependency graph for zstd_ldm.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Macros
#define	ZSTD_LDM_DEFAULT_WINDOW_LOG   ZSTD_WINDOWLOG_LIMIT_DEFAULT

Functions
void	ZSTD_ldm_fillHashTable (ldmState_t *state, const BYTE *ip, const BYTE *iend, ldmParams_t const *params)
size_t	ZSTD_ldm_generateSequences (ldmState_t *ldms, rawSeqStore_t *sequences, ldmParams_t const *params, void const *src, size_t srcSize)
size_t	ZSTD_ldm_blockCompress (rawSeqStore_t *rawSeqStore, ZSTD_matchState_t *ms, seqStore_t *seqStore, U32 rep[ZSTD_REP_NUM], void const *src, size_t srcSize)
void	ZSTD_ldm_skipSequences (rawSeqStore_t *rawSeqStore, size_t srcSize, U32 const minMatch)
size_t	ZSTD_ldm_getTableSize (ldmParams_t params)
size_t	ZSTD_ldm_getMaxNbSeq (ldmParams_t params, size_t maxChunkSize)
void	ZSTD_ldm_adjustParameters (ldmParams_t *params, ZSTD_compressionParameters const *cParams)

Macro Definition Documentation

ZSTD_LDM_DEFAULT_WINDOW_LOG

#define ZSTD_LDM_DEFAULT_WINDOW_LOG   ZSTD_WINDOWLOG_LIMIT_DEFAULT

Definition at line 25 of file zstd_ldm.h.

Function Documentation

ZSTD_ldm_adjustParameters()

void ZSTD_ldm_adjustParameters	(	ldmParams_t *	params,
		ZSTD_compressionParameters const *	cParams
	)

ZSTD_ldm_adjustParameters() : If the params->hashRateLog is not set, set it to its default value based on windowLog and params->hashLog.

Ensures that params->bucketSizeLog is <= params->hashLog (setting it to params->hashLog if it is not).

Ensures that the minMatchLength >= targetLength during optimal parsing.

Definition at line 22 of file zstd_ldm.c.

{
    params->windowLog = cParams->windowLog;
    ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
    DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
    if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
    if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
    if (cParams->strategy >= ZSTD_btopt) {
      /* Get out of the way of the optimal parser */
      U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
      assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
      assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
      params->minMatchLength = minMatch;
    }
    if (params->hashLog == 0) {
        params->hashLog = MAX(ZSTD_HASHLOG_MIN, params->windowLog - LDM_HASH_RLOG);
        assert(params->hashLog <= ZSTD_HASHLOG_MAX);
    }
    if (params->hashRateLog == 0) {
        params->hashRateLog = params->windowLog < params->hashLog
                                   ? 0
                                   : params->windowLog - params->hashLog;
    }
    params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
}

Referenced by ZSTD_resetCCtx_internal().

ZSTD_ldm_blockCompress()

size_t ZSTD_ldm_blockCompress	(	rawSeqStore_t *	rawSeqStore,
		ZSTD_matchState_t *	ms,
		seqStore_t *	seqStore,
		U32	rep[ZSTD_REP_NUM],
		void const *	src,
		size_t	srcSize
	)

ZSTD_ldm_blockCompress():

Compresses a block using the predefined sequences, along with a secondary block compressor. The literals section of every sequence is passed to the secondary block compressor, and those sequences are interspersed with the predefined sequences. Returns the length of the last literals. Updates rawSeqStore.pos to indicate how many sequences have been consumed. rawSeqStore.seq may also be updated to split the last sequence between two blocks.

Returns

The length of the last literals.

NOTE: The source must be at most the maximum block size, but the predefined sequences can be any size, and may be longer than the block. In the case that they are longer than the block, the last sequences may need to be split into two. We handle that case correctly, and update rawSeqStore appropriately. NOTE: This function does not return any errors.

Definition at line 565 of file zstd_ldm.c.

{
    const ZSTD_compressionParameters* const cParams = &ms->cParams;
    unsigned const minMatch = cParams->minMatch;
    ZSTD_blockCompressor const blockCompressor =
        ZSTD_selectBlockCompressor(cParams->strategy, ZSTD_matchState_dictMode(ms));
    /* Input bounds */
    BYTE const* const istart = (BYTE const*)src;
    BYTE const* const iend = istart + srcSize;
    /* Input positions */
    BYTE const* ip = istart;
 
    DEBUGLOG(5, "ZSTD_ldm_blockCompress: srcSize=%zu", srcSize);
    assert(rawSeqStore->pos <= rawSeqStore->size);
    assert(rawSeqStore->size <= rawSeqStore->capacity);
    /* Loop through each sequence and apply the block compressor to the lits */
    while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
        /* maybeSplitSequence updates rawSeqStore->pos */
        rawSeq const sequence = maybeSplitSequence(rawSeqStore,
                                                   (U32)(iend - ip), minMatch);
        int i;
        /* End signal */
        if (sequence.offset == 0)
            break;
 
        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, "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);
            ip += sequence.litLength;
            /* Update the repcodes */
            for (i = ZSTD_REP_NUM - 1; i > 0; i--)
                rep[i] = rep[i-1];
            rep[0] = sequence.offset;
            /* Store the sequence */
            ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, iend,
                          sequence.offset + ZSTD_REP_MOVE,
                          sequence.matchLength - MINMATCH);
            ip += sequence.matchLength;
        }
    }
    /* Fill the tables for the block compressor */
    ZSTD_ldm_limitTableUpdate(ms, ip);
    ZSTD_ldm_fillFastTables(ms, ip);
    /* Compress the last literals */
    return blockCompressor(ms, seqStore, rep, ip, iend - ip);
}

Referenced by ZSTD_buildSeqStore().

ZSTD_ldm_fillHashTable()

void ZSTD_ldm_fillHashTable	(	ldmState_t *	state,
		const BYTE *	ip,
		const BYTE *	iend,
		ldmParams_t const *	params
	)

Definition at line 227 of file zstd_ldm.c.

{
    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);
    }
}

Referenced by ZSTD_loadDictionaryContent().

ZSTD_ldm_generateSequences()

size_t ZSTD_ldm_generateSequences	(	ldmState_t *	ldms,
		rawSeqStore_t *	sequences,
		ldmParams_t const *	params,
		void const *	src,
		size_t	srcSize
	)

ZSTD_ldm_generateSequences():

Generates the sequences using the long distance match finder. Generates long range matching sequences in sequences, which parse a prefix of the source. sequences must be large enough to store every sequence, which can be checked with ZSTD_ldm_getMaxNbSeq().

Returns

0 or an error code.

NOTE: The user must have called ZSTD_window_update() for all of the input they have, even if they pass it to ZSTD_ldm_generateSequences() in chunks. NOTE: This function returns an error if it runs out of space to store sequences.

Definition at line 429 of file zstd_ldm.c.

{
    U32 const maxDist = 1U << params->windowLog;
    BYTE const* const istart = (BYTE const*)src;
    BYTE const* const iend = istart + srcSize;
    size_t const kMaxChunkSize = 1 << 20;
    size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
    size_t chunk;
    size_t leftoverSize = 0;
 
    assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
    /* Check that ZSTD_window_update() has been called for this chunk prior
     * to passing it to this function.
     */
    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 maximum distance and handle overflow correction.
     */
    assert(sequences->pos <= sequences->size);
    assert(sequences->size <= sequences->capacity);
    for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
        BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
        size_t const remaining = (size_t)(iend - chunkStart);
        BYTE const *const chunkEnd =
            (remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
        size_t const chunkSize = chunkEnd - chunkStart;
        size_t newLeftoverSize;
        size_t const prevSize = sequences->size;
 
        assert(chunkStart < iend);
        /* 1. Perform overflow correction if necessary. */
        if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
            U32 const ldmHSize = 1U << params->hashLog;
            U32 const correction = ZSTD_window_correctOverflow(
                &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.
         *
         * kMaxChunkSize should be small enough that we don't lose too much of
         * the window through early invalidation.
         * 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, &ldmState->loadedDictEnd, NULL);
        /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
        newLeftoverSize = ZSTD_ldm_generateSequences_internal(
            ldmState, sequences, params, chunkStart, chunkSize);
        if (ZSTD_isError(newLeftoverSize))
            return newLeftoverSize;
        /* 4. We add the leftover literals from previous iterations to the first
         *    newly generated sequence, or add the `newLeftoverSize` if none are
         *    generated.
         */
        /* Prepend the leftover literals from the last call */
        if (prevSize < sequences->size) {
            sequences->seq[prevSize].litLength += (U32)leftoverSize;
            leftoverSize = newLeftoverSize;
        } else {
            assert(newLeftoverSize == chunkSize);
            leftoverSize += chunkSize;
        }
    }
    return 0;
}

Referenced by ZSTD_buildSeqStore().

ZSTD_ldm_getMaxNbSeq()

size_t ZSTD_ldm_getMaxNbSeq	(	ldmParams_t	params,
		size_t	maxChunkSize
	)

ZSTD_ldm_getSeqSpace() : Return an upper bound on the number of sequences that can be produced by the long distance matcher, or 0 if LDM is disabled.

Definition at line 59 of file zstd_ldm.c.

{
    return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
}

Referenced by ZSTD_estimateCCtxSize_usingCCtxParams(), and ZSTD_resetCCtx_internal().

ZSTD_ldm_getTableSize()

size_t ZSTD_ldm_getTableSize

(

ldmParams_t

params

)

ZSTD_ldm_getTableSize() : Estimate the space needed for long distance matching tables or 0 if LDM is disabled.

Definition at line 49 of file zstd_ldm.c.

{
    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 = ZSTD_cwksp_alloc_size(ldmBucketSize)
                           + ZSTD_cwksp_alloc_size(ldmHSize * sizeof(ldmEntry_t));
    return params.enableLdm ? totalSize : 0;
}

Referenced by ZSTD_estimateCCtxSize_usingCCtxParams(), and ZSTD_resetCCtx_internal().

ZSTD_ldm_skipSequences()

void ZSTD_ldm_skipSequences	(	rawSeqStore_t *	rawSeqStore,
		size_t	srcSize,
		U32 const	minMatch
	)

ZSTD_ldm_skipSequences():

Skip past srcSize bytes worth of sequences in rawSeqStore. Avoids emitting matches less than minMatch bytes. Must be called for data with is not passed to ZSTD_ldm_blockCompress().

Definition at line 506 of file zstd_ldm.c.

                                                                                            {
    while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
        rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
        if (srcSize <= seq->litLength) {
            /* Skip past srcSize literals */
            seq->litLength -= (U32)srcSize;
            return;
        }
        srcSize -= seq->litLength;
        seq->litLength = 0;
        if (srcSize < seq->matchLength) {
            /* Skip past the first srcSize of the match */
            seq->matchLength -= (U32)srcSize;
            if (seq->matchLength < minMatch) {
                /* The match is too short, omit it */
                if (rawSeqStore->pos + 1 < rawSeqStore->size) {
                    seq[1].litLength += seq[0].matchLength;
                }
                rawSeqStore->pos++;
            }
            return;
        }
        srcSize -= seq->matchLength;
        seq->matchLength = 0;
        rawSeqStore->pos++;
    }
}

Referenced by maybeSplitSequence(), and ZSTD_buildSeqStore().