ReactOS  0.4.15-dev-1386-g5cb9f87
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_DEFAULTMAX
 

Functions

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)
 
U64 ZSTD_ldm_getHashPower (U32 minMatchLength)
 
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_DEFAULTMAX

Definition at line 24 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->hashEveryLog 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 21 of file zstd_ldm.c.

23 {
24  params->windowLog = cParams->windowLog;
25  ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
26  DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
27  if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
28  if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
29  if (cParams->strategy >= ZSTD_btopt) {
30  /* Get out of the way of the optimal parser */
31  U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
32  assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
33  assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
34  params->minMatchLength = minMatch;
35  }
36  if (params->hashLog == 0) {
37  params->hashLog = MAX(ZSTD_HASHLOG_MIN, params->windowLog - LDM_HASH_RLOG);
38  assert(params->hashLog <= ZSTD_HASHLOG_MAX);
39  }
40  if (params->hashEveryLog == 0) {
41  params->hashEveryLog = params->windowLog < params->hashLog
42  ? 0
43  : params->windowLog - params->hashLog;
44  }
45  params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
46 }
#define assert(x)
Definition: debug.h:53
#define LDM_MIN_MATCH_LENGTH
Definition: zstd_ldm.c:17
T MIN(T a, T b)
Definition: polytest.cpp:79
#define ZSTD_STATIC_ASSERT(c)
Definition: zstd_internal.h:43
#define DEBUGLOG(l,...)
Definition: debug.h:115
GLenum const GLfloat * params
Definition: glext.h:5645
#define LDM_BUCKET_SIZE_LOG
Definition: zstd_ldm.c:16
T MAX(T a, T b)
Definition: polytest.cpp:85
#define LDM_HASH_RLOG
Definition: zstd_ldm.c:18
unsigned int U32
Definition: xxhash.c:195

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 591 of file zstd_ldm.c.

594 {
595  const ZSTD_compressionParameters* const cParams = &ms->cParams;
596  unsigned const minMatch = cParams->searchLength;
597  ZSTD_blockCompressor const blockCompressor =
599  /* Input bounds */
600  BYTE const* const istart = (BYTE const*)src;
601  BYTE const* const iend = istart + srcSize;
602  /* Input positions */
603  BYTE const* ip = istart;
604 
605  DEBUGLOG(5, "ZSTD_ldm_blockCompress: srcSize=%zu", srcSize);
606  assert(rawSeqStore->pos <= rawSeqStore->size);
607  assert(rawSeqStore->size <= rawSeqStore->capacity);
608  /* Loop through each sequence and apply the block compressor to the lits */
609  while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
610  /* maybeSplitSequence updates rawSeqStore->pos */
611  rawSeq const sequence = maybeSplitSequence(rawSeqStore,
612  (U32)(iend - ip), minMatch);
613  int i;
614  /* End signal */
615  if (sequence.offset == 0)
616  break;
617 
618  assert(sequence.offset <= (1U << cParams->windowLog));
619  assert(ip + sequence.litLength + sequence.matchLength <= iend);
620 
621  /* Fill tables for block compressor */
624  /* Run the block compressor */
625  DEBUGLOG(5, "calling block compressor on segment of size %u", sequence.litLength);
626  {
627  size_t const newLitLength =
628  blockCompressor(ms, seqStore, rep, ip, sequence.litLength);
629  ip += sequence.litLength;
630  /* Update the repcodes */
631  for (i = ZSTD_REP_NUM - 1; i > 0; i--)
632  rep[i] = rep[i-1];
633  rep[0] = sequence.offset;
634  /* Store the sequence */
635  ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength,
636  sequence.offset + ZSTD_REP_MOVE,
637  sequence.matchLength - MINMATCH);
638  ip += sequence.matchLength;
639  }
640  }
641  /* Fill the tables for the block compressor */
644  /* Compress the last literals */
645  return blockCompressor(ms, seqStore, rep, ip, iend - ip);
646 }
#define U(x)
Definition: wordpad.c:45
#define assert(x)
Definition: debug.h:53
#define DEBUGLOG(l,...)
Definition: debug.h:115
ZSTD_compressionParameters cParams
ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMode_e dictMode)
#define ZSTD_REP_MOVE
Definition: zstd_internal.h:63
Definition: dhcpd.h:61
MEM_STATIC void ZSTD_storeSeq(seqStore_t *seqStorePtr, size_t litLength, const void *literals, U32 offsetCode, size_t mlBase)
size_t(* ZSTD_blockCompressor)(ZSTD_matchState_t *bs, seqStore_t *seqStore, U32 rep[ZSTD_REP_NUM], void const *src, size_t srcSize)
static struct message * sequence
Definition: subclass.c:46
GLenum src
Definition: glext.h:6340
static rawSeq maybeSplitSequence(rawSeqStore_t *rawSeqStore, U32 const remaining, U32 const minMatch)
Definition: zstd_ldm.c:567
unsigned char BYTE
Definition: xxhash.c:193
static void ZSTD_ldm_limitTableUpdate(ZSTD_matchState_t *ms, const BYTE *anchor)
Definition: zstd_ldm.c:281
static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t *ms, void const *end)
Definition: zstd_ldm.c:220
GLsizei GLenum const GLvoid GLsizei GLenum GLbyte GLbyte GLbyte GLdouble GLdouble GLdouble GLfloat GLfloat GLfloat GLint GLint GLint GLshort GLshort GLshort GLubyte GLubyte GLubyte GLuint GLuint GLuint GLushort GLushort GLushort GLbyte GLbyte GLbyte GLbyte GLdouble GLdouble GLdouble GLdouble GLfloat GLfloat GLfloat GLfloat GLint GLint GLint GLint GLshort GLshort GLshort GLshort GLubyte GLubyte GLubyte GLubyte GLuint GLuint GLuint GLuint GLushort GLushort GLushort GLushort GLboolean const GLdouble const GLfloat const GLint const GLshort const GLbyte const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLdouble const GLfloat const GLfloat const GLint const GLint const GLshort const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort const GLdouble const GLfloat const GLint const GLshort GLenum GLenum GLenum GLfloat GLenum GLint GLenum GLenum GLenum GLfloat GLenum GLenum GLint GLenum GLfloat GLenum GLint GLint GLushort GLenum GLenum GLfloat GLenum GLenum GLint GLfloat const GLubyte GLenum GLenum GLenum const GLfloat GLenum GLenum const GLint GLenum GLint GLint GLsizei GLsizei GLint GLenum GLenum const GLvoid GLenum GLenum const GLfloat GLenum GLenum const GLint GLenum GLenum const GLdouble GLenum GLenum const GLfloat GLenum GLenum const GLint GLsizei GLuint GLfloat GLuint GLbitfield GLfloat GLint GLuint GLboolean GLenum GLfloat GLenum GLbitfield GLenum GLfloat GLfloat GLint GLint const GLfloat GLenum GLfloat GLfloat GLint GLint GLfloat GLfloat GLint GLint const GLfloat GLint GLfloat GLfloat GLint GLfloat GLfloat GLint GLfloat GLfloat const GLdouble const GLfloat const GLdouble const GLfloat GLint i
Definition: glfuncs.h:248
MEM_STATIC ZSTD_dictMode_e ZSTD_matchState_dictMode(const ZSTD_matchState_t *ms)
#define MINMATCH
Definition: zstd_internal.h:96
unsigned int U32
Definition: xxhash.c:195
#define ZSTD_REP_NUM
Definition: zstd_internal.h:62

Referenced by ZSTD_compressBlock_internal().

◆ 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 463 of file zstd_ldm.c.

466 {
467  U32 const maxDist = 1U << params->windowLog;
468  BYTE const* const istart = (BYTE const*)src;
469  BYTE const* const iend = istart + srcSize;
470  size_t const kMaxChunkSize = 1 << 20;
471  size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
472  size_t chunk;
473  size_t leftoverSize = 0;
474 
475  assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
476  /* Check that ZSTD_window_update() has been called for this chunk prior
477  * to passing it to this function.
478  */
479  assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize);
480  /* The input could be very large (in zstdmt), so it must be broken up into
481  * chunks to enforce the maximmum distance and handle overflow correction.
482  */
483  assert(sequences->pos <= sequences->size);
484  assert(sequences->size <= sequences->capacity);
485  for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
486  BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
487  size_t const remaining = (size_t)(iend - chunkStart);
488  BYTE const *const chunkEnd =
489  (remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
490  size_t const chunkSize = chunkEnd - chunkStart;
491  size_t newLeftoverSize;
492  size_t const prevSize = sequences->size;
493 
494  assert(chunkStart < iend);
495  /* 1. Perform overflow correction if necessary. */
496  if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
497  U32 const ldmHSize = 1U << params->hashLog;
498  U32 const correction = ZSTD_window_correctOverflow(
499  &ldmState->window, /* cycleLog */ 0, maxDist, src);
500  ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
501  }
502  /* 2. We enforce the maximum offset allowed.
503  *
504  * kMaxChunkSize should be small enough that we don't lose too much of
505  * the window through early invalidation.
506  * TODO: * Test the chunk size.
507  * * Try invalidation after the sequence generation and test the
508  * the offset against maxDist directly.
509  */
510  ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, NULL, NULL);
511  /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
512  newLeftoverSize = ZSTD_ldm_generateSequences_internal(
513  ldmState, sequences, params, chunkStart, chunkSize);
514  if (ZSTD_isError(newLeftoverSize))
515  return newLeftoverSize;
516  /* 4. We add the leftover literals from previous iterations to the first
517  * newly generated sequence, or add the `newLeftoverSize` if none are
518  * generated.
519  */
520  /* Prepend the leftover literals from the last call */
521  if (prevSize < sequences->size) {
522  sequences->seq[prevSize].litLength += (U32)leftoverSize;
523  leftoverSize = newLeftoverSize;
524  } else {
525  assert(newLeftoverSize == chunkSize);
526  leftoverSize += chunkSize;
527  }
528  }
529  return 0;
530 }
#define U(x)
Definition: wordpad.c:45
#define assert(x)
Definition: debug.h:53
static void ZSTD_ldm_reduceTable(ldmEntry_t *const table, U32 const size, U32 const reducerValue)
Definition: zstd_ldm.c:453
GLenum const GLfloat * params
Definition: glext.h:5645
static size_t ZSTD_ldm_generateSequences_internal(ldmState_t *ldmState, rawSeqStore_t *rawSeqStore, ldmParams_t const *params, void const *src, size_t srcSize)
Definition: zstd_ldm.c:290
MEM_STATIC void ZSTD_window_enforceMaxDist(ZSTD_window_t *window, void const *srcEnd, U32 maxDist, U32 *loadedDictEndPtr, const ZSTD_matchState_t **dictMatchStatePtr)
__kernel_size_t size_t
Definition: linux.h:237
GLsizeiptr size
Definition: glext.h:5919
MEM_STATIC U32 ZSTD_window_needOverflowCorrection(ZSTD_window_t const window, void const *srcEnd)
GLenum src
Definition: glext.h:6340
unsigned char BYTE
Definition: xxhash.c:193
static struct msg_sequence * sequences[NUM_MSG_SEQUENCES]
Definition: button.c:54
MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t *window, U32 cycleLog, U32 maxDist, void const *src)
#define ZSTD_isError
#define ZSTD_CHUNKSIZE_MAX
#define NULL
Definition: types.h:112
unsigned int U32
Definition: xxhash.c:195

Referenced by ZSTD_compressBlock_internal().

◆ ZSTD_ldm_getHashPower()

U64 ZSTD_ldm_getHashPower ( U32  minMatchLength)

ZSTD_ldm_getTableSize() : Return prime8bytes^(minMatchLength-1)

Definition at line 180 of file zstd_ldm.c.

180  {
181  DEBUGLOG(4, "ZSTD_ldm_getHashPower: mml=%u", minMatchLength);
182  assert(minMatchLength >= ZSTD_LDM_MINMATCH_MIN);
183  return ZSTD_ldm_ipow(prime8bytes, minMatchLength - 1);
184 }
#define assert(x)
Definition: debug.h:53
#define DEBUGLOG(l,...)
Definition: debug.h:115
static const U64 prime8bytes
static U64 ZSTD_ldm_ipow(U64 base, U64 exp)
Definition: zstd_ldm.c:169

Referenced by ZSTD_resetCCtx_internal().

◆ 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 58 of file zstd_ldm.c.

59 {
60  return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
61 }
GLenum const GLfloat * params
Definition: glext.h:5645

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 48 of file zstd_ldm.c.

49 {
50  size_t const ldmHSize = ((size_t)1) << params.hashLog;
51  size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog);
52  size_t const ldmBucketSize =
53  ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
54  size_t const totalSize = ldmBucketSize + ldmHSize * sizeof(ldmEntry_t);
55  return params.enableLdm ? totalSize : 0;
56 }
T MIN(T a, T b)
Definition: polytest.cpp:79
GLenum const GLfloat * params
Definition: glext.h:5645
__kernel_size_t size_t
Definition: linux.h:237

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 532 of file zstd_ldm.c.

532  {
533  while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
534  rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
535  if (srcSize <= seq->litLength) {
536  /* Skip past srcSize literals */
537  seq->litLength -= (U32)srcSize;
538  return;
539  }
540  srcSize -= seq->litLength;
541  seq->litLength = 0;
542  if (srcSize < seq->matchLength) {
543  /* Skip past the first srcSize of the match */
544  seq->matchLength -= (U32)srcSize;
545  if (seq->matchLength < minMatch) {
546  /* The match is too short, omit it */
547  if (rawSeqStore->pos + 1 < rawSeqStore->size) {
548  seq[1].litLength += seq[0].matchLength;
549  }
550  rawSeqStore->pos++;
551  }
552  return;
553  }
554  srcSize -= seq->matchLength;
555  seq->matchLength = 0;
556  rawSeqStore->pos++;
557  }
558 }
unsigned int U32
Definition: xxhash.c:195

Referenced by maybeSplitSequence(), and ZSTD_compressBlock_internal().