Path: blob/main/sys/contrib/openzfs/module/zstd/lib/compress/zstd_compress.c
48774 views
// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0-only1/*2* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.3* All rights reserved.4*5* This source code is licensed under both the BSD-style license (found in the6* LICENSE file in the root directory of this source tree) and the GPLv2 (found7* 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*/1011/*-*************************************12* Dependencies13***************************************/14#include <limits.h> /* INT_MAX */15#include <string.h> /* memset */16#include "../common/cpu.h"17#include "../common/mem.h"18#include "hist.h" /* HIST_countFast_wksp */19#define FSE_STATIC_LINKING_ONLY /* FSE_encodeSymbol */20#include "../common/fse.h"21#define HUF_STATIC_LINKING_ONLY22#include "../common/huf.h"23#include "zstd_compress_internal.h"24#include "zstd_compress_sequences.h"25#include "zstd_compress_literals.h"26#include "zstd_fast.h"27#include "zstd_double_fast.h"28#include "zstd_lazy.h"29#include "zstd_opt.h"30#include "zstd_ldm.h"31#include "zstd_compress_superblock.h"323334/*-*************************************35* Helper functions36***************************************/37/* ZSTD_compressBound()38* Note that the result from this function is only compatible with the "normal"39* full-block strategy.40* When there are a lot of small blocks due to frequent flush in streaming mode41* the overhead of headers can make the compressed data to be larger than the42* return value of ZSTD_compressBound().43*/44size_t ZSTD_compressBound(size_t srcSize) {45return ZSTD_COMPRESSBOUND(srcSize);46}474849/*-*************************************50* Context memory management51***************************************/52struct ZSTD_CDict_s {53const void* dictContent;54size_t dictContentSize;55U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */56ZSTD_cwksp workspace;57ZSTD_matchState_t matchState;58ZSTD_compressedBlockState_t cBlockState;59ZSTD_customMem customMem;60U32 dictID;61int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */62}; /* typedef'd to ZSTD_CDict within "zstd.h" */6364ZSTD_CCtx* ZSTD_createCCtx(void)65{66return ZSTD_createCCtx_advanced(ZSTD_defaultCMem);67}6869static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager)70{71assert(cctx != NULL);72memset(cctx, 0, sizeof(*cctx));73cctx->customMem = memManager;74cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());75{ size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters);76assert(!ZSTD_isError(err));77(void)err;78}79}8081ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)82{83ZSTD_STATIC_ASSERT(zcss_init==0);84ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1));85if (!customMem.customAlloc ^ !customMem.customFree) return NULL;86{ ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_malloc(sizeof(ZSTD_CCtx), customMem);87if (!cctx) return NULL;88ZSTD_initCCtx(cctx, customMem);89return cctx;90}91}9293ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize)94{95ZSTD_cwksp ws;96ZSTD_CCtx* cctx;97if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL; /* minimum size */98if ((size_t)workspace & 7) return NULL; /* must be 8-aligned */99ZSTD_cwksp_init(&ws, workspace, workspaceSize);100101cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));102if (cctx == NULL) return NULL;103104memset(cctx, 0, sizeof(ZSTD_CCtx));105ZSTD_cwksp_move(&cctx->workspace, &ws);106cctx->staticSize = workspaceSize;107108/* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */109if (!ZSTD_cwksp_check_available(&cctx->workspace, HUF_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL;110cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));111cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));112cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, HUF_WORKSPACE_SIZE);113cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());114return cctx;115}116117/**118* Clears and frees all of the dictionaries in the CCtx.119*/120static void ZSTD_clearAllDicts(ZSTD_CCtx* cctx)121{122ZSTD_free(cctx->localDict.dictBuffer, cctx->customMem);123ZSTD_freeCDict(cctx->localDict.cdict);124memset(&cctx->localDict, 0, sizeof(cctx->localDict));125memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));126cctx->cdict = NULL;127}128129static size_t ZSTD_sizeof_localDict(ZSTD_localDict dict)130{131size_t const bufferSize = dict.dictBuffer != NULL ? dict.dictSize : 0;132size_t const cdictSize = ZSTD_sizeof_CDict(dict.cdict);133return bufferSize + cdictSize;134}135136static void ZSTD_freeCCtxContent(ZSTD_CCtx* cctx)137{138assert(cctx != NULL);139assert(cctx->staticSize == 0);140ZSTD_clearAllDicts(cctx);141#ifdef ZSTD_MULTITHREAD142ZSTDMT_freeCCtx(cctx->mtctx); cctx->mtctx = NULL;143#endif144ZSTD_cwksp_free(&cctx->workspace, cctx->customMem);145}146147size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)148{149if (cctx==NULL) return 0; /* support free on NULL */150RETURN_ERROR_IF(cctx->staticSize, memory_allocation,151"not compatible with static CCtx");152{153int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx);154ZSTD_freeCCtxContent(cctx);155if (!cctxInWorkspace) {156ZSTD_free(cctx, cctx->customMem);157}158}159return 0;160}161162163static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)164{165#ifdef ZSTD_MULTITHREAD166return ZSTDMT_sizeof_CCtx(cctx->mtctx);167#else168(void)cctx;169return 0;170#endif171}172173174size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)175{176if (cctx==NULL) return 0; /* support sizeof on NULL */177/* cctx may be in the workspace */178return (cctx->workspace.workspace == cctx ? 0 : sizeof(*cctx))179+ ZSTD_cwksp_sizeof(&cctx->workspace)180+ ZSTD_sizeof_localDict(cctx->localDict)181+ ZSTD_sizeof_mtctx(cctx);182}183184size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)185{186return ZSTD_sizeof_CCtx(zcs); /* same object */187}188189/* private API call, for dictBuilder only */190const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); }191192static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams(193ZSTD_compressionParameters cParams)194{195ZSTD_CCtx_params cctxParams;196memset(&cctxParams, 0, sizeof(cctxParams));197cctxParams.cParams = cParams;198cctxParams.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */199assert(!ZSTD_checkCParams(cParams));200cctxParams.fParams.contentSizeFlag = 1;201return cctxParams;202}203204static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced(205ZSTD_customMem customMem)206{207ZSTD_CCtx_params* params;208if (!customMem.customAlloc ^ !customMem.customFree) return NULL;209params = (ZSTD_CCtx_params*)ZSTD_calloc(210sizeof(ZSTD_CCtx_params), customMem);211if (!params) { return NULL; }212params->customMem = customMem;213params->compressionLevel = ZSTD_CLEVEL_DEFAULT;214params->fParams.contentSizeFlag = 1;215return params;216}217218ZSTD_CCtx_params* ZSTD_createCCtxParams(void)219{220return ZSTD_createCCtxParams_advanced(ZSTD_defaultCMem);221}222223size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params)224{225if (params == NULL) { return 0; }226ZSTD_free(params, params->customMem);227return 0;228}229230size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params)231{232return ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT);233}234235size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) {236RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");237memset(cctxParams, 0, sizeof(*cctxParams));238cctxParams->compressionLevel = compressionLevel;239cctxParams->fParams.contentSizeFlag = 1;240return 0;241}242243size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)244{245RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!");246FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");247memset(cctxParams, 0, sizeof(*cctxParams));248assert(!ZSTD_checkCParams(params.cParams));249cctxParams->cParams = params.cParams;250cctxParams->fParams = params.fParams;251cctxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */252return 0;253}254255/* ZSTD_assignParamsToCCtxParams() :256* params is presumed valid at this stage */257static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams(258const ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params)259{260ZSTD_CCtx_params ret = *cctxParams;261assert(!ZSTD_checkCParams(params->cParams));262ret.cParams = params->cParams;263ret.fParams = params->fParams;264ret.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */265return ret;266}267268ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)269{270ZSTD_bounds bounds = { 0, 0, 0 };271272switch(param)273{274case ZSTD_c_compressionLevel:275bounds.lowerBound = ZSTD_minCLevel();276bounds.upperBound = ZSTD_maxCLevel();277return bounds;278279case ZSTD_c_windowLog:280bounds.lowerBound = ZSTD_WINDOWLOG_MIN;281bounds.upperBound = ZSTD_WINDOWLOG_MAX;282return bounds;283284case ZSTD_c_hashLog:285bounds.lowerBound = ZSTD_HASHLOG_MIN;286bounds.upperBound = ZSTD_HASHLOG_MAX;287return bounds;288289case ZSTD_c_chainLog:290bounds.lowerBound = ZSTD_CHAINLOG_MIN;291bounds.upperBound = ZSTD_CHAINLOG_MAX;292return bounds;293294case ZSTD_c_searchLog:295bounds.lowerBound = ZSTD_SEARCHLOG_MIN;296bounds.upperBound = ZSTD_SEARCHLOG_MAX;297return bounds;298299case ZSTD_c_minMatch:300bounds.lowerBound = ZSTD_MINMATCH_MIN;301bounds.upperBound = ZSTD_MINMATCH_MAX;302return bounds;303304case ZSTD_c_targetLength:305bounds.lowerBound = ZSTD_TARGETLENGTH_MIN;306bounds.upperBound = ZSTD_TARGETLENGTH_MAX;307return bounds;308309case ZSTD_c_strategy:310bounds.lowerBound = ZSTD_STRATEGY_MIN;311bounds.upperBound = ZSTD_STRATEGY_MAX;312return bounds;313314case ZSTD_c_contentSizeFlag:315bounds.lowerBound = 0;316bounds.upperBound = 1;317return bounds;318319case ZSTD_c_checksumFlag:320bounds.lowerBound = 0;321bounds.upperBound = 1;322return bounds;323324case ZSTD_c_dictIDFlag:325bounds.lowerBound = 0;326bounds.upperBound = 1;327return bounds;328329case ZSTD_c_nbWorkers:330bounds.lowerBound = 0;331#ifdef ZSTD_MULTITHREAD332bounds.upperBound = ZSTDMT_NBWORKERS_MAX;333#else334bounds.upperBound = 0;335#endif336return bounds;337338case ZSTD_c_jobSize:339bounds.lowerBound = 0;340#ifdef ZSTD_MULTITHREAD341bounds.upperBound = ZSTDMT_JOBSIZE_MAX;342#else343bounds.upperBound = 0;344#endif345return bounds;346347case ZSTD_c_overlapLog:348#ifdef ZSTD_MULTITHREAD349bounds.lowerBound = ZSTD_OVERLAPLOG_MIN;350bounds.upperBound = ZSTD_OVERLAPLOG_MAX;351#else352bounds.lowerBound = 0;353bounds.upperBound = 0;354#endif355return bounds;356357case ZSTD_c_enableLongDistanceMatching:358bounds.lowerBound = 0;359bounds.upperBound = 1;360return bounds;361362case ZSTD_c_ldmHashLog:363bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN;364bounds.upperBound = ZSTD_LDM_HASHLOG_MAX;365return bounds;366367case ZSTD_c_ldmMinMatch:368bounds.lowerBound = ZSTD_LDM_MINMATCH_MIN;369bounds.upperBound = ZSTD_LDM_MINMATCH_MAX;370return bounds;371372case ZSTD_c_ldmBucketSizeLog:373bounds.lowerBound = ZSTD_LDM_BUCKETSIZELOG_MIN;374bounds.upperBound = ZSTD_LDM_BUCKETSIZELOG_MAX;375return bounds;376377case ZSTD_c_ldmHashRateLog:378bounds.lowerBound = ZSTD_LDM_HASHRATELOG_MIN;379bounds.upperBound = ZSTD_LDM_HASHRATELOG_MAX;380return bounds;381382/* experimental parameters */383case ZSTD_c_rsyncable:384bounds.lowerBound = 0;385bounds.upperBound = 1;386return bounds;387388case ZSTD_c_forceMaxWindow :389bounds.lowerBound = 0;390bounds.upperBound = 1;391return bounds;392393case ZSTD_c_format:394ZSTD_STATIC_ASSERT(ZSTD_f_zstd1 < ZSTD_f_zstd1_magicless);395bounds.lowerBound = ZSTD_f_zstd1;396bounds.upperBound = ZSTD_f_zstd1_magicless; /* note : how to ensure at compile time that this is the highest value enum ? */397return bounds;398399case ZSTD_c_forceAttachDict:400ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceCopy);401bounds.lowerBound = ZSTD_dictDefaultAttach;402bounds.upperBound = ZSTD_dictForceLoad; /* note : how to ensure at compile time that this is the highest value enum ? */403return bounds;404405case ZSTD_c_literalCompressionMode:406ZSTD_STATIC_ASSERT(ZSTD_lcm_auto < ZSTD_lcm_huffman && ZSTD_lcm_huffman < ZSTD_lcm_uncompressed);407bounds.lowerBound = ZSTD_lcm_auto;408bounds.upperBound = ZSTD_lcm_uncompressed;409return bounds;410411case ZSTD_c_targetCBlockSize:412bounds.lowerBound = ZSTD_TARGETCBLOCKSIZE_MIN;413bounds.upperBound = ZSTD_TARGETCBLOCKSIZE_MAX;414return bounds;415416case ZSTD_c_srcSizeHint:417bounds.lowerBound = ZSTD_SRCSIZEHINT_MIN;418bounds.upperBound = ZSTD_SRCSIZEHINT_MAX;419return bounds;420421default:422bounds.error = ERROR(parameter_unsupported);423return bounds;424}425}426427/* ZSTD_cParam_clampBounds:428* Clamps the value into the bounded range.429*/430static size_t ZSTD_cParam_clampBounds(ZSTD_cParameter cParam, int* value)431{432ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);433if (ZSTD_isError(bounds.error)) return bounds.error;434if (*value < bounds.lowerBound) *value = bounds.lowerBound;435if (*value > bounds.upperBound) *value = bounds.upperBound;436return 0;437}438439#define BOUNDCHECK(cParam, val) { \440RETURN_ERROR_IF(!ZSTD_cParam_withinBounds(cParam,val), \441parameter_outOfBound, "Param out of bounds"); \442}443444445static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param)446{447switch(param)448{449case ZSTD_c_compressionLevel:450case ZSTD_c_hashLog:451case ZSTD_c_chainLog:452case ZSTD_c_searchLog:453case ZSTD_c_minMatch:454case ZSTD_c_targetLength:455case ZSTD_c_strategy:456return 1;457458case ZSTD_c_format:459case ZSTD_c_windowLog:460case ZSTD_c_contentSizeFlag:461case ZSTD_c_checksumFlag:462case ZSTD_c_dictIDFlag:463case ZSTD_c_forceMaxWindow :464case ZSTD_c_nbWorkers:465case ZSTD_c_jobSize:466case ZSTD_c_overlapLog:467case ZSTD_c_rsyncable:468case ZSTD_c_enableLongDistanceMatching:469case ZSTD_c_ldmHashLog:470case ZSTD_c_ldmMinMatch:471case ZSTD_c_ldmBucketSizeLog:472case ZSTD_c_ldmHashRateLog:473case ZSTD_c_forceAttachDict:474case ZSTD_c_literalCompressionMode:475case ZSTD_c_targetCBlockSize:476case ZSTD_c_srcSizeHint:477default:478return 0;479}480}481482size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)483{484DEBUGLOG(4, "ZSTD_CCtx_setParameter (%i, %i)", (int)param, value);485if (cctx->streamStage != zcss_init) {486if (ZSTD_isUpdateAuthorized(param)) {487cctx->cParamsChanged = 1;488} else {489RETURN_ERROR(stage_wrong, "can only set params in ctx init stage");490} }491492switch(param)493{494case ZSTD_c_nbWorkers:495RETURN_ERROR_IF((value!=0) && cctx->staticSize, parameter_unsupported,496"MT not compatible with static alloc");497break;498499case ZSTD_c_compressionLevel:500case ZSTD_c_windowLog:501case ZSTD_c_hashLog:502case ZSTD_c_chainLog:503case ZSTD_c_searchLog:504case ZSTD_c_minMatch:505case ZSTD_c_targetLength:506case ZSTD_c_strategy:507case ZSTD_c_ldmHashRateLog:508case ZSTD_c_format:509case ZSTD_c_contentSizeFlag:510case ZSTD_c_checksumFlag:511case ZSTD_c_dictIDFlag:512case ZSTD_c_forceMaxWindow:513case ZSTD_c_forceAttachDict:514case ZSTD_c_literalCompressionMode:515case ZSTD_c_jobSize:516case ZSTD_c_overlapLog:517case ZSTD_c_rsyncable:518case ZSTD_c_enableLongDistanceMatching:519case ZSTD_c_ldmHashLog:520case ZSTD_c_ldmMinMatch:521case ZSTD_c_ldmBucketSizeLog:522case ZSTD_c_targetCBlockSize:523case ZSTD_c_srcSizeHint:524break;525526default: RETURN_ERROR(parameter_unsupported, "unknown parameter");527}528return ZSTD_CCtxParams_setParameter(&cctx->requestedParams, param, value);529}530531size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,532ZSTD_cParameter param, int value)533{534DEBUGLOG(4, "ZSTD_CCtxParams_setParameter (%i, %i)", (int)param, value);535switch(param)536{537case ZSTD_c_format :538BOUNDCHECK(ZSTD_c_format, value);539CCtxParams->format = (ZSTD_format_e)value;540return (size_t)CCtxParams->format;541542case ZSTD_c_compressionLevel : {543FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");544if (value) { /* 0 : does not change current level */545CCtxParams->compressionLevel = value;546}547if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel;548return 0; /* return type (size_t) cannot represent negative values */549}550551case ZSTD_c_windowLog :552if (value!=0) /* 0 => use default */553BOUNDCHECK(ZSTD_c_windowLog, value);554CCtxParams->cParams.windowLog = (U32)value;555return CCtxParams->cParams.windowLog;556557case ZSTD_c_hashLog :558if (value!=0) /* 0 => use default */559BOUNDCHECK(ZSTD_c_hashLog, value);560CCtxParams->cParams.hashLog = (U32)value;561return CCtxParams->cParams.hashLog;562563case ZSTD_c_chainLog :564if (value!=0) /* 0 => use default */565BOUNDCHECK(ZSTD_c_chainLog, value);566CCtxParams->cParams.chainLog = (U32)value;567return CCtxParams->cParams.chainLog;568569case ZSTD_c_searchLog :570if (value!=0) /* 0 => use default */571BOUNDCHECK(ZSTD_c_searchLog, value);572CCtxParams->cParams.searchLog = (U32)value;573return (size_t)value;574575case ZSTD_c_minMatch :576if (value!=0) /* 0 => use default */577BOUNDCHECK(ZSTD_c_minMatch, value);578CCtxParams->cParams.minMatch = value;579return CCtxParams->cParams.minMatch;580581case ZSTD_c_targetLength :582BOUNDCHECK(ZSTD_c_targetLength, value);583CCtxParams->cParams.targetLength = value;584return CCtxParams->cParams.targetLength;585586case ZSTD_c_strategy :587if (value!=0) /* 0 => use default */588BOUNDCHECK(ZSTD_c_strategy, value);589CCtxParams->cParams.strategy = (ZSTD_strategy)value;590return (size_t)CCtxParams->cParams.strategy;591592case ZSTD_c_contentSizeFlag :593/* Content size written in frame header _when known_ (default:1) */594DEBUGLOG(4, "set content size flag = %u", (value!=0));595CCtxParams->fParams.contentSizeFlag = value != 0;596return CCtxParams->fParams.contentSizeFlag;597598case ZSTD_c_checksumFlag :599/* A 32-bits content checksum will be calculated and written at end of frame (default:0) */600CCtxParams->fParams.checksumFlag = value != 0;601return CCtxParams->fParams.checksumFlag;602603case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */604DEBUGLOG(4, "set dictIDFlag = %u", (value!=0));605CCtxParams->fParams.noDictIDFlag = !value;606return !CCtxParams->fParams.noDictIDFlag;607608case ZSTD_c_forceMaxWindow :609CCtxParams->forceWindow = (value != 0);610return CCtxParams->forceWindow;611612case ZSTD_c_forceAttachDict : {613const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value;614BOUNDCHECK(ZSTD_c_forceAttachDict, pref);615CCtxParams->attachDictPref = pref;616return CCtxParams->attachDictPref;617}618619case ZSTD_c_literalCompressionMode : {620const ZSTD_literalCompressionMode_e lcm = (ZSTD_literalCompressionMode_e)value;621BOUNDCHECK(ZSTD_c_literalCompressionMode, lcm);622CCtxParams->literalCompressionMode = lcm;623return CCtxParams->literalCompressionMode;624}625626case ZSTD_c_nbWorkers :627#ifndef ZSTD_MULTITHREAD628RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");629return 0;630#else631FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");632CCtxParams->nbWorkers = value;633return CCtxParams->nbWorkers;634#endif635636case ZSTD_c_jobSize :637#ifndef ZSTD_MULTITHREAD638RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");639return 0;640#else641/* Adjust to the minimum non-default value. */642if (value != 0 && value < ZSTDMT_JOBSIZE_MIN)643value = ZSTDMT_JOBSIZE_MIN;644FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), "");645assert(value >= 0);646CCtxParams->jobSize = value;647return CCtxParams->jobSize;648#endif649650case ZSTD_c_overlapLog :651#ifndef ZSTD_MULTITHREAD652RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");653return 0;654#else655FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");656CCtxParams->overlapLog = value;657return CCtxParams->overlapLog;658#endif659660case ZSTD_c_rsyncable :661#ifndef ZSTD_MULTITHREAD662RETURN_ERROR_IF(value!=0, parameter_unsupported, "not compiled with multithreading");663return 0;664#else665FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(ZSTD_c_overlapLog, &value), "");666CCtxParams->rsyncable = value;667return CCtxParams->rsyncable;668#endif669670case ZSTD_c_enableLongDistanceMatching :671CCtxParams->ldmParams.enableLdm = (value!=0);672return CCtxParams->ldmParams.enableLdm;673674case ZSTD_c_ldmHashLog :675if (value!=0) /* 0 ==> auto */676BOUNDCHECK(ZSTD_c_ldmHashLog, value);677CCtxParams->ldmParams.hashLog = value;678return CCtxParams->ldmParams.hashLog;679680case ZSTD_c_ldmMinMatch :681if (value!=0) /* 0 ==> default */682BOUNDCHECK(ZSTD_c_ldmMinMatch, value);683CCtxParams->ldmParams.minMatchLength = value;684return CCtxParams->ldmParams.minMatchLength;685686case ZSTD_c_ldmBucketSizeLog :687if (value!=0) /* 0 ==> default */688BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value);689CCtxParams->ldmParams.bucketSizeLog = value;690return CCtxParams->ldmParams.bucketSizeLog;691692case ZSTD_c_ldmHashRateLog :693RETURN_ERROR_IF(value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN,694parameter_outOfBound, "Param out of bounds!");695CCtxParams->ldmParams.hashRateLog = value;696return CCtxParams->ldmParams.hashRateLog;697698case ZSTD_c_targetCBlockSize :699if (value!=0) /* 0 ==> default */700BOUNDCHECK(ZSTD_c_targetCBlockSize, value);701CCtxParams->targetCBlockSize = value;702return CCtxParams->targetCBlockSize;703704case ZSTD_c_srcSizeHint :705if (value!=0) /* 0 ==> default */706BOUNDCHECK(ZSTD_c_srcSizeHint, value);707CCtxParams->srcSizeHint = value;708return CCtxParams->srcSizeHint;709710default: RETURN_ERROR(parameter_unsupported, "unknown parameter");711}712}713714size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value)715{716return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value);717}718719size_t ZSTD_CCtxParams_getParameter(720ZSTD_CCtx_params* CCtxParams, ZSTD_cParameter param, int* value)721{722switch(param)723{724case ZSTD_c_format :725*value = CCtxParams->format;726break;727case ZSTD_c_compressionLevel :728*value = CCtxParams->compressionLevel;729break;730case ZSTD_c_windowLog :731*value = (int)CCtxParams->cParams.windowLog;732break;733case ZSTD_c_hashLog :734*value = (int)CCtxParams->cParams.hashLog;735break;736case ZSTD_c_chainLog :737*value = (int)CCtxParams->cParams.chainLog;738break;739case ZSTD_c_searchLog :740*value = CCtxParams->cParams.searchLog;741break;742case ZSTD_c_minMatch :743*value = CCtxParams->cParams.minMatch;744break;745case ZSTD_c_targetLength :746*value = CCtxParams->cParams.targetLength;747break;748case ZSTD_c_strategy :749*value = (unsigned)CCtxParams->cParams.strategy;750break;751case ZSTD_c_contentSizeFlag :752*value = CCtxParams->fParams.contentSizeFlag;753break;754case ZSTD_c_checksumFlag :755*value = CCtxParams->fParams.checksumFlag;756break;757case ZSTD_c_dictIDFlag :758*value = !CCtxParams->fParams.noDictIDFlag;759break;760case ZSTD_c_forceMaxWindow :761*value = CCtxParams->forceWindow;762break;763case ZSTD_c_forceAttachDict :764*value = CCtxParams->attachDictPref;765break;766case ZSTD_c_literalCompressionMode :767*value = CCtxParams->literalCompressionMode;768break;769case ZSTD_c_nbWorkers :770#ifndef ZSTD_MULTITHREAD771assert(CCtxParams->nbWorkers == 0);772#endif773*value = CCtxParams->nbWorkers;774break;775case ZSTD_c_jobSize :776#ifndef ZSTD_MULTITHREAD777RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");778#else779assert(CCtxParams->jobSize <= INT_MAX);780*value = (int)CCtxParams->jobSize;781break;782#endif783case ZSTD_c_overlapLog :784#ifndef ZSTD_MULTITHREAD785RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");786#else787*value = CCtxParams->overlapLog;788break;789#endif790case ZSTD_c_rsyncable :791#ifndef ZSTD_MULTITHREAD792RETURN_ERROR(parameter_unsupported, "not compiled with multithreading");793#else794*value = CCtxParams->rsyncable;795break;796#endif797case ZSTD_c_enableLongDistanceMatching :798*value = CCtxParams->ldmParams.enableLdm;799break;800case ZSTD_c_ldmHashLog :801*value = CCtxParams->ldmParams.hashLog;802break;803case ZSTD_c_ldmMinMatch :804*value = CCtxParams->ldmParams.minMatchLength;805break;806case ZSTD_c_ldmBucketSizeLog :807*value = CCtxParams->ldmParams.bucketSizeLog;808break;809case ZSTD_c_ldmHashRateLog :810*value = CCtxParams->ldmParams.hashRateLog;811break;812case ZSTD_c_targetCBlockSize :813*value = (int)CCtxParams->targetCBlockSize;814break;815case ZSTD_c_srcSizeHint :816*value = (int)CCtxParams->srcSizeHint;817break;818default: RETURN_ERROR(parameter_unsupported, "unknown parameter");819}820return 0;821}822823/** ZSTD_CCtx_setParametersUsingCCtxParams() :824* just applies `params` into `cctx`825* no action is performed, parameters are merely stored.826* If ZSTDMT is enabled, parameters are pushed to cctx->mtctx.827* This is possible even if a compression is ongoing.828* In which case, new parameters will be applied on the fly, starting with next compression job.829*/830size_t ZSTD_CCtx_setParametersUsingCCtxParams(831ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params)832{833DEBUGLOG(4, "ZSTD_CCtx_setParametersUsingCCtxParams");834RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,835"The context is in the wrong stage!");836RETURN_ERROR_IF(cctx->cdict, stage_wrong,837"Can't override parameters with cdict attached (some must "838"be inherited from the cdict).");839840cctx->requestedParams = *params;841return 0;842}843844ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize)845{846DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %u bytes", (U32)pledgedSrcSize);847RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,848"Can't set pledgedSrcSize when not in init stage.");849cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;850return 0;851}852853/**854* Initializes the local dict using the requested parameters.855* NOTE: This does not use the pledged src size, because it may be used for more856* than one compression.857*/858static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx)859{860ZSTD_localDict* const dl = &cctx->localDict;861ZSTD_compressionParameters const cParams = ZSTD_getCParamsFromCCtxParams(862&cctx->requestedParams, ZSTD_CONTENTSIZE_UNKNOWN, dl->dictSize);863if (dl->dict == NULL) {864/* No local dictionary. */865assert(dl->dictBuffer == NULL);866assert(dl->cdict == NULL);867assert(dl->dictSize == 0);868return 0;869}870if (dl->cdict != NULL) {871assert(cctx->cdict == dl->cdict);872/* Local dictionary already initialized. */873return 0;874}875assert(dl->dictSize > 0);876assert(cctx->cdict == NULL);877assert(cctx->prefixDict.dict == NULL);878879dl->cdict = ZSTD_createCDict_advanced(880dl->dict,881dl->dictSize,882ZSTD_dlm_byRef,883dl->dictContentType,884cParams,885cctx->customMem);886RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed");887cctx->cdict = dl->cdict;888return 0;889}890891size_t ZSTD_CCtx_loadDictionary_advanced(892ZSTD_CCtx* cctx, const void* dict, size_t dictSize,893ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType)894{895RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,896"Can't load a dictionary when ctx is not in init stage.");897RETURN_ERROR_IF(cctx->staticSize, memory_allocation,898"no malloc for static CCtx");899DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)", (U32)dictSize);900ZSTD_clearAllDicts(cctx); /* in case one already exists */901if (dict == NULL || dictSize == 0) /* no dictionary mode */902return 0;903if (dictLoadMethod == ZSTD_dlm_byRef) {904cctx->localDict.dict = dict;905} else {906void* dictBuffer = ZSTD_malloc(dictSize, cctx->customMem);907RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!");908memcpy(dictBuffer, dict, dictSize);909cctx->localDict.dictBuffer = dictBuffer;910cctx->localDict.dict = dictBuffer;911}912cctx->localDict.dictSize = dictSize;913cctx->localDict.dictContentType = dictContentType;914return 0;915}916917ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(918ZSTD_CCtx* cctx, const void* dict, size_t dictSize)919{920return ZSTD_CCtx_loadDictionary_advanced(921cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto);922}923924ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)925{926return ZSTD_CCtx_loadDictionary_advanced(927cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto);928}929930931size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)932{933RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,934"Can't ref a dict when ctx not in init stage.");935/* Free the existing local cdict (if any) to save memory. */936ZSTD_clearAllDicts(cctx);937cctx->cdict = cdict;938return 0;939}940941size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize)942{943return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dct_rawContent);944}945946size_t ZSTD_CCtx_refPrefix_advanced(947ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType)948{949RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,950"Can't ref a prefix when ctx not in init stage.");951ZSTD_clearAllDicts(cctx);952if (prefix != NULL && prefixSize > 0) {953cctx->prefixDict.dict = prefix;954cctx->prefixDict.dictSize = prefixSize;955cctx->prefixDict.dictContentType = dictContentType;956}957return 0;958}959960/*! ZSTD_CCtx_reset() :961* Also dumps dictionary */962size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)963{964if ( (reset == ZSTD_reset_session_only)965|| (reset == ZSTD_reset_session_and_parameters) ) {966cctx->streamStage = zcss_init;967cctx->pledgedSrcSizePlusOne = 0;968}969if ( (reset == ZSTD_reset_parameters)970|| (reset == ZSTD_reset_session_and_parameters) ) {971RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong,972"Can't reset parameters only when not in init stage.");973ZSTD_clearAllDicts(cctx);974return ZSTD_CCtxParams_reset(&cctx->requestedParams);975}976return 0;977}978979980/** ZSTD_checkCParams() :981control CParam values remain within authorized range.982@return : 0, or an error code if one value is beyond authorized range */983size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)984{985BOUNDCHECK(ZSTD_c_windowLog, (int)cParams.windowLog);986BOUNDCHECK(ZSTD_c_chainLog, (int)cParams.chainLog);987BOUNDCHECK(ZSTD_c_hashLog, (int)cParams.hashLog);988BOUNDCHECK(ZSTD_c_searchLog, (int)cParams.searchLog);989BOUNDCHECK(ZSTD_c_minMatch, (int)cParams.minMatch);990BOUNDCHECK(ZSTD_c_targetLength,(int)cParams.targetLength);991BOUNDCHECK(ZSTD_c_strategy, cParams.strategy);992return 0;993}994995/** ZSTD_clampCParams() :996* make CParam values within valid range.997* @return : valid CParams */998static ZSTD_compressionParameters999ZSTD_clampCParams(ZSTD_compressionParameters cParams)1000{1001# define CLAMP_TYPE(cParam, val, type) { \1002ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); \1003if ((int)val<bounds.lowerBound) val=(type)bounds.lowerBound; \1004else if ((int)val>bounds.upperBound) val=(type)bounds.upperBound; \1005}1006# define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, unsigned)1007CLAMP(ZSTD_c_windowLog, cParams.windowLog);1008CLAMP(ZSTD_c_chainLog, cParams.chainLog);1009CLAMP(ZSTD_c_hashLog, cParams.hashLog);1010CLAMP(ZSTD_c_searchLog, cParams.searchLog);1011CLAMP(ZSTD_c_minMatch, cParams.minMatch);1012CLAMP(ZSTD_c_targetLength,cParams.targetLength);1013CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy);1014return cParams;1015}10161017/** ZSTD_cycleLog() :1018* condition for correct operation : hashLog > 1 */1019U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)1020{1021U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);1022return hashLog - btScale;1023}10241025/** ZSTD_adjustCParams_internal() :1026* optimize `cPar` for a specified input (`srcSize` and `dictSize`).1027* mostly downsize to reduce memory consumption and initialization latency.1028* `srcSize` can be ZSTD_CONTENTSIZE_UNKNOWN when not known.1029* note : `srcSize==0` means 0!1030* condition : cPar is presumed validated (can be checked using ZSTD_checkCParams()). */1031static ZSTD_compressionParameters1032ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,1033unsigned long long srcSize,1034size_t dictSize)1035{1036static const U64 minSrcSize = 513; /* (1<<9) + 1 */1037static const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1);1038assert(ZSTD_checkCParams(cPar)==0);10391040if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN)1041srcSize = minSrcSize;10421043/* resize windowLog if input is small enough, to use less memory */1044if ( (srcSize < maxWindowResize)1045&& (dictSize < maxWindowResize) ) {1046U32 const tSize = (U32)(srcSize + dictSize);1047static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN;1048U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN :1049ZSTD_highbit32(tSize-1) + 1;1050if (cPar.windowLog > srcLog) cPar.windowLog = srcLog;1051}1052if (cPar.hashLog > cPar.windowLog+1) cPar.hashLog = cPar.windowLog+1;1053{ U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);1054if (cycleLog > cPar.windowLog)1055cPar.chainLog -= (cycleLog - cPar.windowLog);1056}10571058if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN)1059cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */10601061return cPar;1062}10631064ZSTD_compressionParameters1065ZSTD_adjustCParams(ZSTD_compressionParameters cPar,1066unsigned long long srcSize,1067size_t dictSize)1068{1069cPar = ZSTD_clampCParams(cPar); /* resulting cPar is necessarily valid (all parameters within range) */1070if (srcSize == 0) srcSize = ZSTD_CONTENTSIZE_UNKNOWN;1071return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize);1072}10731074static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize);1075static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize);10761077ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(1078const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize)1079{1080ZSTD_compressionParameters cParams;1081if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) {1082srcSizeHint = CCtxParams->srcSizeHint;1083}1084cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize);1085if (CCtxParams->ldmParams.enableLdm) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;1086if (CCtxParams->cParams.windowLog) cParams.windowLog = CCtxParams->cParams.windowLog;1087if (CCtxParams->cParams.hashLog) cParams.hashLog = CCtxParams->cParams.hashLog;1088if (CCtxParams->cParams.chainLog) cParams.chainLog = CCtxParams->cParams.chainLog;1089if (CCtxParams->cParams.searchLog) cParams.searchLog = CCtxParams->cParams.searchLog;1090if (CCtxParams->cParams.minMatch) cParams.minMatch = CCtxParams->cParams.minMatch;1091if (CCtxParams->cParams.targetLength) cParams.targetLength = CCtxParams->cParams.targetLength;1092if (CCtxParams->cParams.strategy) cParams.strategy = CCtxParams->cParams.strategy;1093assert(!ZSTD_checkCParams(cParams));1094/* srcSizeHint == 0 means 0 */1095return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize);1096}10971098static size_t1099ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,1100const U32 forCCtx)1101{1102size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);1103size_t const hSize = ((size_t)1) << cParams->hashLog;1104U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;1105size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;1106/* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't1107* surrounded by redzones in ASAN. */1108size_t const tableSpace = chainSize * sizeof(U32)1109+ hSize * sizeof(U32)1110+ h3Size * sizeof(U32);1111size_t const optPotentialSpace =1112ZSTD_cwksp_alloc_size((MaxML+1) * sizeof(U32))1113+ ZSTD_cwksp_alloc_size((MaxLL+1) * sizeof(U32))1114+ ZSTD_cwksp_alloc_size((MaxOff+1) * sizeof(U32))1115+ ZSTD_cwksp_alloc_size((1<<Litbits) * sizeof(U32))1116+ ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t))1117+ ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));1118size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))1119? optPotentialSpace1120: 0;1121DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",1122(U32)chainSize, (U32)hSize, (U32)h3Size);1123return tableSpace + optSpace;1124}11251126size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)1127{1128RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");1129{ ZSTD_compressionParameters const cParams =1130ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0);1131size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);1132U32 const divider = (cParams.minMatch==3) ? 3 : 4;1133size_t const maxNbSeq = blockSize / divider;1134size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)1135+ ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))1136+ 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));1137size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);1138size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));1139size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1);11401141size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams);1142size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq));11431144/* estimateCCtxSize is for one-shot compression. So no buffers should1145* be needed. However, we still allocate two 0-sized buffers, which can1146* take space under ASAN. */1147size_t const bufferSpace = ZSTD_cwksp_alloc_size(0)1148+ ZSTD_cwksp_alloc_size(0);11491150size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx));11511152size_t const neededSpace =1153cctxSpace +1154entropySpace +1155blockStateSpace +1156ldmSpace +1157ldmSeqSpace +1158matchStateSize +1159tokenSpace +1160bufferSpace;11611162DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);1163return neededSpace;1164}1165}11661167size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams)1168{1169ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);1170return ZSTD_estimateCCtxSize_usingCCtxParams(¶ms);1171}11721173static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel)1174{1175ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0);1176return ZSTD_estimateCCtxSize_usingCParams(cParams);1177}11781179size_t ZSTD_estimateCCtxSize(int compressionLevel)1180{1181int level;1182size_t memBudget = 0;1183for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {1184size_t const newMB = ZSTD_estimateCCtxSize_internal(level);1185if (newMB > memBudget) memBudget = newMB;1186}1187return memBudget;1188}11891190size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params)1191{1192RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");1193{ ZSTD_compressionParameters const cParams =1194ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0);1195size_t const CCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(params);1196size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);1197size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize;1198size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;1199size_t const streamingSize = ZSTD_cwksp_alloc_size(inBuffSize)1200+ ZSTD_cwksp_alloc_size(outBuffSize);12011202return CCtxSize + streamingSize;1203}1204}12051206size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams)1207{1208ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);1209return ZSTD_estimateCStreamSize_usingCCtxParams(¶ms);1210}12111212static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel)1213{1214ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0);1215return ZSTD_estimateCStreamSize_usingCParams(cParams);1216}12171218size_t ZSTD_estimateCStreamSize(int compressionLevel)1219{1220int level;1221size_t memBudget = 0;1222for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) {1223size_t const newMB = ZSTD_estimateCStreamSize_internal(level);1224if (newMB > memBudget) memBudget = newMB;1225}1226return memBudget;1227}12281229/* ZSTD_getFrameProgression():1230* tells how much data has been consumed (input) and produced (output) for current frame.1231* able to count progression inside worker threads (non-blocking mode).1232*/1233ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx)1234{1235#ifdef ZSTD_MULTITHREAD1236if (cctx->appliedParams.nbWorkers > 0) {1237return ZSTDMT_getFrameProgression(cctx->mtctx);1238}1239#endif1240{ ZSTD_frameProgression fp;1241size_t const buffered = (cctx->inBuff == NULL) ? 0 :1242cctx->inBuffPos - cctx->inToCompress;1243if (buffered) assert(cctx->inBuffPos >= cctx->inToCompress);1244assert(buffered <= ZSTD_BLOCKSIZE_MAX);1245fp.ingested = cctx->consumedSrcSize + buffered;1246fp.consumed = cctx->consumedSrcSize;1247fp.produced = cctx->producedCSize;1248fp.flushed = cctx->producedCSize; /* simplified; some data might still be left within streaming output buffer */1249fp.currentJobID = 0;1250fp.nbActiveWorkers = 0;1251return fp;1252} }12531254/*! ZSTD_toFlushNow()1255* Only useful for multithreading scenarios currently (nbWorkers >= 1).1256*/1257size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx)1258{1259#ifdef ZSTD_MULTITHREAD1260if (cctx->appliedParams.nbWorkers > 0) {1261return ZSTDMT_toFlushNow(cctx->mtctx);1262}1263#endif1264(void)cctx;1265return 0; /* over-simplification; could also check if context is currently running in streaming mode, and in which case, report how many bytes are left to be flushed within output buffer */1266}12671268static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,1269ZSTD_compressionParameters cParams2)1270{1271(void)cParams1;1272(void)cParams2;1273assert(cParams1.windowLog == cParams2.windowLog);1274assert(cParams1.chainLog == cParams2.chainLog);1275assert(cParams1.hashLog == cParams2.hashLog);1276assert(cParams1.searchLog == cParams2.searchLog);1277assert(cParams1.minMatch == cParams2.minMatch);1278assert(cParams1.targetLength == cParams2.targetLength);1279assert(cParams1.strategy == cParams2.strategy);1280}12811282void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs)1283{1284int i;1285for (i = 0; i < ZSTD_REP_NUM; ++i)1286bs->rep[i] = repStartValue[i];1287bs->entropy.huf.repeatMode = HUF_repeat_none;1288bs->entropy.fse.offcode_repeatMode = FSE_repeat_none;1289bs->entropy.fse.matchlength_repeatMode = FSE_repeat_none;1290bs->entropy.fse.litlength_repeatMode = FSE_repeat_none;1291}12921293/*! ZSTD_invalidateMatchState()1294* Invalidate all the matches in the match finder tables.1295* Requires nextSrc and base to be set (can be NULL).1296*/1297static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms)1298{1299ZSTD_window_clear(&ms->window);13001301ms->nextToUpdate = ms->window.dictLimit;1302ms->loadedDictEnd = 0;1303ms->opt.litLengthSum = 0; /* force reset of btopt stats */1304ms->dictMatchState = NULL;1305}13061307/**1308* Indicates whether this compression proceeds directly from user-provided1309* source buffer to user-provided destination buffer (ZSTDb_not_buffered), or1310* whether the context needs to buffer the input/output (ZSTDb_buffered).1311*/1312typedef enum {1313ZSTDb_not_buffered,1314ZSTDb_buffered1315} ZSTD_buffered_policy_e;13161317/**1318* Controls, for this matchState reset, whether the tables need to be cleared /1319* prepared for the coming compression (ZSTDcrp_makeClean), or whether the1320* tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a1321* subsequent operation will overwrite the table space anyways (e.g., copying1322* the matchState contents in from a CDict).1323*/1324typedef enum {1325ZSTDcrp_makeClean,1326ZSTDcrp_leaveDirty1327} ZSTD_compResetPolicy_e;13281329/**1330* Controls, for this matchState reset, whether indexing can continue where it1331* left off (ZSTDirp_continue), or whether it needs to be restarted from zero1332* (ZSTDirp_reset).1333*/1334typedef enum {1335ZSTDirp_continue,1336ZSTDirp_reset1337} ZSTD_indexResetPolicy_e;13381339typedef enum {1340ZSTD_resetTarget_CDict,1341ZSTD_resetTarget_CCtx1342} ZSTD_resetTarget_e;13431344static size_t1345ZSTD_reset_matchState(ZSTD_matchState_t* ms,1346ZSTD_cwksp* ws,1347const ZSTD_compressionParameters* cParams,1348const ZSTD_compResetPolicy_e crp,1349const ZSTD_indexResetPolicy_e forceResetIndex,1350const ZSTD_resetTarget_e forWho)1351{1352size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);1353size_t const hSize = ((size_t)1) << cParams->hashLog;1354U32 const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;1355size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;13561357DEBUGLOG(4, "reset indices : %u", forceResetIndex == ZSTDirp_reset);1358if (forceResetIndex == ZSTDirp_reset) {1359ZSTD_window_init(&ms->window);1360ZSTD_cwksp_mark_tables_dirty(ws);1361}13621363ms->hashLog3 = hashLog3;13641365ZSTD_invalidateMatchState(ms);13661367assert(!ZSTD_cwksp_reserve_failed(ws)); /* check that allocation hasn't already failed */13681369ZSTD_cwksp_clear_tables(ws);13701371DEBUGLOG(5, "reserving table space");1372/* table Space */1373ms->hashTable = (U32*)ZSTD_cwksp_reserve_table(ws, hSize * sizeof(U32));1374ms->chainTable = (U32*)ZSTD_cwksp_reserve_table(ws, chainSize * sizeof(U32));1375ms->hashTable3 = (U32*)ZSTD_cwksp_reserve_table(ws, h3Size * sizeof(U32));1376RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,1377"failed a workspace allocation in ZSTD_reset_matchState");13781379DEBUGLOG(4, "reset table : %u", crp!=ZSTDcrp_leaveDirty);1380if (crp!=ZSTDcrp_leaveDirty) {1381/* reset tables only */1382ZSTD_cwksp_clean_tables(ws);1383}13841385/* opt parser space */1386if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) {1387DEBUGLOG(4, "reserving optimal parser space");1388ms->opt.litFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (1<<Litbits) * sizeof(unsigned));1389ms->opt.litLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxLL+1) * sizeof(unsigned));1390ms->opt.matchLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxML+1) * sizeof(unsigned));1391ms->opt.offCodeFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxOff+1) * sizeof(unsigned));1392ms->opt.matchTable = (ZSTD_match_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t));1393ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));1394}13951396ms->cParams = *cParams;13971398RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,1399"failed a workspace allocation in ZSTD_reset_matchState");14001401return 0;1402}14031404/* ZSTD_indexTooCloseToMax() :1405* minor optimization : prefer memset() rather than reduceIndex()1406* which is measurably slow in some circumstances (reported for Visual Studio).1407* Works when re-using a context for a lot of smallish inputs :1408* if all inputs are smaller than ZSTD_INDEXOVERFLOW_MARGIN,1409* memset() will be triggered before reduceIndex().1410*/1411#define ZSTD_INDEXOVERFLOW_MARGIN (16 MB)1412static int ZSTD_indexTooCloseToMax(ZSTD_window_t w)1413{1414return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN);1415}14161417/*! ZSTD_resetCCtx_internal() :1418note : `params` are assumed fully validated at this stage */1419static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,1420ZSTD_CCtx_params params,1421U64 const pledgedSrcSize,1422ZSTD_compResetPolicy_e const crp,1423ZSTD_buffered_policy_e const zbuff)1424{1425ZSTD_cwksp* const ws = &zc->workspace;1426DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u",1427(U32)pledgedSrcSize, params.cParams.windowLog);1428assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));14291430zc->isFirstBlock = 1;14311432if (params.ldmParams.enableLdm) {1433/* Adjust long distance matching parameters */1434ZSTD_ldm_adjustParameters(¶ms.ldmParams, ¶ms.cParams);1435assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);1436assert(params.ldmParams.hashRateLog < 32);1437zc->ldmState.hashPower = ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength);1438}14391440{ size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params.cParams.windowLog), pledgedSrcSize));1441size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);1442U32 const divider = (params.cParams.minMatch==3) ? 3 : 4;1443size_t const maxNbSeq = blockSize / divider;1444size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)1445+ ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))1446+ 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));1447size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;1448size_t const buffInSize = (zbuff==ZSTDb_buffered) ? windowSize + blockSize : 0;1449size_t const matchStateSize = ZSTD_sizeof_matchState(¶ms.cParams, /* forCCtx */ 1);1450size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params.ldmParams, blockSize);14511452ZSTD_indexResetPolicy_e needsIndexReset = zc->initialized ? ZSTDirp_continue : ZSTDirp_reset;14531454if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) {1455needsIndexReset = ZSTDirp_reset;1456}14571458if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);14591460/* Check if workspace is large enough, alloc a new one if needed */1461{ size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;1462size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);1463size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));1464size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + ZSTD_cwksp_alloc_size(buffOutSize);1465size_t const ldmSpace = ZSTD_ldm_getTableSize(params.ldmParams);1466size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq));14671468size_t const neededSpace =1469cctxSpace +1470entropySpace +1471blockStateSpace +1472ldmSpace +1473ldmSeqSpace +1474matchStateSize +1475tokenSpace +1476bufferSpace;14771478int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;1479int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);14801481DEBUGLOG(4, "Need %zuKB workspace, including %zuKB for match state, and %zuKB for buffers",1482neededSpace>>10, matchStateSize>>10, bufferSpace>>10);1483DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);14841485if (workspaceTooSmall || workspaceWasteful) {1486DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB",1487ZSTD_cwksp_sizeof(ws) >> 10,1488neededSpace >> 10);14891490RETURN_ERROR_IF(zc->staticSize, memory_allocation, "static cctx : no resize");14911492needsIndexReset = ZSTDirp_reset;14931494ZSTD_cwksp_free(ws, zc->customMem);1495FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem), "");14961497DEBUGLOG(5, "reserving object space");1498/* Statically sized space.1499* entropyWorkspace never moves,1500* though prev/next block swap places */1501assert(ZSTD_cwksp_check_available(ws, 2 * sizeof(ZSTD_compressedBlockState_t)));1502zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));1503RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock");1504zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));1505RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock");1506zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, HUF_WORKSPACE_SIZE);1507RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate entropyWorkspace");1508} }15091510ZSTD_cwksp_clear(ws);15111512/* init params */1513zc->appliedParams = params;1514zc->blockState.matchState.cParams = params.cParams;1515zc->pledgedSrcSizePlusOne = pledgedSrcSize+1;1516zc->consumedSrcSize = 0;1517zc->producedCSize = 0;1518if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)1519zc->appliedParams.fParams.contentSizeFlag = 0;1520DEBUGLOG(4, "pledged content size : %u ; flag : %u",1521(unsigned)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);1522zc->blockSize = blockSize;15231524XXH64_reset(&zc->xxhState, 0);1525zc->stage = ZSTDcs_init;1526zc->dictID = 0;15271528ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock);15291530/* ZSTD_wildcopy() is used to copy into the literals buffer,1531* so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.1532*/1533zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);1534zc->seqStore.maxNbLit = blockSize;15351536/* buffers */1537zc->inBuffSize = buffInSize;1538zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);1539zc->outBuffSize = buffOutSize;1540zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);15411542/* ldm bucketOffsets table */1543if (params.ldmParams.enableLdm) {1544/* TODO: avoid memset? */1545size_t const ldmBucketSize =1546((size_t)1) << (params.ldmParams.hashLog -1547params.ldmParams.bucketSizeLog);1548zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, ldmBucketSize);1549memset(zc->ldmState.bucketOffsets, 0, ldmBucketSize);1550}15511552/* sequences storage */1553ZSTD_referenceExternalSequences(zc, NULL, 0);1554zc->seqStore.maxNbSeq = maxNbSeq;1555zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));1556zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));1557zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));1558zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));15591560FORWARD_IF_ERROR(ZSTD_reset_matchState(1561&zc->blockState.matchState,1562ws,1563¶ms.cParams,1564crp,1565needsIndexReset,1566ZSTD_resetTarget_CCtx), "");15671568/* ldm hash table */1569if (params.ldmParams.enableLdm) {1570/* TODO: avoid memset? */1571size_t const ldmHSize = ((size_t)1) << params.ldmParams.hashLog;1572zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t));1573memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t));1574zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq));1575zc->maxNbLdmSequences = maxNbLdmSeq;15761577ZSTD_window_init(&zc->ldmState.window);1578ZSTD_window_clear(&zc->ldmState.window);1579zc->ldmState.loadedDictEnd = 0;1580}15811582DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));1583zc->initialized = 1;15841585return 0;1586}1587}15881589/* ZSTD_invalidateRepCodes() :1590* ensures next compression will not use repcodes from previous block.1591* Note : only works with regular variant;1592* do not use with extDict variant ! */1593void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) {1594int i;1595for (i=0; i<ZSTD_REP_NUM; i++) cctx->blockState.prevCBlock->rep[i] = 0;1596assert(!ZSTD_window_hasExtDict(cctx->blockState.matchState.window));1597}15981599/* These are the approximate sizes for each strategy past which copying the1600* dictionary tables into the working context is faster than using them1601* in-place.1602*/1603static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = {16048 KB, /* unused */16058 KB, /* ZSTD_fast */160616 KB, /* ZSTD_dfast */160732 KB, /* ZSTD_greedy */160832 KB, /* ZSTD_lazy */160932 KB, /* ZSTD_lazy2 */161032 KB, /* ZSTD_btlazy2 */161132 KB, /* ZSTD_btopt */16128 KB, /* ZSTD_btultra */16138 KB /* ZSTD_btultra2 */1614};16151616static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict,1617const ZSTD_CCtx_params* params,1618U64 pledgedSrcSize)1619{1620size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy];1621return ( pledgedSrcSize <= cutoff1622|| pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN1623|| params->attachDictPref == ZSTD_dictForceAttach )1624&& params->attachDictPref != ZSTD_dictForceCopy1625&& !params->forceWindow; /* dictMatchState isn't correctly1626* handled in _enforceMaxDist */1627}16281629static size_t1630ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,1631const ZSTD_CDict* cdict,1632ZSTD_CCtx_params params,1633U64 pledgedSrcSize,1634ZSTD_buffered_policy_e zbuff)1635{1636{ const ZSTD_compressionParameters* const cdict_cParams = &cdict->matchState.cParams;1637unsigned const windowLog = params.cParams.windowLog;1638assert(windowLog != 0);1639/* Resize working context table params for input only, since the dict1640* has its own tables. */1641/* pledgeSrcSize == 0 means 0! */1642params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0);1643params.cParams.windowLog = windowLog;1644FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,1645ZSTDcrp_makeClean, zbuff), "");1646assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);1647}16481649{ const U32 cdictEnd = (U32)( cdict->matchState.window.nextSrc1650- cdict->matchState.window.base);1651const U32 cdictLen = cdictEnd - cdict->matchState.window.dictLimit;1652if (cdictLen == 0) {1653/* don't even attach dictionaries with no contents */1654DEBUGLOG(4, "skipping attaching empty dictionary");1655} else {1656DEBUGLOG(4, "attaching dictionary into context");1657cctx->blockState.matchState.dictMatchState = &cdict->matchState;16581659/* prep working match state so dict matches never have negative indices1660* when they are translated to the working context's index space. */1661if (cctx->blockState.matchState.window.dictLimit < cdictEnd) {1662cctx->blockState.matchState.window.nextSrc =1663cctx->blockState.matchState.window.base + cdictEnd;1664ZSTD_window_clear(&cctx->blockState.matchState.window);1665}1666/* loadedDictEnd is expressed within the referential of the active context */1667cctx->blockState.matchState.loadedDictEnd = cctx->blockState.matchState.window.dictLimit;1668} }16691670cctx->dictID = cdict->dictID;16711672/* copy block state */1673memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));16741675return 0;1676}16771678static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,1679const ZSTD_CDict* cdict,1680ZSTD_CCtx_params params,1681U64 pledgedSrcSize,1682ZSTD_buffered_policy_e zbuff)1683{1684const ZSTD_compressionParameters *cdict_cParams = &cdict->matchState.cParams;16851686DEBUGLOG(4, "copying dictionary into context");16871688{ unsigned const windowLog = params.cParams.windowLog;1689assert(windowLog != 0);1690/* Copy only compression parameters related to tables. */1691params.cParams = *cdict_cParams;1692params.cParams.windowLog = windowLog;1693FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,1694ZSTDcrp_leaveDirty, zbuff), "");1695assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);1696assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog);1697assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog);1698}16991700ZSTD_cwksp_mark_tables_dirty(&cctx->workspace);17011702/* copy tables */1703{ size_t const chainSize = (cdict_cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cdict_cParams->chainLog);1704size_t const hSize = (size_t)1 << cdict_cParams->hashLog;17051706memcpy(cctx->blockState.matchState.hashTable,1707cdict->matchState.hashTable,1708hSize * sizeof(U32));1709memcpy(cctx->blockState.matchState.chainTable,1710cdict->matchState.chainTable,1711chainSize * sizeof(U32));1712}17131714/* Zero the hashTable3, since the cdict never fills it */1715{ int const h3log = cctx->blockState.matchState.hashLog3;1716size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;1717assert(cdict->matchState.hashLog3 == 0);1718memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32));1719}17201721ZSTD_cwksp_mark_tables_clean(&cctx->workspace);17221723/* copy dictionary offsets */1724{ ZSTD_matchState_t const* srcMatchState = &cdict->matchState;1725ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState;1726dstMatchState->window = srcMatchState->window;1727dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;1728dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;1729}17301731cctx->dictID = cdict->dictID;17321733/* copy block state */1734memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState));17351736return 0;1737}17381739/* We have a choice between copying the dictionary context into the working1740* context, or referencing the dictionary context from the working context1741* in-place. We decide here which strategy to use. */1742static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,1743const ZSTD_CDict* cdict,1744const ZSTD_CCtx_params* params,1745U64 pledgedSrcSize,1746ZSTD_buffered_policy_e zbuff)1747{17481749DEBUGLOG(4, "ZSTD_resetCCtx_usingCDict (pledgedSrcSize=%u)",1750(unsigned)pledgedSrcSize);17511752if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) {1753return ZSTD_resetCCtx_byAttachingCDict(1754cctx, cdict, *params, pledgedSrcSize, zbuff);1755} else {1756return ZSTD_resetCCtx_byCopyingCDict(1757cctx, cdict, *params, pledgedSrcSize, zbuff);1758}1759}17601761/*! ZSTD_copyCCtx_internal() :1762* Duplicate an existing context `srcCCtx` into another one `dstCCtx`.1763* Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).1764* The "context", in this case, refers to the hash and chain tables,1765* entropy tables, and dictionary references.1766* `windowLog` value is enforced if != 0, otherwise value is copied from srcCCtx.1767* @return : 0, or an error code */1768static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,1769const ZSTD_CCtx* srcCCtx,1770ZSTD_frameParameters fParams,1771U64 pledgedSrcSize,1772ZSTD_buffered_policy_e zbuff)1773{1774DEBUGLOG(5, "ZSTD_copyCCtx_internal");1775RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong,1776"Can't copy a ctx that's not in init stage.");17771778memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));1779{ ZSTD_CCtx_params params = dstCCtx->requestedParams;1780/* Copy only compression parameters related to tables. */1781params.cParams = srcCCtx->appliedParams.cParams;1782params.fParams = fParams;1783ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize,1784ZSTDcrp_leaveDirty, zbuff);1785assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog);1786assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy);1787assert(dstCCtx->appliedParams.cParams.hashLog == srcCCtx->appliedParams.cParams.hashLog);1788assert(dstCCtx->appliedParams.cParams.chainLog == srcCCtx->appliedParams.cParams.chainLog);1789assert(dstCCtx->blockState.matchState.hashLog3 == srcCCtx->blockState.matchState.hashLog3);1790}17911792ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace);17931794/* copy tables */1795{ size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog);1796size_t const hSize = (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;1797int const h3log = srcCCtx->blockState.matchState.hashLog3;1798size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;17991800memcpy(dstCCtx->blockState.matchState.hashTable,1801srcCCtx->blockState.matchState.hashTable,1802hSize * sizeof(U32));1803memcpy(dstCCtx->blockState.matchState.chainTable,1804srcCCtx->blockState.matchState.chainTable,1805chainSize * sizeof(U32));1806memcpy(dstCCtx->blockState.matchState.hashTable3,1807srcCCtx->blockState.matchState.hashTable3,1808h3Size * sizeof(U32));1809}18101811ZSTD_cwksp_mark_tables_clean(&dstCCtx->workspace);18121813/* copy dictionary offsets */1814{1815const ZSTD_matchState_t* srcMatchState = &srcCCtx->blockState.matchState;1816ZSTD_matchState_t* dstMatchState = &dstCCtx->blockState.matchState;1817dstMatchState->window = srcMatchState->window;1818dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;1819dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd;1820}1821dstCCtx->dictID = srcCCtx->dictID;18221823/* copy block state */1824memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock));18251826return 0;1827}18281829/*! ZSTD_copyCCtx() :1830* Duplicate an existing context `srcCCtx` into another one `dstCCtx`.1831* Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).1832* pledgedSrcSize==0 means "unknown".1833* @return : 0, or an error code */1834size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize)1835{1836ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };1837ZSTD_buffered_policy_e const zbuff = (ZSTD_buffered_policy_e)(srcCCtx->inBuffSize>0);1838ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1);1839if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;1840fParams.contentSizeFlag = (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN);18411842return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx,1843fParams, pledgedSrcSize,1844zbuff);1845}184618471848#define ZSTD_ROWSIZE 161849/*! ZSTD_reduceTable() :1850* reduce table indexes by `reducerValue`, or squash to zero.1851* PreserveMark preserves "unsorted mark" for btlazy2 strategy.1852* It must be set to a clear 0/1 value, to remove branch during inlining.1853* Presume table size is a multiple of ZSTD_ROWSIZE1854* to help auto-vectorization */1855FORCE_INLINE_TEMPLATE void1856ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerValue, int const preserveMark)1857{1858int const nbRows = (int)size / ZSTD_ROWSIZE;1859int cellNb = 0;1860int rowNb;1861assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */1862assert(size < (1U<<31)); /* can be casted to int */18631864#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)1865/* To validate that the table re-use logic is sound, and that we don't1866* access table space that we haven't cleaned, we re-"poison" the table1867* space every time we mark it dirty.1868*1869* This function however is intended to operate on those dirty tables and1870* re-clean them. So when this function is used correctly, we can unpoison1871* the memory it operated on. This introduces a blind spot though, since1872* if we now try to operate on __actually__ poisoned memory, we will not1873* detect that. */1874__msan_unpoison(table, size * sizeof(U32));1875#endif18761877for (rowNb=0 ; rowNb < nbRows ; rowNb++) {1878int column;1879for (column=0; column<ZSTD_ROWSIZE; column++) {1880if (preserveMark) {1881U32 const adder = (table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) ? reducerValue : 0;1882table[cellNb] += adder;1883}1884if (table[cellNb] < reducerValue) table[cellNb] = 0;1885else table[cellNb] -= reducerValue;1886cellNb++;1887} }1888}18891890static void ZSTD_reduceTable(U32* const table, U32 const size, U32 const reducerValue)1891{1892ZSTD_reduceTable_internal(table, size, reducerValue, 0);1893}18941895static void ZSTD_reduceTable_btlazy2(U32* const table, U32 const size, U32 const reducerValue)1896{1897ZSTD_reduceTable_internal(table, size, reducerValue, 1);1898}18991900/*! ZSTD_reduceIndex() :1901* rescale all indexes to avoid future overflow (indexes are U32) */1902static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* params, const U32 reducerValue)1903{1904{ U32 const hSize = (U32)1 << params->cParams.hashLog;1905ZSTD_reduceTable(ms->hashTable, hSize, reducerValue);1906}19071908if (params->cParams.strategy != ZSTD_fast) {1909U32 const chainSize = (U32)1 << params->cParams.chainLog;1910if (params->cParams.strategy == ZSTD_btlazy2)1911ZSTD_reduceTable_btlazy2(ms->chainTable, chainSize, reducerValue);1912else1913ZSTD_reduceTable(ms->chainTable, chainSize, reducerValue);1914}19151916if (ms->hashLog3) {1917U32 const h3Size = (U32)1 << ms->hashLog3;1918ZSTD_reduceTable(ms->hashTable3, h3Size, reducerValue);1919}1920}192119221923/*-*******************************************************1924* Block entropic compression1925*********************************************************/19261927/* See doc/zstd_compression_format.md for detailed format description */19281929void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)1930{1931const seqDef* const sequences = seqStorePtr->sequencesStart;1932BYTE* const llCodeTable = seqStorePtr->llCode;1933BYTE* const ofCodeTable = seqStorePtr->ofCode;1934BYTE* const mlCodeTable = seqStorePtr->mlCode;1935U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);1936U32 u;1937assert(nbSeq <= seqStorePtr->maxNbSeq);1938for (u=0; u<nbSeq; u++) {1939U32 const llv = sequences[u].litLength;1940U32 const mlv = sequences[u].matchLength;1941llCodeTable[u] = (BYTE)ZSTD_LLcode(llv);1942ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offset);1943mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv);1944}1945if (seqStorePtr->longLengthID==1)1946llCodeTable[seqStorePtr->longLengthPos] = MaxLL;1947if (seqStorePtr->longLengthID==2)1948mlCodeTable[seqStorePtr->longLengthPos] = MaxML;1949}19501951/* ZSTD_useTargetCBlockSize():1952* Returns if target compressed block size param is being used.1953* If used, compression will do best effort to make a compressed block size to be around targetCBlockSize.1954* Returns 1 if true, 0 otherwise. */1955static int ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params* cctxParams)1956{1957DEBUGLOG(5, "ZSTD_useTargetCBlockSize (targetCBlockSize=%zu)", cctxParams->targetCBlockSize);1958return (cctxParams->targetCBlockSize != 0);1959}19601961/* ZSTD_compressSequences_internal():1962* actually compresses both literals and sequences */1963MEM_STATIC size_t1964ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,1965const ZSTD_entropyCTables_t* prevEntropy,1966ZSTD_entropyCTables_t* nextEntropy,1967const ZSTD_CCtx_params* cctxParams,1968void* dst, size_t dstCapacity,1969void* entropyWorkspace, size_t entropyWkspSize,1970const int bmi2)1971{1972const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;1973ZSTD_strategy const strategy = cctxParams->cParams.strategy;1974unsigned count[MaxSeq+1];1975FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable;1976FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable;1977FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable;1978U32 LLtype, Offtype, MLtype; /* compressed, raw or rle */1979const seqDef* const sequences = seqStorePtr->sequencesStart;1980const BYTE* const ofCodeTable = seqStorePtr->ofCode;1981const BYTE* const llCodeTable = seqStorePtr->llCode;1982const BYTE* const mlCodeTable = seqStorePtr->mlCode;1983BYTE* const ostart = (BYTE*)dst;1984BYTE* const oend = ostart + dstCapacity;1985BYTE* op = ostart;1986size_t const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart);1987BYTE* seqHead;1988BYTE* lastNCount = NULL;19891990DEBUGLOG(5, "ZSTD_compressSequences_internal (nbSeq=%zu)", nbSeq);1991ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));19921993/* Compress literals */1994{ const BYTE* const literals = seqStorePtr->litStart;1995size_t const litSize = (size_t)(seqStorePtr->lit - literals);1996size_t const cSize = ZSTD_compressLiterals(1997&prevEntropy->huf, &nextEntropy->huf,1998cctxParams->cParams.strategy,1999ZSTD_disableLiteralsCompression(cctxParams),2000op, dstCapacity,2001literals, litSize,2002entropyWorkspace, entropyWkspSize,2003bmi2);2004FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed");2005assert(cSize <= dstCapacity);2006op += cSize;2007}20082009/* Sequences Header */2010RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,2011dstSize_tooSmall, "Can't fit seq hdr in output buf!");2012if (nbSeq < 128) {2013*op++ = (BYTE)nbSeq;2014} else if (nbSeq < LONGNBSEQ) {2015op[0] = (BYTE)((nbSeq>>8) + 0x80);2016op[1] = (BYTE)nbSeq;2017op+=2;2018} else {2019op[0]=0xFF;2020MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ));2021op+=3;2022}2023assert(op <= oend);2024if (nbSeq==0) {2025/* Copy the old tables over as if we repeated them */2026memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse));2027return (size_t)(op - ostart);2028}20292030/* seqHead : flags for FSE encoding type */2031seqHead = op++;2032assert(op <= oend);20332034/* convert length/distances into codes */2035ZSTD_seqToCodes(seqStorePtr);2036/* build CTable for Literal Lengths */2037{ unsigned max = MaxLL;2038size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */2039DEBUGLOG(5, "Building LL table");2040nextEntropy->fse.litlength_repeatMode = prevEntropy->fse.litlength_repeatMode;2041LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode,2042count, max, mostFrequent, nbSeq,2043LLFSELog, prevEntropy->fse.litlengthCTable,2044LL_defaultNorm, LL_defaultNormLog,2045ZSTD_defaultAllowed, strategy);2046assert(set_basic < set_compressed && set_rle < set_compressed);2047assert(!(LLtype < set_compressed && nextEntropy->fse.litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */2048{ size_t const countSize = ZSTD_buildCTable(2049op, (size_t)(oend - op),2050CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,2051count, max, llCodeTable, nbSeq,2052LL_defaultNorm, LL_defaultNormLog, MaxLL,2053prevEntropy->fse.litlengthCTable,2054sizeof(prevEntropy->fse.litlengthCTable),2055entropyWorkspace, entropyWkspSize);2056FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");2057if (LLtype == set_compressed)2058lastNCount = op;2059op += countSize;2060assert(op <= oend);2061} }2062/* build CTable for Offsets */2063{ unsigned max = MaxOff;2064size_t const mostFrequent = HIST_countFast_wksp(2065count, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */2066/* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */2067ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;2068DEBUGLOG(5, "Building OF table");2069nextEntropy->fse.offcode_repeatMode = prevEntropy->fse.offcode_repeatMode;2070Offtype = ZSTD_selectEncodingType(&nextEntropy->fse.offcode_repeatMode,2071count, max, mostFrequent, nbSeq,2072OffFSELog, prevEntropy->fse.offcodeCTable,2073OF_defaultNorm, OF_defaultNormLog,2074defaultPolicy, strategy);2075assert(!(Offtype < set_compressed && nextEntropy->fse.offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */2076{ size_t const countSize = ZSTD_buildCTable(2077op, (size_t)(oend - op),2078CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,2079count, max, ofCodeTable, nbSeq,2080OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,2081prevEntropy->fse.offcodeCTable,2082sizeof(prevEntropy->fse.offcodeCTable),2083entropyWorkspace, entropyWkspSize);2084FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");2085if (Offtype == set_compressed)2086lastNCount = op;2087op += countSize;2088assert(op <= oend);2089} }2090/* build CTable for MatchLengths */2091{ unsigned max = MaxML;2092size_t const mostFrequent = HIST_countFast_wksp(2093count, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */2094DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));2095nextEntropy->fse.matchlength_repeatMode = prevEntropy->fse.matchlength_repeatMode;2096MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode,2097count, max, mostFrequent, nbSeq,2098MLFSELog, prevEntropy->fse.matchlengthCTable,2099ML_defaultNorm, ML_defaultNormLog,2100ZSTD_defaultAllowed, strategy);2101assert(!(MLtype < set_compressed && nextEntropy->fse.matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */2102{ size_t const countSize = ZSTD_buildCTable(2103op, (size_t)(oend - op),2104CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,2105count, max, mlCodeTable, nbSeq,2106ML_defaultNorm, ML_defaultNormLog, MaxML,2107prevEntropy->fse.matchlengthCTable,2108sizeof(prevEntropy->fse.matchlengthCTable),2109entropyWorkspace, entropyWkspSize);2110FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");2111if (MLtype == set_compressed)2112lastNCount = op;2113op += countSize;2114assert(op <= oend);2115} }21162117*seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));21182119{ size_t const bitstreamSize = ZSTD_encodeSequences(2120op, (size_t)(oend - op),2121CTable_MatchLength, mlCodeTable,2122CTable_OffsetBits, ofCodeTable,2123CTable_LitLength, llCodeTable,2124sequences, nbSeq,2125longOffsets, bmi2);2126FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");2127op += bitstreamSize;2128assert(op <= oend);2129/* zstd versions <= 1.3.4 mistakenly report corruption when2130* FSE_readNCount() receives a buffer < 4 bytes.2131* Fixed by https://github.com/facebook/zstd/pull/1146.2132* This can happen when the last set_compressed table present is 22133* bytes and the bitstream is only one byte.2134* In this exceedingly rare case, we will simply emit an uncompressed2135* block, since it isn't worth optimizing.2136*/2137if (lastNCount && (op - lastNCount) < 4) {2138/* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */2139assert(op - lastNCount == 3);2140DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "2141"emitting an uncompressed block.");2142return 0;2143}2144}21452146DEBUGLOG(5, "compressed block size : %u", (unsigned)(op - ostart));2147return (size_t)(op - ostart);2148}21492150MEM_STATIC size_t2151ZSTD_compressSequences(seqStore_t* seqStorePtr,2152const ZSTD_entropyCTables_t* prevEntropy,2153ZSTD_entropyCTables_t* nextEntropy,2154const ZSTD_CCtx_params* cctxParams,2155void* dst, size_t dstCapacity,2156size_t srcSize,2157void* entropyWorkspace, size_t entropyWkspSize,2158int bmi2)2159{2160size_t const cSize = ZSTD_compressSequences_internal(2161seqStorePtr, prevEntropy, nextEntropy, cctxParams,2162dst, dstCapacity,2163entropyWorkspace, entropyWkspSize, bmi2);2164if (cSize == 0) return 0;2165/* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.2166* Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.2167*/2168if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity))2169return 0; /* block not compressed */2170FORWARD_IF_ERROR(cSize, "ZSTD_compressSequences_internal failed");21712172/* Check compressibility */2173{ size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy);2174if (cSize >= maxCSize) return 0; /* block not compressed */2175}21762177return cSize;2178}21792180/* ZSTD_selectBlockCompressor() :2181* Not static, but internal use only (used by long distance matcher)2182* assumption : strat is a valid strategy */2183ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMode_e dictMode)2184{2185static const ZSTD_blockCompressor blockCompressor[3][ZSTD_STRATEGY_MAX+1] = {2186{ ZSTD_compressBlock_fast /* default for 0 */,2187ZSTD_compressBlock_fast,2188ZSTD_compressBlock_doubleFast,2189ZSTD_compressBlock_greedy,2190ZSTD_compressBlock_lazy,2191ZSTD_compressBlock_lazy2,2192ZSTD_compressBlock_btlazy2,2193ZSTD_compressBlock_btopt,2194ZSTD_compressBlock_btultra,2195ZSTD_compressBlock_btultra2 },2196{ ZSTD_compressBlock_fast_extDict /* default for 0 */,2197ZSTD_compressBlock_fast_extDict,2198ZSTD_compressBlock_doubleFast_extDict,2199ZSTD_compressBlock_greedy_extDict,2200ZSTD_compressBlock_lazy_extDict,2201ZSTD_compressBlock_lazy2_extDict,2202ZSTD_compressBlock_btlazy2_extDict,2203ZSTD_compressBlock_btopt_extDict,2204ZSTD_compressBlock_btultra_extDict,2205ZSTD_compressBlock_btultra_extDict },2206{ ZSTD_compressBlock_fast_dictMatchState /* default for 0 */,2207ZSTD_compressBlock_fast_dictMatchState,2208ZSTD_compressBlock_doubleFast_dictMatchState,2209ZSTD_compressBlock_greedy_dictMatchState,2210ZSTD_compressBlock_lazy_dictMatchState,2211ZSTD_compressBlock_lazy2_dictMatchState,2212ZSTD_compressBlock_btlazy2_dictMatchState,2213ZSTD_compressBlock_btopt_dictMatchState,2214ZSTD_compressBlock_btultra_dictMatchState,2215ZSTD_compressBlock_btultra_dictMatchState }2216};2217ZSTD_blockCompressor selectedCompressor;2218ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);22192220assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));2221selectedCompressor = blockCompressor[(int)dictMode][(int)strat];2222assert(selectedCompressor != NULL);2223return selectedCompressor;2224}22252226static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr,2227const BYTE* anchor, size_t lastLLSize)2228{2229memcpy(seqStorePtr->lit, anchor, lastLLSize);2230seqStorePtr->lit += lastLLSize;2231}22322233void ZSTD_resetSeqStore(seqStore_t* ssPtr)2234{2235ssPtr->lit = ssPtr->litStart;2236ssPtr->sequences = ssPtr->sequencesStart;2237ssPtr->longLengthID = 0;2238}22392240typedef enum { ZSTDbss_compress, ZSTDbss_noCompress } ZSTD_buildSeqStore_e;22412242static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)2243{2244ZSTD_matchState_t* const ms = &zc->blockState.matchState;2245DEBUGLOG(5, "ZSTD_buildSeqStore (srcSize=%zu)", srcSize);2246assert(srcSize <= ZSTD_BLOCKSIZE_MAX);2247/* Assert that we have correctly flushed the ctx params into the ms's copy */2248ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams);2249if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) {2250ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch);2251return ZSTDbss_noCompress; /* don't even attempt compression below a certain srcSize */2252}2253ZSTD_resetSeqStore(&(zc->seqStore));2254/* required for optimal parser to read stats from dictionary */2255ms->opt.symbolCosts = &zc->blockState.prevCBlock->entropy;2256/* tell the optimal parser how we expect to compress literals */2257ms->opt.literalCompressionMode = zc->appliedParams.literalCompressionMode;2258/* a gap between an attached dict and the current window is not safe,2259* they must remain adjacent,2260* and when that stops being the case, the dict must be unset */2261assert(ms->dictMatchState == NULL || ms->loadedDictEnd == ms->window.dictLimit);22622263/* limited update after a very long match */2264{ const BYTE* const base = ms->window.base;2265const BYTE* const istart = (const BYTE*)src;2266const U32 current = (U32)(istart-base);2267if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */2268if (current > ms->nextToUpdate + 384)2269ms->nextToUpdate = current - MIN(192, (U32)(current - ms->nextToUpdate - 384));2270}22712272/* select and store sequences */2273{ ZSTD_dictMode_e const dictMode = ZSTD_matchState_dictMode(ms);2274size_t lastLLSize;2275{ int i;2276for (i = 0; i < ZSTD_REP_NUM; ++i)2277zc->blockState.nextCBlock->rep[i] = zc->blockState.prevCBlock->rep[i];2278}2279if (zc->externSeqStore.pos < zc->externSeqStore.size) {2280assert(!zc->appliedParams.ldmParams.enableLdm);2281/* Updates ldmSeqStore.pos */2282lastLLSize =2283ZSTD_ldm_blockCompress(&zc->externSeqStore,2284ms, &zc->seqStore,2285zc->blockState.nextCBlock->rep,2286src, srcSize);2287assert(zc->externSeqStore.pos <= zc->externSeqStore.size);2288} else if (zc->appliedParams.ldmParams.enableLdm) {2289rawSeqStore_t ldmSeqStore = {NULL, 0, 0, 0};22902291ldmSeqStore.seq = zc->ldmSequences;2292ldmSeqStore.capacity = zc->maxNbLdmSequences;2293/* Updates ldmSeqStore.size */2294FORWARD_IF_ERROR(ZSTD_ldm_generateSequences(&zc->ldmState, &ldmSeqStore,2295&zc->appliedParams.ldmParams,2296src, srcSize), "");2297/* Updates ldmSeqStore.pos */2298lastLLSize =2299ZSTD_ldm_blockCompress(&ldmSeqStore,2300ms, &zc->seqStore,2301zc->blockState.nextCBlock->rep,2302src, srcSize);2303assert(ldmSeqStore.pos == ldmSeqStore.size);2304} else { /* not long range mode */2305ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, dictMode);2306lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize);2307}2308{ const BYTE* const lastLiterals = (const BYTE*)src + srcSize - lastLLSize;2309ZSTD_storeLastLiterals(&zc->seqStore, lastLiterals, lastLLSize);2310} }2311return ZSTDbss_compress;2312}23132314static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc)2315{2316const seqStore_t* seqStore = ZSTD_getSeqStore(zc);2317const seqDef* seqs = seqStore->sequencesStart;2318size_t seqsSize = seqStore->sequences - seqs;23192320ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];2321size_t i; size_t position; int repIdx;23222323assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);2324for (i = 0, position = 0; i < seqsSize; ++i) {2325outSeqs[i].offset = seqs[i].offset;2326outSeqs[i].litLength = seqs[i].litLength;2327outSeqs[i].matchLength = seqs[i].matchLength + MINMATCH;23282329if (i == seqStore->longLengthPos) {2330if (seqStore->longLengthID == 1) {2331outSeqs[i].litLength += 0x10000;2332} else if (seqStore->longLengthID == 2) {2333outSeqs[i].matchLength += 0x10000;2334}2335}23362337if (outSeqs[i].offset <= ZSTD_REP_NUM) {2338outSeqs[i].rep = outSeqs[i].offset;2339repIdx = (unsigned int)i - outSeqs[i].offset;23402341if (outSeqs[i].litLength == 0) {2342if (outSeqs[i].offset < 3) {2343--repIdx;2344} else {2345repIdx = (unsigned int)i - 1;2346}2347++outSeqs[i].rep;2348}2349assert(repIdx >= -3);2350outSeqs[i].offset = repIdx >= 0 ? outSeqs[repIdx].offset : repStartValue[-repIdx - 1];2351if (outSeqs[i].rep == 4) {2352--outSeqs[i].offset;2353}2354} else {2355outSeqs[i].offset -= ZSTD_REP_NUM;2356}23572358position += outSeqs[i].litLength;2359outSeqs[i].matchPos = (unsigned int)position;2360position += outSeqs[i].matchLength;2361}2362zc->seqCollector.seqIndex += seqsSize;2363}23642365size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,2366size_t outSeqsSize, const void* src, size_t srcSize)2367{2368const size_t dstCapacity = ZSTD_compressBound(srcSize);2369void* dst = ZSTD_malloc(dstCapacity, ZSTD_defaultCMem);2370SeqCollector seqCollector;23712372RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!");23732374seqCollector.collectSequences = 1;2375seqCollector.seqStart = outSeqs;2376seqCollector.seqIndex = 0;2377seqCollector.maxSequences = outSeqsSize;2378zc->seqCollector = seqCollector;23792380ZSTD_compress2(zc, dst, dstCapacity, src, srcSize);2381ZSTD_free(dst, ZSTD_defaultCMem);2382return zc->seqCollector.seqIndex;2383}23842385/* Returns true if the given block is a RLE block */2386static int ZSTD_isRLE(const BYTE *ip, size_t length) {2387size_t i;2388if (length < 2) return 1;2389for (i = 1; i < length; ++i) {2390if (ip[0] != ip[i]) return 0;2391}2392return 1;2393}23942395/* Returns true if the given block may be RLE.2396* This is just a heuristic based on the compressibility.2397* It may return both false positives and false negatives.2398*/2399static int ZSTD_maybeRLE(seqStore_t const* seqStore)2400{2401size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart);2402size_t const nbLits = (size_t)(seqStore->lit - seqStore->litStart);24032404return nbSeqs < 4 && nbLits < 10;2405}24062407static void ZSTD_confirmRepcodesAndEntropyTables(ZSTD_CCtx* zc)2408{2409ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;2410zc->blockState.prevCBlock = zc->blockState.nextCBlock;2411zc->blockState.nextCBlock = tmp;2412}24132414static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,2415void* dst, size_t dstCapacity,2416const void* src, size_t srcSize, U32 frame)2417{2418/* This the upper bound for the length of an rle block.2419* This isn't the actual upper bound. Finding the real threshold2420* needs further investigation.2421*/2422const U32 rleMaxLength = 25;2423size_t cSize;2424const BYTE* ip = (const BYTE*)src;2425BYTE* op = (BYTE*)dst;2426DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",2427(unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,2428(unsigned)zc->blockState.matchState.nextToUpdate);24292430{ const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);2431FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");2432if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }2433}24342435if (zc->seqCollector.collectSequences) {2436ZSTD_copyBlockSequences(zc);2437return 0;2438}24392440/* encode sequences and literals */2441cSize = ZSTD_compressSequences(&zc->seqStore,2442&zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,2443&zc->appliedParams,2444dst, dstCapacity,2445srcSize,2446zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */,2447zc->bmi2);24482449if (frame &&2450/* We don't want to emit our first block as a RLE even if it qualifies because2451* doing so will cause the decoder (cli only) to throw a "should consume all input error."2452* This is only an issue for zstd <= v1.4.32453*/2454!zc->isFirstBlock &&2455cSize < rleMaxLength &&2456ZSTD_isRLE(ip, srcSize))2457{2458cSize = 1;2459op[0] = ip[0];2460}24612462out:2463if (!ZSTD_isError(cSize) && cSize > 1) {2464ZSTD_confirmRepcodesAndEntropyTables(zc);2465}2466/* We check that dictionaries have offset codes available for the first2467* block. After the first block, the offcode table might not have large2468* enough codes to represent the offsets in the data.2469*/2470if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)2471zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;24722473return cSize;2474}24752476static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc,2477void* dst, size_t dstCapacity,2478const void* src, size_t srcSize,2479const size_t bss, U32 lastBlock)2480{2481DEBUGLOG(6, "Attempting ZSTD_compressSuperBlock()");2482if (bss == ZSTDbss_compress) {2483if (/* We don't want to emit our first block as a RLE even if it qualifies because2484* doing so will cause the decoder (cli only) to throw a "should consume all input error."2485* This is only an issue for zstd <= v1.4.32486*/2487!zc->isFirstBlock &&2488ZSTD_maybeRLE(&zc->seqStore) &&2489ZSTD_isRLE((BYTE const*)src, srcSize))2490{2491return ZSTD_rleCompressBlock(dst, dstCapacity, *(BYTE const*)src, srcSize, lastBlock);2492}2493/* Attempt superblock compression.2494*2495* Note that compressed size of ZSTD_compressSuperBlock() is not bound by the2496* standard ZSTD_compressBound(). This is a problem, because even if we have2497* space now, taking an extra byte now could cause us to run out of space later2498* and violate ZSTD_compressBound().2499*2500* Define blockBound(blockSize) = blockSize + ZSTD_blockHeaderSize.2501*2502* In order to respect ZSTD_compressBound() we must attempt to emit a raw2503* uncompressed block in these cases:2504* * cSize == 0: Return code for an uncompressed block.2505* * cSize == dstSize_tooSmall: We may have expanded beyond blockBound(srcSize).2506* ZSTD_noCompressBlock() will return dstSize_tooSmall if we are really out of2507* output space.2508* * cSize >= blockBound(srcSize): We have expanded the block too much so2509* emit an uncompressed block.2510*/2511{2512size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock);2513if (cSize != ERROR(dstSize_tooSmall)) {2514size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy);2515FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed");2516if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) {2517ZSTD_confirmRepcodesAndEntropyTables(zc);2518return cSize;2519}2520}2521}2522}25232524DEBUGLOG(6, "Resorting to ZSTD_noCompressBlock()");2525/* Superblock compression failed, attempt to emit a single no compress block.2526* The decoder will be able to stream this block since it is uncompressed.2527*/2528return ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock);2529}25302531static size_t ZSTD_compressBlock_targetCBlockSize(ZSTD_CCtx* zc,2532void* dst, size_t dstCapacity,2533const void* src, size_t srcSize,2534U32 lastBlock)2535{2536size_t cSize = 0;2537const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize);2538DEBUGLOG(5, "ZSTD_compressBlock_targetCBlockSize (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u, srcSize=%zu)",2539(unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, (unsigned)zc->blockState.matchState.nextToUpdate, srcSize);2540FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed");25412542cSize = ZSTD_compressBlock_targetCBlockSize_body(zc, dst, dstCapacity, src, srcSize, bss, lastBlock);2543FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize_body failed");25442545if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)2546zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;25472548return cSize;2549}25502551static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms,2552ZSTD_cwksp* ws,2553ZSTD_CCtx_params const* params,2554void const* ip,2555void const* iend)2556{2557if (ZSTD_window_needOverflowCorrection(ms->window, iend)) {2558U32 const maxDist = (U32)1 << params->cParams.windowLog;2559U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy);2560U32 const correction = ZSTD_window_correctOverflow(&ms->window, cycleLog, maxDist, ip);2561ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);2562ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);2563ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);2564ZSTD_cwksp_mark_tables_dirty(ws);2565ZSTD_reduceIndex(ms, params, correction);2566ZSTD_cwksp_mark_tables_clean(ws);2567if (ms->nextToUpdate < correction) ms->nextToUpdate = 0;2568else ms->nextToUpdate -= correction;2569/* invalidate dictionaries on overflow correction */2570ms->loadedDictEnd = 0;2571ms->dictMatchState = NULL;2572}2573}25742575/*! ZSTD_compress_frameChunk() :2576* Compress a chunk of data into one or multiple blocks.2577* All blocks will be terminated, all input will be consumed.2578* Function will issue an error if there is not enough `dstCapacity` to hold the compressed content.2579* Frame is supposed already started (header already produced)2580* @return : compressed size, or an error code2581*/2582static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,2583void* dst, size_t dstCapacity,2584const void* src, size_t srcSize,2585U32 lastFrameChunk)2586{2587size_t blockSize = cctx->blockSize;2588size_t remaining = srcSize;2589const BYTE* ip = (const BYTE*)src;2590BYTE* const ostart = (BYTE*)dst;2591BYTE* op = ostart;2592U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog;25932594assert(cctx->appliedParams.cParams.windowLog <= ZSTD_WINDOWLOG_MAX);25952596DEBUGLOG(5, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize);2597if (cctx->appliedParams.fParams.checksumFlag && srcSize)2598XXH64_update(&cctx->xxhState, src, srcSize);25992600while (remaining) {2601ZSTD_matchState_t* const ms = &cctx->blockState.matchState;2602U32 const lastBlock = lastFrameChunk & (blockSize >= remaining);26032604RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE,2605dstSize_tooSmall,2606"not enough space to store compressed block");2607if (remaining < blockSize) blockSize = remaining;26082609ZSTD_overflowCorrectIfNeeded(2610ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize);2611ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);26122613/* Ensure hash/chain table insertion resumes no sooner than lowlimit */2614if (ms->nextToUpdate < ms->window.lowLimit) ms->nextToUpdate = ms->window.lowLimit;26152616{ size_t cSize;2617if (ZSTD_useTargetCBlockSize(&cctx->appliedParams)) {2618cSize = ZSTD_compressBlock_targetCBlockSize(cctx, op, dstCapacity, ip, blockSize, lastBlock);2619FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed");2620assert(cSize > 0);2621assert(cSize <= blockSize + ZSTD_blockHeaderSize);2622} else {2623cSize = ZSTD_compressBlock_internal(cctx,2624op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,2625ip, blockSize, 1 /* frame */);2626FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_internal failed");26272628if (cSize == 0) { /* block is not compressible */2629cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);2630FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");2631} else {2632U32 const cBlockHeader = cSize == 1 ?2633lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :2634lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);2635MEM_writeLE24(op, cBlockHeader);2636cSize += ZSTD_blockHeaderSize;2637}2638}263926402641ip += blockSize;2642assert(remaining >= blockSize);2643remaining -= blockSize;2644op += cSize;2645assert(dstCapacity >= cSize);2646dstCapacity -= cSize;2647cctx->isFirstBlock = 0;2648DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u",2649(unsigned)cSize);2650} }26512652if (lastFrameChunk && (op>ostart)) cctx->stage = ZSTDcs_ending;2653return (size_t)(op-ostart);2654}265526562657static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,2658const ZSTD_CCtx_params* params, U64 pledgedSrcSize, U32 dictID)2659{ BYTE* const op = (BYTE*)dst;2660U32 const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536); /* 0-3 */2661U32 const dictIDSizeCode = params->fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength; /* 0-3 */2662U32 const checksumFlag = params->fParams.checksumFlag>0;2663U32 const windowSize = (U32)1 << params->cParams.windowLog;2664U32 const singleSegment = params->fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);2665BYTE const windowLogByte = (BYTE)((params->cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);2666U32 const fcsCode = params->fParams.contentSizeFlag ?2667(pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0; /* 0-3 */2668BYTE const frameHeaderDescriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );2669size_t pos=0;26702671assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));2672RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall,2673"dst buf is too small to fit worst-case frame header size.");2674DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",2675!params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);26762677if (params->format == ZSTD_f_zstd1) {2678MEM_writeLE32(dst, ZSTD_MAGICNUMBER);2679pos = 4;2680}2681op[pos++] = frameHeaderDescriptionByte;2682if (!singleSegment) op[pos++] = windowLogByte;2683switch(dictIDSizeCode)2684{2685default: assert(0); /* impossible */2686case 0 : break;2687case 1 : op[pos] = (BYTE)(dictID); pos++; break;2688case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break;2689case 3 : MEM_writeLE32(op+pos, dictID); pos+=4; break;2690}2691switch(fcsCode)2692{2693default: assert(0); /* impossible */2694case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break;2695case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break;2696case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break;2697case 3 : MEM_writeLE64(op+pos, (U64)(pledgedSrcSize)); pos+=8; break;2698}2699return pos;2700}27012702/* ZSTD_writeLastEmptyBlock() :2703* output an empty Block with end-of-frame mark to complete a frame2704* @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))2705* or an error code if `dstCapacity` is too small (<ZSTD_blockHeaderSize)2706*/2707size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity)2708{2709RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall,2710"dst buf is too small to write frame trailer empty block.");2711{ U32 const cBlockHeader24 = 1 /*lastBlock*/ + (((U32)bt_raw)<<1); /* 0 size */2712MEM_writeLE24(dst, cBlockHeader24);2713return ZSTD_blockHeaderSize;2714}2715}27162717size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq)2718{2719RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong,2720"wrong cctx stage");2721RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm,2722parameter_unsupported,2723"incompatible with ldm");2724cctx->externSeqStore.seq = seq;2725cctx->externSeqStore.size = nbSeq;2726cctx->externSeqStore.capacity = nbSeq;2727cctx->externSeqStore.pos = 0;2728return 0;2729}273027312732static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,2733void* dst, size_t dstCapacity,2734const void* src, size_t srcSize,2735U32 frame, U32 lastFrameChunk)2736{2737ZSTD_matchState_t* const ms = &cctx->blockState.matchState;2738size_t fhSize = 0;27392740DEBUGLOG(5, "ZSTD_compressContinue_internal, stage: %u, srcSize: %u",2741cctx->stage, (unsigned)srcSize);2742RETURN_ERROR_IF(cctx->stage==ZSTDcs_created, stage_wrong,2743"missing init (ZSTD_compressBegin)");27442745if (frame && (cctx->stage==ZSTDcs_init)) {2746fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,2747cctx->pledgedSrcSizePlusOne-1, cctx->dictID);2748FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");2749assert(fhSize <= dstCapacity);2750dstCapacity -= fhSize;2751dst = (char*)dst + fhSize;2752cctx->stage = ZSTDcs_ongoing;2753}27542755if (!srcSize) return fhSize; /* do not generate an empty block if no input */27562757if (!ZSTD_window_update(&ms->window, src, srcSize)) {2758ms->nextToUpdate = ms->window.dictLimit;2759}2760if (cctx->appliedParams.ldmParams.enableLdm) {2761ZSTD_window_update(&cctx->ldmState.window, src, srcSize);2762}27632764if (!frame) {2765/* overflow check and correction for block mode */2766ZSTD_overflowCorrectIfNeeded(2767ms, &cctx->workspace, &cctx->appliedParams,2768src, (BYTE const*)src + srcSize);2769}27702771DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (unsigned)cctx->blockSize);2772{ size_t const cSize = frame ?2773ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :2774ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */);2775FORWARD_IF_ERROR(cSize, "%s", frame ? "ZSTD_compress_frameChunk failed" : "ZSTD_compressBlock_internal failed");2776cctx->consumedSrcSize += srcSize;2777cctx->producedCSize += (cSize + fhSize);2778assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));2779if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */2780ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);2781RETURN_ERROR_IF(2782cctx->consumedSrcSize+1 > cctx->pledgedSrcSizePlusOne,2783srcSize_wrong,2784"error : pledgedSrcSize = %u, while realSrcSize >= %u",2785(unsigned)cctx->pledgedSrcSizePlusOne-1,2786(unsigned)cctx->consumedSrcSize);2787}2788return cSize + fhSize;2789}2790}27912792size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,2793void* dst, size_t dstCapacity,2794const void* src, size_t srcSize)2795{2796DEBUGLOG(5, "ZSTD_compressContinue (srcSize=%u)", (unsigned)srcSize);2797return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 1 /* frame mode */, 0 /* last chunk */);2798}279928002801size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)2802{2803ZSTD_compressionParameters const cParams = cctx->appliedParams.cParams;2804assert(!ZSTD_checkCParams(cParams));2805return MIN (ZSTD_BLOCKSIZE_MAX, (U32)1 << cParams.windowLog);2806}28072808size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)2809{2810DEBUGLOG(5, "ZSTD_compressBlock: srcSize = %u", (unsigned)srcSize);2811{ size_t const blockSizeMax = ZSTD_getBlockSize(cctx);2812RETURN_ERROR_IF(srcSize > blockSizeMax, srcSize_wrong, "input is larger than a block"); }28132814return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);2815}28162817/*! ZSTD_loadDictionaryContent() :2818* @return : 0, or an error code2819*/2820static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,2821ldmState_t* ls,2822ZSTD_cwksp* ws,2823ZSTD_CCtx_params const* params,2824const void* src, size_t srcSize,2825ZSTD_dictTableLoadMethod_e dtlm)2826{2827const BYTE* ip = (const BYTE*) src;2828const BYTE* const iend = ip + srcSize;28292830ZSTD_window_update(&ms->window, src, srcSize);2831ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base);28322833if (params->ldmParams.enableLdm && ls != NULL) {2834ZSTD_window_update(&ls->window, src, srcSize);2835ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base);2836}28372838/* Assert that we the ms params match the params we're being given */2839ZSTD_assertEqualCParams(params->cParams, ms->cParams);28402841if (srcSize <= HASH_READ_SIZE) return 0;28422843while (iend - ip > HASH_READ_SIZE) {2844size_t const remaining = (size_t)(iend - ip);2845size_t const chunk = MIN(remaining, ZSTD_CHUNKSIZE_MAX);2846const BYTE* const ichunk = ip + chunk;28472848ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, ichunk);28492850if (params->ldmParams.enableLdm && ls != NULL)2851ZSTD_ldm_fillHashTable(ls, (const BYTE*)src, (const BYTE*)src + srcSize, ¶ms->ldmParams);28522853switch(params->cParams.strategy)2854{2855case ZSTD_fast:2856ZSTD_fillHashTable(ms, ichunk, dtlm);2857break;2858case ZSTD_dfast:2859ZSTD_fillDoubleHashTable(ms, ichunk, dtlm);2860break;28612862case ZSTD_greedy:2863case ZSTD_lazy:2864case ZSTD_lazy2:2865if (chunk >= HASH_READ_SIZE)2866ZSTD_insertAndFindFirstIndex(ms, ichunk-HASH_READ_SIZE);2867break;28682869case ZSTD_btlazy2: /* we want the dictionary table fully sorted */2870case ZSTD_btopt:2871case ZSTD_btultra:2872case ZSTD_btultra2:2873if (chunk >= HASH_READ_SIZE)2874ZSTD_updateTree(ms, ichunk-HASH_READ_SIZE, ichunk);2875break;28762877default:2878assert(0); /* not possible : not a valid strategy id */2879}28802881ip = ichunk;2882}28832884ms->nextToUpdate = (U32)(iend - ms->window.base);2885return 0;2886}288728882889/* Dictionaries that assign zero probability to symbols that show up causes problems2890when FSE encoding. Refuse dictionaries that assign zero probability to symbols2891that we may encounter during compression.2892NOTE: This behavior is not standard and could be improved in the future. */2893static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) {2894U32 s;2895RETURN_ERROR_IF(dictMaxSymbolValue < maxSymbolValue, dictionary_corrupted, "dict fse tables don't have all symbols");2896for (s = 0; s <= maxSymbolValue; ++s) {2897RETURN_ERROR_IF(normalizedCounter[s] == 0, dictionary_corrupted, "dict fse tables don't have all symbols");2898}2899return 0;2900}29012902size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,2903short* offcodeNCount, unsigned* offcodeMaxValue,2904const void* const dict, size_t dictSize)2905{2906const BYTE* dictPtr = (const BYTE*)dict; /* skip magic num and dict ID */2907const BYTE* const dictEnd = dictPtr + dictSize;2908dictPtr += 8;2909bs->entropy.huf.repeatMode = HUF_repeat_check;29102911{ unsigned maxSymbolValue = 255;2912unsigned hasZeroWeights = 1;2913size_t const hufHeaderSize = HUF_readCTable((HUF_CElt*)bs->entropy.huf.CTable, &maxSymbolValue, dictPtr,2914dictEnd-dictPtr, &hasZeroWeights);29152916/* We only set the loaded table as valid if it contains all non-zero2917* weights. Otherwise, we set it to check */2918if (!hasZeroWeights)2919bs->entropy.huf.repeatMode = HUF_repeat_valid;29202921RETURN_ERROR_IF(HUF_isError(hufHeaderSize), dictionary_corrupted, "");2922RETURN_ERROR_IF(maxSymbolValue < 255, dictionary_corrupted, "");2923dictPtr += hufHeaderSize;2924}29252926{ unsigned offcodeLog;2927size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr);2928RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, "");2929RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, "");2930/* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */2931/* fill all offset symbols to avoid garbage at end of table */2932RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(2933bs->entropy.fse.offcodeCTable,2934offcodeNCount, MaxOff, offcodeLog,2935workspace, HUF_WORKSPACE_SIZE)),2936dictionary_corrupted, "");2937dictPtr += offcodeHeaderSize;2938}29392940{ short matchlengthNCount[MaxML+1];2941unsigned matchlengthMaxValue = MaxML, matchlengthLog;2942size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr);2943RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, "");2944RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, "");2945/* Every match length code must have non-zero probability */2946FORWARD_IF_ERROR( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML), "");2947RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(2948bs->entropy.fse.matchlengthCTable,2949matchlengthNCount, matchlengthMaxValue, matchlengthLog,2950workspace, HUF_WORKSPACE_SIZE)),2951dictionary_corrupted, "");2952dictPtr += matchlengthHeaderSize;2953}29542955{ short litlengthNCount[MaxLL+1];2956unsigned litlengthMaxValue = MaxLL, litlengthLog;2957size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr);2958RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, "");2959RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, "");2960/* Every literal length code must have non-zero probability */2961FORWARD_IF_ERROR( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL), "");2962RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp(2963bs->entropy.fse.litlengthCTable,2964litlengthNCount, litlengthMaxValue, litlengthLog,2965workspace, HUF_WORKSPACE_SIZE)),2966dictionary_corrupted, "");2967dictPtr += litlengthHeaderSize;2968}29692970RETURN_ERROR_IF(dictPtr+12 > dictEnd, dictionary_corrupted, "");2971bs->rep[0] = MEM_readLE32(dictPtr+0);2972bs->rep[1] = MEM_readLE32(dictPtr+4);2973bs->rep[2] = MEM_readLE32(dictPtr+8);2974dictPtr += 12;29752976return dictPtr - (const BYTE*)dict;2977}29782979/* Dictionary format :2980* See :2981* https://github.com/facebook/zstd/blob/master/doc/zstd_compression_format.md#dictionary-format2982*/2983/*! ZSTD_loadZstdDictionary() :2984* @return : dictID, or an error code2985* assumptions : magic number supposed already checked2986* dictSize supposed >= 82987*/2988static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,2989ZSTD_matchState_t* ms,2990ZSTD_cwksp* ws,2991ZSTD_CCtx_params const* params,2992const void* dict, size_t dictSize,2993ZSTD_dictTableLoadMethod_e dtlm,2994void* workspace)2995{2996const BYTE* dictPtr = (const BYTE*)dict;2997const BYTE* const dictEnd = dictPtr + dictSize;2998short offcodeNCount[MaxOff+1];2999unsigned offcodeMaxValue = MaxOff;3000size_t dictID;3001size_t eSize;30023003ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));3004assert(dictSize >= 8);3005assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY);30063007dictID = params->fParams.noDictIDFlag ? 0 : MEM_readLE32(dictPtr + 4 /* skip magic number */ );3008eSize = ZSTD_loadCEntropy(bs, workspace, offcodeNCount, &offcodeMaxValue, dict, dictSize);3009FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed");3010dictPtr += eSize;30113012{ size_t const dictContentSize = (size_t)(dictEnd - dictPtr);3013U32 offcodeMax = MaxOff;3014if (dictContentSize <= ((U32)-1) - 128 KB) {3015U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */3016offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */3017}3018/* All offset values <= dictContentSize + 128 KB must be representable */3019FORWARD_IF_ERROR(ZSTD_checkDictNCount(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)), "");3020/* All repCodes must be <= dictContentSize and != 0*/3021{ U32 u;3022for (u=0; u<3; u++) {3023RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, "");3024RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, "");3025} }30263027bs->entropy.fse.offcode_repeatMode = FSE_repeat_valid;3028bs->entropy.fse.matchlength_repeatMode = FSE_repeat_valid;3029bs->entropy.fse.litlength_repeatMode = FSE_repeat_valid;3030FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(3031ms, NULL, ws, params, dictPtr, dictContentSize, dtlm), "");3032return dictID;3033}3034}30353036/** ZSTD_compress_insertDictionary() :3037* @return : dictID, or an error code */3038static size_t3039ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,3040ZSTD_matchState_t* ms,3041ldmState_t* ls,3042ZSTD_cwksp* ws,3043const ZSTD_CCtx_params* params,3044const void* dict, size_t dictSize,3045ZSTD_dictContentType_e dictContentType,3046ZSTD_dictTableLoadMethod_e dtlm,3047void* workspace)3048{3049DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize);3050if ((dict==NULL) || (dictSize<8)) {3051RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");3052return 0;3053}30543055ZSTD_reset_compressedBlockState(bs);30563057/* dict restricted modes */3058if (dictContentType == ZSTD_dct_rawContent)3059return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm);30603061if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {3062if (dictContentType == ZSTD_dct_auto) {3063DEBUGLOG(4, "raw content dictionary detected");3064return ZSTD_loadDictionaryContent(3065ms, ls, ws, params, dict, dictSize, dtlm);3066}3067RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, "");3068assert(0); /* impossible */3069}30703071/* dict as full zstd dictionary */3072return ZSTD_loadZstdDictionary(3073bs, ms, ws, params, dict, dictSize, dtlm, workspace);3074}30753076#define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB)3077#define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6)30783079/*! ZSTD_compressBegin_internal() :3080* @return : 0, or an error code */3081static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,3082const void* dict, size_t dictSize,3083ZSTD_dictContentType_e dictContentType,3084ZSTD_dictTableLoadMethod_e dtlm,3085const ZSTD_CDict* cdict,3086const ZSTD_CCtx_params* params, U64 pledgedSrcSize,3087ZSTD_buffered_policy_e zbuff)3088{3089DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params->cParams.windowLog);3090/* params are supposed to be fully validated at this point */3091assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));3092assert(!((dict) && (cdict))); /* either dict or cdict, not both */3093if ( (cdict)3094&& (cdict->dictContentSize > 0)3095&& ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF3096|| pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER3097|| pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN3098|| cdict->compressionLevel == 0)3099&& (params->attachDictPref != ZSTD_dictForceLoad) ) {3100return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff);3101}31023103FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, *params, pledgedSrcSize,3104ZSTDcrp_makeClean, zbuff) , "");3105{ size_t const dictID = cdict ?3106ZSTD_compress_insertDictionary(3107cctx->blockState.prevCBlock, &cctx->blockState.matchState,3108&cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent,3109cdict->dictContentSize, dictContentType, dtlm,3110cctx->entropyWorkspace)3111: ZSTD_compress_insertDictionary(3112cctx->blockState.prevCBlock, &cctx->blockState.matchState,3113&cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize,3114dictContentType, dtlm, cctx->entropyWorkspace);3115FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");3116assert(dictID <= UINT_MAX);3117cctx->dictID = (U32)dictID;3118}3119return 0;3120}31213122size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,3123const void* dict, size_t dictSize,3124ZSTD_dictContentType_e dictContentType,3125ZSTD_dictTableLoadMethod_e dtlm,3126const ZSTD_CDict* cdict,3127const ZSTD_CCtx_params* params,3128unsigned long long pledgedSrcSize)3129{3130DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params->cParams.windowLog);3131/* compression parameters verification and optimization */3132FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) , "");3133return ZSTD_compressBegin_internal(cctx,3134dict, dictSize, dictContentType, dtlm,3135cdict,3136params, pledgedSrcSize,3137ZSTDb_not_buffered);3138}31393140/*! ZSTD_compressBegin_advanced() :3141* @return : 0, or an error code */3142size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,3143const void* dict, size_t dictSize,3144ZSTD_parameters params, unsigned long long pledgedSrcSize)3145{3146ZSTD_CCtx_params const cctxParams =3147ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms);3148return ZSTD_compressBegin_advanced_internal(cctx,3149dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast,3150NULL /*cdict*/,3151&cctxParams, pledgedSrcSize);3152}31533154size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)3155{3156ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);3157ZSTD_CCtx_params const cctxParams =3158ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms);3159DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize);3160return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,3161&cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);3162}31633164size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel)3165{3166return ZSTD_compressBegin_usingDict(cctx, NULL, 0, compressionLevel);3167}316831693170/*! ZSTD_writeEpilogue() :3171* Ends a frame.3172* @return : nb of bytes written into dst (or an error code) */3173static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)3174{3175BYTE* const ostart = (BYTE*)dst;3176BYTE* op = ostart;3177size_t fhSize = 0;31783179DEBUGLOG(4, "ZSTD_writeEpilogue");3180RETURN_ERROR_IF(cctx->stage == ZSTDcs_created, stage_wrong, "init missing");31813182/* special case : empty frame */3183if (cctx->stage == ZSTDcs_init) {3184fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0);3185FORWARD_IF_ERROR(fhSize, "ZSTD_writeFrameHeader failed");3186dstCapacity -= fhSize;3187op += fhSize;3188cctx->stage = ZSTDcs_ongoing;3189}31903191if (cctx->stage != ZSTDcs_ending) {3192/* write one last empty block, make it the "last" block */3193U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1) + 0;3194RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for epilogue");3195MEM_writeLE32(op, cBlockHeader24);3196op += ZSTD_blockHeaderSize;3197dstCapacity -= ZSTD_blockHeaderSize;3198}31993200if (cctx->appliedParams.fParams.checksumFlag) {3201U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);3202RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum");3203DEBUGLOG(4, "ZSTD_writeEpilogue: write checksum : %08X", (unsigned)checksum);3204MEM_writeLE32(op, checksum);3205op += 4;3206}32073208cctx->stage = ZSTDcs_created; /* return to "created but no init" status */3209return op-ostart;3210}32113212size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,3213void* dst, size_t dstCapacity,3214const void* src, size_t srcSize)3215{3216size_t endResult;3217size_t const cSize = ZSTD_compressContinue_internal(cctx,3218dst, dstCapacity, src, srcSize,32191 /* frame mode */, 1 /* last chunk */);3220FORWARD_IF_ERROR(cSize, "ZSTD_compressContinue_internal failed");3221endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);3222FORWARD_IF_ERROR(endResult, "ZSTD_writeEpilogue failed");3223assert(!(cctx->appliedParams.fParams.contentSizeFlag && cctx->pledgedSrcSizePlusOne == 0));3224if (cctx->pledgedSrcSizePlusOne != 0) { /* control src size */3225ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN == (unsigned long long)-1);3226DEBUGLOG(4, "end of frame : controlling src size");3227RETURN_ERROR_IF(3228cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1,3229srcSize_wrong,3230"error : pledgedSrcSize = %u, while realSrcSize = %u",3231(unsigned)cctx->pledgedSrcSizePlusOne-1,3232(unsigned)cctx->consumedSrcSize);3233}3234return cSize + endResult;3235}323632373238static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx,3239void* dst, size_t dstCapacity,3240const void* src, size_t srcSize,3241const void* dict,size_t dictSize,3242const ZSTD_parameters* params)3243{3244ZSTD_CCtx_params const cctxParams =3245ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);3246DEBUGLOG(4, "ZSTD_compress_internal");3247return ZSTD_compress_advanced_internal(cctx,3248dst, dstCapacity,3249src, srcSize,3250dict, dictSize,3251&cctxParams);3252}32533254size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,3255void* dst, size_t dstCapacity,3256const void* src, size_t srcSize,3257const void* dict,size_t dictSize,3258ZSTD_parameters params)3259{3260DEBUGLOG(4, "ZSTD_compress_advanced");3261FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), "");3262return ZSTD_compress_internal(cctx,3263dst, dstCapacity,3264src, srcSize,3265dict, dictSize,3266¶ms);3267}32683269/* Internal */3270size_t ZSTD_compress_advanced_internal(3271ZSTD_CCtx* cctx,3272void* dst, size_t dstCapacity,3273const void* src, size_t srcSize,3274const void* dict,size_t dictSize,3275const ZSTD_CCtx_params* params)3276{3277DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);3278FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,3279dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,3280params, srcSize, ZSTDb_not_buffered) , "");3281return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);3282}32833284size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx,3285void* dst, size_t dstCapacity,3286const void* src, size_t srcSize,3287const void* dict, size_t dictSize,3288int compressionLevel)3289{3290ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0);3291ZSTD_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms);3292DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize);3293assert(params.fParams.contentSizeFlag == 1);3294return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctxParams);3295}32963297size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,3298void* dst, size_t dstCapacity,3299const void* src, size_t srcSize,3300int compressionLevel)3301{3302DEBUGLOG(4, "ZSTD_compressCCtx (srcSize=%u)", (unsigned)srcSize);3303assert(cctx != NULL);3304return ZSTD_compress_usingDict(cctx, dst, dstCapacity, src, srcSize, NULL, 0, compressionLevel);3305}33063307size_t ZSTD_compress(void* dst, size_t dstCapacity,3308const void* src, size_t srcSize,3309int compressionLevel)3310{3311size_t result;3312ZSTD_CCtx ctxBody;3313ZSTD_initCCtx(&ctxBody, ZSTD_defaultCMem);3314result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel);3315ZSTD_freeCCtxContent(&ctxBody); /* can't free ctxBody itself, as it's on stack; free only heap content */3316return result;3317}331833193320/* ===== Dictionary API ===== */33213322/*! ZSTD_estimateCDictSize_advanced() :3323* Estimate amount of memory that will be needed to create a dictionary with following arguments */3324size_t ZSTD_estimateCDictSize_advanced(3325size_t dictSize, ZSTD_compressionParameters cParams,3326ZSTD_dictLoadMethod_e dictLoadMethod)3327{3328DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));3329return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))3330+ ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)3331+ ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0)3332+ (dictLoadMethod == ZSTD_dlm_byRef ? 03333: ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *))));3334}33353336size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)3337{3338ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);3339return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy);3340}33413342size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)3343{3344if (cdict==NULL) return 0; /* support sizeof on NULL */3345DEBUGLOG(5, "sizeof(*cdict) : %u", (unsigned)sizeof(*cdict));3346/* cdict may be in the workspace */3347return (cdict->workspace.workspace == cdict ? 0 : sizeof(*cdict))3348+ ZSTD_cwksp_sizeof(&cdict->workspace);3349}33503351static size_t ZSTD_initCDict_internal(3352ZSTD_CDict* cdict,3353const void* dictBuffer, size_t dictSize,3354ZSTD_dictLoadMethod_e dictLoadMethod,3355ZSTD_dictContentType_e dictContentType,3356ZSTD_compressionParameters cParams)3357{3358DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType);3359assert(!ZSTD_checkCParams(cParams));3360cdict->matchState.cParams = cParams;3361if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {3362cdict->dictContent = dictBuffer;3363} else {3364void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*)));3365RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!");3366cdict->dictContent = internalBuffer;3367memcpy(internalBuffer, dictBuffer, dictSize);3368}3369cdict->dictContentSize = dictSize;33703371cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE);337233733374/* Reset the state to no dictionary */3375ZSTD_reset_compressedBlockState(&cdict->cBlockState);3376FORWARD_IF_ERROR(ZSTD_reset_matchState(3377&cdict->matchState,3378&cdict->workspace,3379&cParams,3380ZSTDcrp_makeClean,3381ZSTDirp_reset,3382ZSTD_resetTarget_CDict), "");3383/* (Maybe) load the dictionary3384* Skips loading the dictionary if it is < 8 bytes.3385*/3386{ ZSTD_CCtx_params params;3387memset(¶ms, 0, sizeof(params));3388params.compressionLevel = ZSTD_CLEVEL_DEFAULT;3389params.fParams.contentSizeFlag = 1;3390params.cParams = cParams;3391{ size_t const dictID = ZSTD_compress_insertDictionary(3392&cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace,3393¶ms, cdict->dictContent, cdict->dictContentSize,3394dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace);3395FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed");3396assert(dictID <= (size_t)(U32)-1);3397cdict->dictID = (U32)dictID;3398}3399}34003401return 0;3402}34033404ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,3405ZSTD_dictLoadMethod_e dictLoadMethod,3406ZSTD_dictContentType_e dictContentType,3407ZSTD_compressionParameters cParams, ZSTD_customMem customMem)3408{3409DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (unsigned)dictContentType);3410if (!customMem.customAlloc ^ !customMem.customFree) return NULL;34113412{ size_t const workspaceSize =3413ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) +3414ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) +3415ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0) +3416(dictLoadMethod == ZSTD_dlm_byRef ? 03417: ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))));3418void* const workspace = ZSTD_malloc(workspaceSize, customMem);3419ZSTD_cwksp ws;3420ZSTD_CDict* cdict;34213422if (!workspace) {3423ZSTD_free(workspace, customMem);3424return NULL;3425}34263427ZSTD_cwksp_init(&ws, workspace, workspaceSize);34283429cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));3430assert(cdict != NULL);3431ZSTD_cwksp_move(&cdict->workspace, &ws);3432cdict->customMem = customMem;3433cdict->compressionLevel = 0; /* signals advanced API usage */34343435if (ZSTD_isError( ZSTD_initCDict_internal(cdict,3436dictBuffer, dictSize,3437dictLoadMethod, dictContentType,3438cParams) )) {3439ZSTD_freeCDict(cdict);3440return NULL;3441}34423443return cdict;3444}3445}34463447ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)3448{3449ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);3450ZSTD_CDict* cdict = ZSTD_createCDict_advanced(dict, dictSize,3451ZSTD_dlm_byCopy, ZSTD_dct_auto,3452cParams, ZSTD_defaultCMem);3453if (cdict)3454cdict->compressionLevel = compressionLevel == 0 ? ZSTD_CLEVEL_DEFAULT : compressionLevel;3455return cdict;3456}34573458ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel)3459{3460ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);3461return ZSTD_createCDict_advanced(dict, dictSize,3462ZSTD_dlm_byRef, ZSTD_dct_auto,3463cParams, ZSTD_defaultCMem);3464}34653466size_t ZSTD_freeCDict(ZSTD_CDict* cdict)3467{3468if (cdict==NULL) return 0; /* support free on NULL */3469{ ZSTD_customMem const cMem = cdict->customMem;3470int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict);3471ZSTD_cwksp_free(&cdict->workspace, cMem);3472if (!cdictInWorkspace) {3473ZSTD_free(cdict, cMem);3474}3475return 0;3476}3477}34783479/*! ZSTD_initStaticCDict_advanced() :3480* Generate a digested dictionary in provided memory area.3481* workspace: The memory area to emplace the dictionary into.3482* Provided pointer must 8-bytes aligned.3483* It must outlive dictionary usage.3484* workspaceSize: Use ZSTD_estimateCDictSize()3485* to determine how large workspace must be.3486* cParams : use ZSTD_getCParams() to transform a compression level3487* into its relevants cParams.3488* @return : pointer to ZSTD_CDict*, or NULL if error (size too small)3489* Note : there is no corresponding "free" function.3490* Since workspace was allocated externally, it must be freed externally.3491*/3492const ZSTD_CDict* ZSTD_initStaticCDict(3493void* workspace, size_t workspaceSize,3494const void* dict, size_t dictSize,3495ZSTD_dictLoadMethod_e dictLoadMethod,3496ZSTD_dictContentType_e dictContentType,3497ZSTD_compressionParameters cParams)3498{3499size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0);3500size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))3501+ (dictLoadMethod == ZSTD_dlm_byRef ? 03502: ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))))3503+ ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)3504+ matchStateSize;3505ZSTD_CDict* cdict;35063507if ((size_t)workspace & 7) return NULL; /* 8-aligned */35083509{3510ZSTD_cwksp ws;3511ZSTD_cwksp_init(&ws, workspace, workspaceSize);3512cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));3513if (cdict == NULL) return NULL;3514ZSTD_cwksp_move(&cdict->workspace, &ws);3515}35163517DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u",3518(unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize));3519if (workspaceSize < neededSize) return NULL;35203521if (ZSTD_isError( ZSTD_initCDict_internal(cdict,3522dict, dictSize,3523dictLoadMethod, dictContentType,3524cParams) ))3525return NULL;35263527return cdict;3528}35293530ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict)3531{3532assert(cdict != NULL);3533return cdict->matchState.cParams;3534}35353536/* ZSTD_compressBegin_usingCDict_advanced() :3537* cdict must be != NULL */3538size_t ZSTD_compressBegin_usingCDict_advanced(3539ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,3540ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)3541{3542DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_advanced");3543RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!");3544{ ZSTD_CCtx_params params = cctx->requestedParams;3545params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF3546|| pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER3547|| pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN3548|| cdict->compressionLevel == 0 )3549&& (params.attachDictPref != ZSTD_dictForceLoad) ?3550ZSTD_getCParamsFromCDict(cdict)3551: ZSTD_getCParams(cdict->compressionLevel,3552pledgedSrcSize,3553cdict->dictContentSize);3554/* Increase window log to fit the entire dictionary and source if the3555* source size is known. Limit the increase to 19, which is the3556* window log for compression level 1 with the largest source size.3557*/3558if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) {3559U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19);3560U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1;3561params.cParams.windowLog = MAX(params.cParams.windowLog, limitedSrcLog);3562}3563params.fParams = fParams;3564return ZSTD_compressBegin_internal(cctx,3565NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast,3566cdict,3567¶ms, pledgedSrcSize,3568ZSTDb_not_buffered);3569}3570}35713572/* ZSTD_compressBegin_usingCDict() :3573* pledgedSrcSize=0 means "unknown"3574* if pledgedSrcSize>0, it will enable contentSizeFlag */3575size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)3576{3577ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };3578DEBUGLOG(4, "ZSTD_compressBegin_usingCDict : dictIDFlag == %u", !fParams.noDictIDFlag);3579return ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN);3580}35813582size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,3583void* dst, size_t dstCapacity,3584const void* src, size_t srcSize,3585const ZSTD_CDict* cdict, ZSTD_frameParameters fParams)3586{3587FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */3588return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);3589}35903591/*! ZSTD_compress_usingCDict() :3592* Compression using a digested Dictionary.3593* Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.3594* Note that compression parameters are decided at CDict creation time3595* while frame parameters are hardcoded */3596size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,3597void* dst, size_t dstCapacity,3598const void* src, size_t srcSize,3599const ZSTD_CDict* cdict)3600{3601ZSTD_frameParameters const fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };3602return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, fParams);3603}3604360536063607/* ******************************************************************3608* Streaming3609********************************************************************/36103611ZSTD_CStream* ZSTD_createCStream(void)3612{3613DEBUGLOG(3, "ZSTD_createCStream");3614return ZSTD_createCStream_advanced(ZSTD_defaultCMem);3615}36163617ZSTD_CStream* ZSTD_initStaticCStream(void *workspace, size_t workspaceSize)3618{3619return ZSTD_initStaticCCtx(workspace, workspaceSize);3620}36213622ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)3623{ /* CStream and CCtx are now same object */3624return ZSTD_createCCtx_advanced(customMem);3625}36263627size_t ZSTD_freeCStream(ZSTD_CStream* zcs)3628{3629return ZSTD_freeCCtx(zcs); /* same object */3630}3631363236333634/*====== Initialization ======*/36353636size_t ZSTD_CStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX; }36373638size_t ZSTD_CStreamOutSize(void)3639{3640return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ;3641}36423643static size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx,3644const void* const dict, size_t const dictSize, ZSTD_dictContentType_e const dictContentType,3645const ZSTD_CDict* const cdict,3646ZSTD_CCtx_params params, unsigned long long const pledgedSrcSize)3647{3648DEBUGLOG(4, "ZSTD_resetCStream_internal");3649/* Finalize the compression parameters */3650params.cParams = ZSTD_getCParamsFromCCtxParams(¶ms, pledgedSrcSize, dictSize);3651/* params are supposed to be fully validated at this point */3652assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));3653assert(!((dict) && (cdict))); /* either dict or cdict, not both */36543655FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,3656dict, dictSize, dictContentType, ZSTD_dtlm_fast,3657cdict,3658¶ms, pledgedSrcSize,3659ZSTDb_buffered) , "");36603661cctx->inToCompress = 0;3662cctx->inBuffPos = 0;3663cctx->inBuffTarget = cctx->blockSize3664+ (cctx->blockSize == pledgedSrcSize); /* for small input: avoid automatic flush on reaching end of block, since it would require to add a 3-bytes null block to end frame */3665cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0;3666cctx->streamStage = zcss_load;3667cctx->frameEnded = 0;3668return 0; /* ready to go */3669}36703671/* ZSTD_resetCStream():3672* pledgedSrcSize == 0 means "unknown" */3673size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss)3674{3675/* temporary : 0 interpreted as "unknown" during transition period.3676* Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.3677* 0 will be interpreted as "empty" in the future.3678*/3679U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;3680DEBUGLOG(4, "ZSTD_resetCStream: pledgedSrcSize = %u", (unsigned)pledgedSrcSize);3681FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3682FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");3683return 0;3684}36853686/*! ZSTD_initCStream_internal() :3687* Note : for lib/compress only. Used by zstdmt_compress.c.3688* Assumption 1 : params are valid3689* Assumption 2 : either dict, or cdict, is defined, not both */3690size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,3691const void* dict, size_t dictSize, const ZSTD_CDict* cdict,3692const ZSTD_CCtx_params* params,3693unsigned long long pledgedSrcSize)3694{3695DEBUGLOG(4, "ZSTD_initCStream_internal");3696FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3697FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");3698assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));3699zcs->requestedParams = *params;3700assert(!((dict) && (cdict))); /* either dict or cdict, not both */3701if (dict) {3702FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");3703} else {3704/* Dictionary is cleared if !cdict */3705FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");3706}3707return 0;3708}37093710/* ZSTD_initCStream_usingCDict_advanced() :3711* same as ZSTD_initCStream_usingCDict(), with control over frame parameters */3712size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,3713const ZSTD_CDict* cdict,3714ZSTD_frameParameters fParams,3715unsigned long long pledgedSrcSize)3716{3717DEBUGLOG(4, "ZSTD_initCStream_usingCDict_advanced");3718FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3719FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");3720zcs->requestedParams.fParams = fParams;3721FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");3722return 0;3723}37243725/* note : cdict must outlive compression session */3726size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)3727{3728DEBUGLOG(4, "ZSTD_initCStream_usingCDict");3729FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3730FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, cdict) , "");3731return 0;3732}373337343735/* ZSTD_initCStream_advanced() :3736* pledgedSrcSize must be exact.3737* if srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN.3738* dict is loaded with default parameters ZSTD_dct_auto and ZSTD_dlm_byCopy. */3739size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,3740const void* dict, size_t dictSize,3741ZSTD_parameters params, unsigned long long pss)3742{3743/* for compatibility with older programs relying on this behavior.3744* Users should now specify ZSTD_CONTENTSIZE_UNKNOWN.3745* This line will be removed in the future.3746*/3747U64 const pledgedSrcSize = (pss==0 && params.fParams.contentSizeFlag==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;3748DEBUGLOG(4, "ZSTD_initCStream_advanced");3749FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3750FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");3751FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , "");3752zcs->requestedParams = ZSTD_assignParamsToCCtxParams(&zcs->requestedParams, ¶ms);3753FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");3754return 0;3755}37563757size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)3758{3759DEBUGLOG(4, "ZSTD_initCStream_usingDict");3760FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3761FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");3762FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , "");3763return 0;3764}37653766size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pss)3767{3768/* temporary : 0 interpreted as "unknown" during transition period.3769* Users willing to specify "unknown" **must** use ZSTD_CONTENTSIZE_UNKNOWN.3770* 0 will be interpreted as "empty" in the future.3771*/3772U64 const pledgedSrcSize = (pss==0) ? ZSTD_CONTENTSIZE_UNKNOWN : pss;3773DEBUGLOG(4, "ZSTD_initCStream_srcSize");3774FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3775FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");3776FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");3777FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , "");3778return 0;3779}37803781size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)3782{3783DEBUGLOG(4, "ZSTD_initCStream");3784FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , "");3785FORWARD_IF_ERROR( ZSTD_CCtx_refCDict(zcs, NULL) , "");3786FORWARD_IF_ERROR( ZSTD_CCtx_setParameter(zcs, ZSTD_c_compressionLevel, compressionLevel) , "");3787return 0;3788}37893790/*====== Compression ======*/37913792static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx)3793{3794size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos;3795if (hintInSize==0) hintInSize = cctx->blockSize;3796return hintInSize;3797}37983799/** ZSTD_compressStream_generic():3800* internal function for all *compressStream*() variants3801* non-static, because can be called from zstdmt_compress.c3802* @return : hint size for next input */3803static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,3804ZSTD_outBuffer* output,3805ZSTD_inBuffer* input,3806ZSTD_EndDirective const flushMode)3807{3808const char* const istart = (const char*)input->src;3809const char* const iend = input->size != 0 ? istart + input->size : istart;3810const char* ip = input->pos != 0 ? istart + input->pos : istart;3811char* const ostart = (char*)output->dst;3812char* const oend = output->size != 0 ? ostart + output->size : ostart;3813char* op = output->pos != 0 ? ostart + output->pos : ostart;3814U32 someMoreWork = 1;38153816/* check expectations */3817DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (unsigned)flushMode);3818assert(zcs->inBuff != NULL);3819assert(zcs->inBuffSize > 0);3820assert(zcs->outBuff != NULL);3821assert(zcs->outBuffSize > 0);3822assert(output->pos <= output->size);3823assert(input->pos <= input->size);38243825while (someMoreWork) {3826switch(zcs->streamStage)3827{3828case zcss_init:3829RETURN_ERROR(init_missing, "call ZSTD_initCStream() first!");38303831case zcss_load:3832if ( (flushMode == ZSTD_e_end)3833&& ((size_t)(oend-op) >= ZSTD_compressBound(iend-ip)) /* enough dstCapacity */3834&& (zcs->inBuffPos == 0) ) {3835/* shortcut to compression pass directly into output buffer */3836size_t const cSize = ZSTD_compressEnd(zcs,3837op, oend-op, ip, iend-ip);3838DEBUGLOG(4, "ZSTD_compressEnd : cSize=%u", (unsigned)cSize);3839FORWARD_IF_ERROR(cSize, "ZSTD_compressEnd failed");3840ip = iend;3841op += cSize;3842zcs->frameEnded = 1;3843ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);3844someMoreWork = 0; break;3845}3846/* complete loading into inBuffer */3847{ size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos;3848size_t const loaded = ZSTD_limitCopy(3849zcs->inBuff + zcs->inBuffPos, toLoad,3850ip, iend-ip);3851zcs->inBuffPos += loaded;3852if (loaded != 0)3853ip += loaded;3854if ( (flushMode == ZSTD_e_continue)3855&& (zcs->inBuffPos < zcs->inBuffTarget) ) {3856/* not enough input to fill full block : stop here */3857someMoreWork = 0; break;3858}3859if ( (flushMode == ZSTD_e_flush)3860&& (zcs->inBuffPos == zcs->inToCompress) ) {3861/* empty */3862someMoreWork = 0; break;3863}3864}3865/* compress current block (note : this stage cannot be stopped in the middle) */3866DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode);3867{ void* cDst;3868size_t cSize;3869size_t const iSize = zcs->inBuffPos - zcs->inToCompress;3870size_t oSize = oend-op;3871unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend);3872if (oSize >= ZSTD_compressBound(iSize))3873cDst = op; /* compress into output buffer, to skip flush stage */3874else3875cDst = zcs->outBuff, oSize = zcs->outBuffSize;3876cSize = lastBlock ?3877ZSTD_compressEnd(zcs, cDst, oSize,3878zcs->inBuff + zcs->inToCompress, iSize) :3879ZSTD_compressContinue(zcs, cDst, oSize,3880zcs->inBuff + zcs->inToCompress, iSize);3881FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed");3882zcs->frameEnded = lastBlock;3883/* prepare next block */3884zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;3885if (zcs->inBuffTarget > zcs->inBuffSize)3886zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;3887DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u",3888(unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize);3889if (!lastBlock)3890assert(zcs->inBuffTarget <= zcs->inBuffSize);3891zcs->inToCompress = zcs->inBuffPos;3892if (cDst == op) { /* no need to flush */3893op += cSize;3894if (zcs->frameEnded) {3895DEBUGLOG(5, "Frame completed directly in outBuffer");3896someMoreWork = 0;3897ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);3898}3899break;3900}3901zcs->outBuffContentSize = cSize;3902zcs->outBuffFlushedSize = 0;3903zcs->streamStage = zcss_flush; /* pass-through to flush stage */3904}3905/* fall-through */3906case zcss_flush:3907DEBUGLOG(5, "flush stage");3908{ size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;3909size_t const flushed = ZSTD_limitCopy(op, (size_t)(oend-op),3910zcs->outBuff + zcs->outBuffFlushedSize, toFlush);3911DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u",3912(unsigned)toFlush, (unsigned)(oend-op), (unsigned)flushed);3913if (flushed)3914op += flushed;3915zcs->outBuffFlushedSize += flushed;3916if (toFlush!=flushed) {3917/* flush not fully completed, presumably because dst is too small */3918assert(op==oend);3919someMoreWork = 0;3920break;3921}3922zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;3923if (zcs->frameEnded) {3924DEBUGLOG(5, "Frame completed on flush");3925someMoreWork = 0;3926ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);3927break;3928}3929zcs->streamStage = zcss_load;3930break;3931}39323933default: /* impossible */3934assert(0);3935}3936}39373938input->pos = ip - istart;3939output->pos = op - ostart;3940if (zcs->frameEnded) return 0;3941return ZSTD_nextInputSizeHint(zcs);3942}39433944static size_t ZSTD_nextInputSizeHint_MTorST(const ZSTD_CCtx* cctx)3945{3946#ifdef ZSTD_MULTITHREAD3947if (cctx->appliedParams.nbWorkers >= 1) {3948assert(cctx->mtctx != NULL);3949return ZSTDMT_nextInputSizeHint(cctx->mtctx);3950}3951#endif3952return ZSTD_nextInputSizeHint(cctx);39533954}39553956size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)3957{3958FORWARD_IF_ERROR( ZSTD_compressStream2(zcs, output, input, ZSTD_e_continue) , "");3959return ZSTD_nextInputSizeHint_MTorST(zcs);3960}396139623963size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,3964ZSTD_outBuffer* output,3965ZSTD_inBuffer* input,3966ZSTD_EndDirective endOp)3967{3968DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp);3969/* check conditions */3970RETURN_ERROR_IF(output->pos > output->size, GENERIC, "invalid buffer");3971RETURN_ERROR_IF(input->pos > input->size, GENERIC, "invalid buffer");3972assert(cctx!=NULL);39733974/* transparent initialization stage */3975if (cctx->streamStage == zcss_init) {3976ZSTD_CCtx_params params = cctx->requestedParams;3977ZSTD_prefixDict const prefixDict = cctx->prefixDict;3978FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */3979memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */3980assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */3981DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage");3982if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = input->size + 1; /* auto-fix pledgedSrcSize */3983params.cParams = ZSTD_getCParamsFromCCtxParams(3984&cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1, 0 /*dictSize*/);398539863987#ifdef ZSTD_MULTITHREAD3988if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) {3989params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */3990}3991if (params.nbWorkers > 0) {3992/* mt context creation */3993if (cctx->mtctx == NULL) {3994DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",3995params.nbWorkers);3996cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem);3997RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!");3998}3999/* mt compression */4000DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers);4001FORWARD_IF_ERROR( ZSTDMT_initCStream_internal(4002cctx->mtctx,4003prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,4004cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , "");4005cctx->streamStage = zcss_load;4006cctx->appliedParams.nbWorkers = params.nbWorkers;4007} else4008#endif4009{ FORWARD_IF_ERROR( ZSTD_resetCStream_internal(cctx,4010prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType,4011cctx->cdict,4012params, cctx->pledgedSrcSizePlusOne-1) , "");4013assert(cctx->streamStage == zcss_load);4014assert(cctx->appliedParams.nbWorkers == 0);4015} }4016/* end of transparent initialization stage */40174018/* compression stage */4019#ifdef ZSTD_MULTITHREAD4020if (cctx->appliedParams.nbWorkers > 0) {4021int const forceMaxProgress = (endOp == ZSTD_e_flush || endOp == ZSTD_e_end);4022size_t flushMin;4023assert(forceMaxProgress || endOp == ZSTD_e_continue /* Protection for a new flush type */);4024if (cctx->cParamsChanged) {4025ZSTDMT_updateCParams_whileCompressing(cctx->mtctx, &cctx->requestedParams);4026cctx->cParamsChanged = 0;4027}4028do {4029flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp);4030if ( ZSTD_isError(flushMin)4031|| (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */4032ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);4033}4034FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed");4035} while (forceMaxProgress && flushMin != 0 && output->pos < output->size);4036DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic");4037/* Either we don't require maximum forward progress, we've finished the4038* flush, or we are out of output space.4039*/4040assert(!forceMaxProgress || flushMin == 0 || output->pos == output->size);4041return flushMin;4042}4043#endif4044FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , "");4045DEBUGLOG(5, "completed ZSTD_compressStream2");4046return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */4047}40484049size_t ZSTD_compressStream2_simpleArgs (4050ZSTD_CCtx* cctx,4051void* dst, size_t dstCapacity, size_t* dstPos,4052const void* src, size_t srcSize, size_t* srcPos,4053ZSTD_EndDirective endOp)4054{4055ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };4056ZSTD_inBuffer input = { src, srcSize, *srcPos };4057/* ZSTD_compressStream2() will check validity of dstPos and srcPos */4058size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp);4059*dstPos = output.pos;4060*srcPos = input.pos;4061return cErr;4062}40634064size_t ZSTD_compress2(ZSTD_CCtx* cctx,4065void* dst, size_t dstCapacity,4066const void* src, size_t srcSize)4067{4068DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize);4069ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only);4070{ size_t oPos = 0;4071size_t iPos = 0;4072size_t const result = ZSTD_compressStream2_simpleArgs(cctx,4073dst, dstCapacity, &oPos,4074src, srcSize, &iPos,4075ZSTD_e_end);4076FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed");4077if (result != 0) { /* compression not completed, due to lack of output space */4078assert(oPos == dstCapacity);4079RETURN_ERROR(dstSize_tooSmall, "");4080}4081assert(iPos == srcSize); /* all input is expected consumed */4082return oPos;4083}4084}40854086/*====== Finalize ======*/40874088/*! ZSTD_flushStream() :4089* @return : amount of data remaining to flush */4090size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)4091{4092ZSTD_inBuffer input = { NULL, 0, 0 };4093return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush);4094}409540964097size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)4098{4099ZSTD_inBuffer input = { NULL, 0, 0 };4100size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end);4101FORWARD_IF_ERROR( remainingToFlush , "ZSTD_compressStream2 failed");4102if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */4103/* single thread mode : attempt to calculate remaining to flush more precisely */4104{ size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;4105size_t const checksumSize = (size_t)(zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4);4106size_t const toFlush = remainingToFlush + lastBlockSize + checksumSize;4107DEBUGLOG(4, "ZSTD_endStream : remaining to flush : %u", (unsigned)toFlush);4108return toFlush;4109}4110}411141124113/*-===== Pre-defined compression levels =====-*/41144115#define ZSTD_MAX_CLEVEL 224116int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }4117int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; }41184119static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEVEL+1] = {4120{ /* "default" - for any srcSize > 256 KB */4121/* W, C, H, S, L, TL, strat */4122{ 19, 12, 13, 1, 6, 1, ZSTD_fast }, /* base for negative levels */4123{ 19, 13, 14, 1, 7, 0, ZSTD_fast }, /* level 1 */4124{ 20, 15, 16, 1, 6, 0, ZSTD_fast }, /* level 2 */4125{ 21, 16, 17, 1, 5, 0, ZSTD_dfast }, /* level 3 */4126{ 21, 18, 18, 1, 5, 0, ZSTD_dfast }, /* level 4 */4127{ 21, 18, 19, 2, 5, 2, ZSTD_greedy }, /* level 5 */4128{ 21, 19, 19, 3, 5, 4, ZSTD_greedy }, /* level 6 */4129{ 21, 19, 19, 3, 5, 8, ZSTD_lazy }, /* level 7 */4130{ 21, 19, 19, 3, 5, 16, ZSTD_lazy2 }, /* level 8 */4131{ 21, 19, 20, 4, 5, 16, ZSTD_lazy2 }, /* level 9 */4132{ 22, 20, 21, 4, 5, 16, ZSTD_lazy2 }, /* level 10 */4133{ 22, 21, 22, 4, 5, 16, ZSTD_lazy2 }, /* level 11 */4134{ 22, 21, 22, 5, 5, 16, ZSTD_lazy2 }, /* level 12 */4135{ 22, 21, 22, 5, 5, 32, ZSTD_btlazy2 }, /* level 13 */4136{ 22, 22, 23, 5, 5, 32, ZSTD_btlazy2 }, /* level 14 */4137{ 22, 23, 23, 6, 5, 32, ZSTD_btlazy2 }, /* level 15 */4138{ 22, 22, 22, 5, 5, 48, ZSTD_btopt }, /* level 16 */4139{ 23, 23, 22, 5, 4, 64, ZSTD_btopt }, /* level 17 */4140{ 23, 23, 22, 6, 3, 64, ZSTD_btultra }, /* level 18 */4141{ 23, 24, 22, 7, 3,256, ZSTD_btultra2}, /* level 19 */4142{ 25, 25, 23, 7, 3,256, ZSTD_btultra2}, /* level 20 */4143{ 26, 26, 24, 7, 3,512, ZSTD_btultra2}, /* level 21 */4144{ 27, 27, 25, 9, 3,999, ZSTD_btultra2}, /* level 22 */4145},4146{ /* for srcSize <= 256 KB */4147/* W, C, H, S, L, T, strat */4148{ 18, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */4149{ 18, 13, 14, 1, 6, 0, ZSTD_fast }, /* level 1 */4150{ 18, 14, 14, 1, 5, 0, ZSTD_dfast }, /* level 2 */4151{ 18, 16, 16, 1, 4, 0, ZSTD_dfast }, /* level 3 */4152{ 18, 16, 17, 2, 5, 2, ZSTD_greedy }, /* level 4.*/4153{ 18, 18, 18, 3, 5, 2, ZSTD_greedy }, /* level 5.*/4154{ 18, 18, 19, 3, 5, 4, ZSTD_lazy }, /* level 6.*/4155{ 18, 18, 19, 4, 4, 4, ZSTD_lazy }, /* level 7 */4156{ 18, 18, 19, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */4157{ 18, 18, 19, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */4158{ 18, 18, 19, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */4159{ 18, 18, 19, 5, 4, 12, ZSTD_btlazy2 }, /* level 11.*/4160{ 18, 19, 19, 7, 4, 12, ZSTD_btlazy2 }, /* level 12.*/4161{ 18, 18, 19, 4, 4, 16, ZSTD_btopt }, /* level 13 */4162{ 18, 18, 19, 4, 3, 32, ZSTD_btopt }, /* level 14.*/4163{ 18, 18, 19, 6, 3,128, ZSTD_btopt }, /* level 15.*/4164{ 18, 19, 19, 6, 3,128, ZSTD_btultra }, /* level 16.*/4165{ 18, 19, 19, 8, 3,256, ZSTD_btultra }, /* level 17.*/4166{ 18, 19, 19, 6, 3,128, ZSTD_btultra2}, /* level 18.*/4167{ 18, 19, 19, 8, 3,256, ZSTD_btultra2}, /* level 19.*/4168{ 18, 19, 19, 10, 3,512, ZSTD_btultra2}, /* level 20.*/4169{ 18, 19, 19, 12, 3,512, ZSTD_btultra2}, /* level 21.*/4170{ 18, 19, 19, 13, 3,999, ZSTD_btultra2}, /* level 22.*/4171},4172{ /* for srcSize <= 128 KB */4173/* W, C, H, S, L, T, strat */4174{ 17, 12, 12, 1, 5, 1, ZSTD_fast }, /* base for negative levels */4175{ 17, 12, 13, 1, 6, 0, ZSTD_fast }, /* level 1 */4176{ 17, 13, 15, 1, 5, 0, ZSTD_fast }, /* level 2 */4177{ 17, 15, 16, 2, 5, 0, ZSTD_dfast }, /* level 3 */4178{ 17, 17, 17, 2, 4, 0, ZSTD_dfast }, /* level 4 */4179{ 17, 16, 17, 3, 4, 2, ZSTD_greedy }, /* level 5 */4180{ 17, 17, 17, 3, 4, 4, ZSTD_lazy }, /* level 6 */4181{ 17, 17, 17, 3, 4, 8, ZSTD_lazy2 }, /* level 7 */4182{ 17, 17, 17, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */4183{ 17, 17, 17, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */4184{ 17, 17, 17, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */4185{ 17, 17, 17, 5, 4, 8, ZSTD_btlazy2 }, /* level 11 */4186{ 17, 18, 17, 7, 4, 12, ZSTD_btlazy2 }, /* level 12 */4187{ 17, 18, 17, 3, 4, 12, ZSTD_btopt }, /* level 13.*/4188{ 17, 18, 17, 4, 3, 32, ZSTD_btopt }, /* level 14.*/4189{ 17, 18, 17, 6, 3,256, ZSTD_btopt }, /* level 15.*/4190{ 17, 18, 17, 6, 3,128, ZSTD_btultra }, /* level 16.*/4191{ 17, 18, 17, 8, 3,256, ZSTD_btultra }, /* level 17.*/4192{ 17, 18, 17, 10, 3,512, ZSTD_btultra }, /* level 18.*/4193{ 17, 18, 17, 5, 3,256, ZSTD_btultra2}, /* level 19.*/4194{ 17, 18, 17, 7, 3,512, ZSTD_btultra2}, /* level 20.*/4195{ 17, 18, 17, 9, 3,512, ZSTD_btultra2}, /* level 21.*/4196{ 17, 18, 17, 11, 3,999, ZSTD_btultra2}, /* level 22.*/4197},4198{ /* for srcSize <= 16 KB */4199/* W, C, H, S, L, T, strat */4200{ 14, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */4201{ 14, 14, 15, 1, 5, 0, ZSTD_fast }, /* level 1 */4202{ 14, 14, 15, 1, 4, 0, ZSTD_fast }, /* level 2 */4203{ 14, 14, 15, 2, 4, 0, ZSTD_dfast }, /* level 3 */4204{ 14, 14, 14, 4, 4, 2, ZSTD_greedy }, /* level 4 */4205{ 14, 14, 14, 3, 4, 4, ZSTD_lazy }, /* level 5.*/4206{ 14, 14, 14, 4, 4, 8, ZSTD_lazy2 }, /* level 6 */4207{ 14, 14, 14, 6, 4, 8, ZSTD_lazy2 }, /* level 7 */4208{ 14, 14, 14, 8, 4, 8, ZSTD_lazy2 }, /* level 8.*/4209{ 14, 15, 14, 5, 4, 8, ZSTD_btlazy2 }, /* level 9.*/4210{ 14, 15, 14, 9, 4, 8, ZSTD_btlazy2 }, /* level 10.*/4211{ 14, 15, 14, 3, 4, 12, ZSTD_btopt }, /* level 11.*/4212{ 14, 15, 14, 4, 3, 24, ZSTD_btopt }, /* level 12.*/4213{ 14, 15, 14, 5, 3, 32, ZSTD_btultra }, /* level 13.*/4214{ 14, 15, 15, 6, 3, 64, ZSTD_btultra }, /* level 14.*/4215{ 14, 15, 15, 7, 3,256, ZSTD_btultra }, /* level 15.*/4216{ 14, 15, 15, 5, 3, 48, ZSTD_btultra2}, /* level 16.*/4217{ 14, 15, 15, 6, 3,128, ZSTD_btultra2}, /* level 17.*/4218{ 14, 15, 15, 7, 3,256, ZSTD_btultra2}, /* level 18.*/4219{ 14, 15, 15, 8, 3,256, ZSTD_btultra2}, /* level 19.*/4220{ 14, 15, 15, 8, 3,512, ZSTD_btultra2}, /* level 20.*/4221{ 14, 15, 15, 9, 3,512, ZSTD_btultra2}, /* level 21.*/4222{ 14, 15, 15, 10, 3,999, ZSTD_btultra2}, /* level 22.*/4223},4224};42254226/*! ZSTD_getCParams_internal() :4227* @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.4228* Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown.4229* Use dictSize == 0 for unknown or unused. */4230static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)4231{4232int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN;4233size_t const addedSize = unknown && dictSize > 0 ? 500 : 0;4234U64 const rSize = unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize;4235U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);4236int row = compressionLevel;4237DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel);4238if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */4239if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */4240if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;4241{ ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];4242if (compressionLevel < 0) cp.targetLength = (unsigned)(-compressionLevel); /* acceleration factor */4243/* refine parameters based on srcSize & dictSize */4244return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize);4245}4246}42474248/*! ZSTD_getCParams() :4249* @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.4250* Size values are optional, provide 0 if not known or unused */4251ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)4252{4253if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;4254return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize);4255}42564257/*! ZSTD_getParams() :4258* same idea as ZSTD_getCParams()4259* @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).4260* Fields of `ZSTD_frameParameters` are set to default values */4261static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) {4262ZSTD_parameters params;4263ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize);4264DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel);4265memset(¶ms, 0, sizeof(params));4266params.cParams = cParams;4267params.fParams.contentSizeFlag = 1;4268return params;4269}42704271/*! ZSTD_getParams() :4272* same idea as ZSTD_getCParams()4273* @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`).4274* Fields of `ZSTD_frameParameters` are set to default values */4275ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) {4276if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN;4277return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize);4278}427942804281