zstd/lib/dictBuilder/zdict.h

306 lines
18 KiB
C
Raw Normal View History

/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
2016-08-30 17:04:33 +00:00
* 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.
2016-08-30 17:04:33 +00:00
*/
2016-02-11 23:07:30 +00:00
#ifndef DICTBUILDER_H_001
#define DICTBUILDER_H_001
#if defined (__cplusplus)
extern "C" {
#endif
2016-08-30 17:04:33 +00:00
/*====== Dependencies ======*/
#include <stddef.h> /* size_t */
2016-12-16 21:27:30 +00:00
/* ===== ZDICTLIB_API : control library symbols visibility ===== */
#ifndef ZDICTLIB_VISIBILITY
# if defined(__GNUC__) && (__GNUC__ >= 4)
# define ZDICTLIB_VISIBILITY __attribute__ ((visibility ("default")))
# else
# define ZDICTLIB_VISIBILITY
# endif
2016-12-16 21:27:30 +00:00
#endif
#if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
# define ZDICTLIB_API __declspec(dllexport) ZDICTLIB_VISIBILITY
#elif defined(ZSTD_DLL_IMPORT) && (ZSTD_DLL_IMPORT==1)
# define ZDICTLIB_API __declspec(dllimport) ZDICTLIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
#else
2016-12-16 21:27:30 +00:00
# define ZDICTLIB_API ZDICTLIB_VISIBILITY
#endif
/*! ZDICT_trainFromBuffer():
* Train a dictionary from an array of samples.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
* Redirect towards ZDICT_optimizeTrainFromBuffer_fastCover() single-threaded, with d=8, steps=4,
2018-09-05 00:12:35 +00:00
* f=20, and accel=1.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* The resulting dictionary will be saved into `dictBuffer`.
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
2019-02-01 23:19:32 +00:00
* Note: Dictionary training will fail if there are not enough samples to construct a
* dictionary, or if most of the samples are too small (< 8 bytes being the lower limit).
* If dictionary training fails, you should use zstd without a dictionary, as the dictionary
* would've been ineffective anyways. If you believe your samples would benefit from a dictionary
* please open an issue with details, and we can look into it.
* Note: ZDICT_trainFromBuffer()'s memory usage is about 6 MB.
* Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
* It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
* In general, it's recommended to provide a few thousands samples, though this can vary a lot.
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
const void* samplesBuffer,
const size_t* samplesSizes, unsigned nbSamples);
typedef struct {
int compressionLevel; /*< optimize for a specific zstd compression level; 0 means default */
unsigned notificationLevel; /*< Write log to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
unsigned dictID; /*< force dictID value; 0 means auto mode (32-bits random value) */
} ZDICT_params_t;
/*! ZDICT_finalizeDictionary():
* Given a custom content as a basis for dictionary, and a set of samples,
* finalize dictionary by adding headers and statistics according to the zstd
* dictionary format.
*
* Samples must be stored concatenated in a flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each
* sample in order. The samples are used to construct the statistics, so they
* should be representative of what you will compress with this dictionary.
*
* The compression level can be set in `parameters`. You should pass the
* compression level you expect to use in production. The statistics for each
* compression level differ, so tuning the dictionary for the compression level
* can help quite a bit.
*
* You can set an explicit dictionary ID in `parameters`, or allow us to pick
* a random dictionary ID for you, but we can't guarantee no collisions.
*
* The dstDictBuffer and the dictContent may overlap, and the content will be
* appended to the end of the header. If the header + the content doesn't fit in
* maxDictSize the beginning of the content is truncated to make room, since it
* is presumed that the most profitable content is at the end of the dictionary,
* since that is the cheapest to reference.
*
* `dictContentSize` must be >= ZDICT_CONTENTSIZE_MIN bytes.
* `maxDictSize` must be >= max(dictContentSize, ZSTD_DICTSIZE_MIN).
*
* @return: size of dictionary stored into `dstDictBuffer` (<= `maxDictSize`),
* or an error code, which can be tested by ZDICT_isError().
* Note: ZDICT_finalizeDictionary() will push notifications into stderr if
* instructed to, using notificationLevel>0.
* NOTE: This function currently may fail in several edge cases including:
* * Not enough samples
* * Samples are uncompressible
* * Samples are all exactly the same
*/
ZDICTLIB_API size_t ZDICT_finalizeDictionary(void* dstDictBuffer, size_t maxDictSize,
const void* dictContent, size_t dictContentSize,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
ZDICT_params_t parameters);
/*====== Helper functions ======*/
2016-08-19 12:23:58 +00:00
ZDICTLIB_API unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize); /**< extracts dictID; @return zero if error (not a valid dictionary) */
ZDICTLIB_API size_t ZDICT_getDictHeaderSize(const void* dictBuffer, size_t dictSize); /* returns dict header size; returns a ZSTD error code on failure */
ZDICTLIB_API unsigned ZDICT_isError(size_t errorCode);
ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
2016-02-11 23:07:30 +00:00
#ifdef ZDICT_STATIC_LINKING_ONLY
/* ====================================================================================
* The definitions in this section are considered experimental.
* They should never be used with a dynamic library, as they may change in the future.
* They are provided for advanced usages.
* Use them only in association with static linking.
* ==================================================================================== */
#define ZDICT_CONTENTSIZE_MIN 128
#define ZDICT_DICTSIZE_MIN 256
/*! ZDICT_cover_params_t:
* k and d are the only required parameters.
* For others, value 0 means default.
*/
2017-01-01 05:07:44 +00:00
typedef struct {
2017-01-02 20:40:43 +00:00
unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
2017-01-27 19:56:02 +00:00
unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (1.0), 1.0 when all samples are used for both training and testing */
unsigned shrinkDict; /* Train dictionaries to shrink in size starting from the minimum size and selects the smallest dictionary that is shrinkDictMaxRegression% worse than the largest dictionary. 0 means no shrinking and 1 means shrinking */
unsigned shrinkDictMaxRegression; /* Sets shrinkDictMaxRegression so that a smaller dictionary can be at worse shrinkDictMaxRegression% worse than the max dict size dictionary. */
ZDICT_params_t zParams;
} ZDICT_cover_params_t;
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
typedef struct {
unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
2018-09-05 00:12:35 +00:00
unsigned f; /* log of size of frequency array : constraint: 0 < f <= 31 : 1 means default(20)*/
unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (0.75), 1.0 when all samples are used for both training and testing */
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
unsigned accel; /* Acceleration level: constraint: 0 < accel <= 10, higher means faster and less accurate, 0 means default(1) */
unsigned shrinkDict; /* Train dictionaries to shrink in size starting from the minimum size and selects the smallest dictionary that is shrinkDictMaxRegression% worse than the largest dictionary. 0 means no shrinking and 1 means shrinking */
unsigned shrinkDictMaxRegression; /* Sets shrinkDictMaxRegression so that a smaller dictionary can be at worse shrinkDictMaxRegression% worse than the max dict size dictionary. */
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
ZDICT_params_t zParams;
} ZDICT_fastCover_params_t;
/*! ZDICT_trainFromBuffer_cover():
* Train a dictionary from an array of samples using the COVER algorithm.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* The resulting dictionary will be saved into `dictBuffer`.
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
2019-02-01 23:19:32 +00:00
* See ZDICT_trainFromBuffer() for details on failure modes.
* Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
* Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
* It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
* In general, it's recommended to provide a few thousands samples, though this can vary a lot.
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
void *dictBuffer, size_t dictBufferCapacity,
const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
ZDICT_cover_params_t parameters);
/*! ZDICT_optimizeTrainFromBuffer_cover():
* The same requirements as above hold for all the parameters except `parameters`.
* This function tries many parameter combinations and picks the best parameters.
* `*parameters` is filled with the best parameters found,
* dictionary constructed with those parameters is stored in `dictBuffer`.
*
* All of the parameters d, k, steps are optional.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
* If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
* if steps is zero it defaults to its default value.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
* If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
*
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
2019-02-01 23:19:32 +00:00
* or an error code, which can be tested with ZDICT_isError().
* On success `*parameters` contains the parameters selected.
* See ZDICT_trainFromBuffer() for details on failure modes.
* Note: ZDICT_optimizeTrainFromBuffer_cover() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
*/
ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
void* dictBuffer, size_t dictBufferCapacity,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
ZDICT_cover_params_t* parameters);
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
/*! ZDICT_trainFromBuffer_fastCover():
* Train a dictionary from an array of samples using a modified version of COVER algorithm.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* d and k are required.
* All other parameters are optional, will use default values if not provided
* The resulting dictionary will be saved into `dictBuffer`.
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
2019-02-01 23:19:32 +00:00
* See ZDICT_trainFromBuffer() for details on failure modes.
* Note: ZDICT_trainFromBuffer_fastCover() requires 6 * 2^f bytes of memory.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
* Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
* It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
* In general, it's recommended to provide a few thousands samples, though this can vary a lot.
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer_fastCover(void *dictBuffer,
size_t dictBufferCapacity, const void *samplesBuffer,
const size_t *samplesSizes, unsigned nbSamples,
ZDICT_fastCover_params_t parameters);
/*! ZDICT_optimizeTrainFromBuffer_fastCover():
* The same requirements as above hold for all the parameters except `parameters`.
* This function tries many parameter combinations (specifically, k and d combinations)
* and picks the best parameters. `*parameters` is filled with the best parameters found,
* dictionary constructed with those parameters is stored in `dictBuffer`.
* All of the parameters d, k, steps, f, and accel are optional.
* If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
* if steps is zero it defaults to its default value.
* If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
2018-09-05 00:12:35 +00:00
* If f is zero, default value of 20 is used.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
* If accel is zero, default value of 1 is used.
*
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
2019-02-01 23:19:32 +00:00
* or an error code, which can be tested with ZDICT_isError().
* On success `*parameters` contains the parameters selected.
* See ZDICT_trainFromBuffer() for details on failure modes.
* Note: ZDICT_optimizeTrainFromBuffer_fastCover() requires about 6 * 2^f bytes of memory for each thread.
Merge fastCover into DictBuilder (#1274) * Minor fix * Run non-optimize FASTCOVER 5 times in benchmark * Merge fastCover into dictBuilder * Fix mixed declaration issue * Add fastcover to symbol.c * Add fastCover.c and cover.h to build * Change fastCover.c to fastcover.c * Update benchmark to run FASTCOVER in dictBuilder * Undo spliting fastcover_param into cover_param and f * Remove convert param functions * Assign f to parameter * Add zdict.h to Makefile in lib * Add cover.h to BUCK * Cast 1 to U64 before shifting * Remove trimming of zero freq head and tail in selectSegment and rebenchmark * Remove f as a separate parameter of tryParam * Read 8 bytes when d is 6 * Add trimming off zero frequency head and tail * Use best functions from COVER and remove trimming part(which leads to worse compression ratio after previous bugs were fixed) * Add finalize= argument to FASTCOVER to specify percentage of training samples passed to ZDICT_finalizeDictionary * Change nbDmer to always read 8 bytes even when d=6 * Add skip=# argument to allow skipping dmers in computeFrequency in FASTCOVER * Update comments and benchmarking result * Change default method of ZDICT_trainFromBuffer to ZDICT_optimizeTrainFromBuffer_fastCover * Add dictType enum and fix bug about passing zParam when converting to coverParam * Combine finalize and skip into a single parameter * Update acceleration parameters and benchmark on 3 sample sets * Change default splitPoint of FASTCOVER to 0.75 and benchmark first 3 sample sets * Initialize variables outside of for loop in benchmark.c * Update benchmark result for hg-manifest * Remove cover.h from install-includes * Add explanation of f * Set default compression level for trainFromBuffer to 3 * Add assertion of fastCoverParams in DiB_trainFromFiles * Add checkTotalCompressedSize function + some minor fixes * Add test for multithreading fastCovr * Initialize segmentFreqs in every FASTCOVER_selectSegment and move mutex_unnlock to end of COVER_best_finish * Free segmentFreqs * Initialize segmentFreqs before calling FASTCOVER_buildDictionary instead of in FASTCOVER_selectSegment * Add FASTCOVER_MEMMULT * Minor fix * Update benchmarking result
2018-08-23 19:06:20 +00:00
*/
ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_fastCover(void* dictBuffer,
size_t dictBufferCapacity, const void* samplesBuffer,
const size_t* samplesSizes, unsigned nbSamples,
ZDICT_fastCover_params_t* parameters);
typedef struct {
unsigned selectivityLevel; /* 0 means default; larger => select more => larger dictionary */
ZDICT_params_t zParams;
} ZDICT_legacy_params_t;
/*! ZDICT_trainFromBuffer_legacy():
* Train a dictionary from an array of samples.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* The resulting dictionary will be saved into `dictBuffer`.
* `parameters` is optional and can be provided with values set to 0 to mean "default".
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
2019-02-01 23:19:32 +00:00
* See ZDICT_trainFromBuffer() for details on failure modes.
* Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
* It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
* In general, it's recommended to provide a few thousands samples, though this can vary a lot.
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
* Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer_legacy(
void *dictBuffer, size_t dictBufferCapacity,
const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
ZDICT_legacy_params_t parameters);
2016-12-22 19:18:43 +00:00
/* Deprecation warnings */
/* It is generally possible to disable deprecation warnings from compiler,
for example with -Wno-deprecated-declarations for gcc
or _CRT_SECURE_NO_WARNINGS in Visual.
Otherwise, it's also possible to manually define ZDICT_DISABLE_DEPRECATE_WARNINGS */
#ifdef ZDICT_DISABLE_DEPRECATE_WARNINGS
# define ZDICT_DEPRECATED(message) ZDICTLIB_API /* disable deprecation warnings */
2016-12-22 19:18:43 +00:00
#else
# define ZDICT_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
# if defined (__cplusplus) && (__cplusplus >= 201402) /* C++14 or greater */
# define ZDICT_DEPRECATED(message) [[deprecated(message)]] ZDICTLIB_API
2016-12-22 19:18:43 +00:00
# elif (ZDICT_GCC_VERSION >= 405) || defined(__clang__)
# define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated(message)))
2016-12-22 19:18:43 +00:00
# elif (ZDICT_GCC_VERSION >= 301)
# define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated))
2016-12-22 19:18:43 +00:00
# elif defined(_MSC_VER)
# define ZDICT_DEPRECATED(message) ZDICTLIB_API __declspec(deprecated(message))
2016-12-22 19:18:43 +00:00
# else
# pragma message("WARNING: You need to implement ZDICT_DEPRECATED for this compiler")
# define ZDICT_DEPRECATED(message) ZDICTLIB_API
2016-12-22 19:18:43 +00:00
# endif
#endif /* ZDICT_DISABLE_DEPRECATE_WARNINGS */
ZDICT_DEPRECATED("use ZDICT_finalizeDictionary() instead")
size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
#endif /* ZDICT_STATIC_LINKING_ONLY */
2016-02-11 23:07:30 +00:00
#if defined (__cplusplus)
}
#endif
2016-07-13 15:38:39 +00:00
#endif /* DICTBUILDER_H_001 */