diff options
Diffstat (limited to 'contrib/zstd/zstd_compress.c')
| -rw-r--r-- | contrib/zstd/zstd_compress.c | 4543 |
1 files changed, 3596 insertions, 947 deletions
diff --git a/contrib/zstd/zstd_compress.c b/contrib/zstd/zstd_compress.c index 526cb21f6..e11252c6a 100644 --- a/contrib/zstd/zstd_compress.c +++ b/contrib/zstd/zstd_compress.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2020, Yann Collet, Facebook, Inc. + * Copyright (c) Meta Platforms, Inc. and affiliates. * All rights reserved. * * This source code is licensed under both the BSD-style license (found in the @@ -11,14 +11,11 @@ /*-************************************* * Dependencies ***************************************/ -#include <limits.h> /* INT_MAX */ -#include <string.h> /* memset */ -#include "cpu.h" +#include "zstd_deps.h" /* INT_MAX, ZSTD_memset, ZSTD_memcpy */ #include "mem.h" #include "hist.h" /* HIST_countFast_wksp */ #define FSE_STATIC_LINKING_ONLY /* FSE_encodeSymbol */ #include "fse.h" -#define HUF_STATIC_LINKING_ONLY #include "huf.h" #include "zstd_compress_internal.h" #include "zstd_compress_sequences.h" @@ -29,20 +26,49 @@ #include "zstd_opt.h" #include "zstd_ldm.h" #include "zstd_compress_superblock.h" +#include "bits.h" /* ZSTD_highbit32 */ + +/* *************************************************************** +* Tuning parameters +*****************************************************************/ +/*! + * COMPRESS_HEAPMODE : + * Select how default decompression function ZSTD_compress() allocates its context, + * on stack (0, default), or into heap (1). + * Note that functions with explicit context such as ZSTD_compressCCtx() are unaffected. + */ +#ifndef ZSTD_COMPRESS_HEAPMODE +# define ZSTD_COMPRESS_HEAPMODE 0 +#endif +/*! + * ZSTD_HASHLOG3_MAX : + * Maximum size of the hash table dedicated to find 3-bytes matches, + * in log format, aka 17 => 1 << 17 == 128Ki positions. + * This structure is only used in zstd_opt. + * Since allocation is centralized for all strategies, it has to be known here. + * The actual (selected) size of the hash table is then stored in ZSTD_matchState_t.hashLog3, + * so that zstd_opt.c doesn't need to know about this constant. + */ +#ifndef ZSTD_HASHLOG3_MAX +# define ZSTD_HASHLOG3_MAX 17 +#endif /*-************************************* * Helper functions ***************************************/ /* ZSTD_compressBound() - * Note that the result from this function is only compatible with the "normal" - * full-block strategy. - * When there are a lot of small blocks due to frequent flush in streaming mode - * the overhead of headers can make the compressed data to be larger than the - * return value of ZSTD_compressBound(). + * Note that the result from this function is only valid for + * the one-pass compression functions. + * When employing the streaming mode, + * if flushes are frequently altering the size of blocks, + * the overhead from block headers can make the compressed data larger + * than the return value of ZSTD_compressBound(). */ size_t ZSTD_compressBound(size_t srcSize) { - return ZSTD_COMPRESSBOUND(srcSize); + size_t const r = ZSTD_COMPRESSBOUND(srcSize); + if (r==0) return ERROR(srcSize_wrong); + return r; } @@ -52,6 +78,7 @@ size_t ZSTD_compressBound(size_t srcSize) { struct ZSTD_CDict_s { const void* dictContent; size_t dictContentSize; + ZSTD_dictContentType_e dictContentType; /* The dictContentType the CDict was created with */ U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */ ZSTD_cwksp workspace; ZSTD_matchState_t matchState; @@ -59,6 +86,10 @@ struct ZSTD_CDict_s { ZSTD_customMem customMem; U32 dictID; int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */ + ZSTD_paramSwitch_e useRowMatchFinder; /* Indicates whether the CDict was created with params that would use + * row-based matchfinder. Unless the cdict is reloaded, we will use + * the same greedy/lazy matchfinder at compression time. + */ }; /* typedef'd to ZSTD_CDict within "zstd.h" */ ZSTD_CCtx* ZSTD_createCCtx(void) @@ -69,9 +100,9 @@ ZSTD_CCtx* ZSTD_createCCtx(void) static void ZSTD_initCCtx(ZSTD_CCtx* cctx, ZSTD_customMem memManager) { assert(cctx != NULL); - memset(cctx, 0, sizeof(*cctx)); + ZSTD_memset(cctx, 0, sizeof(*cctx)); cctx->customMem = memManager; - cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid()); + cctx->bmi2 = ZSTD_cpuSupportsBmi2(); { size_t const err = ZSTD_CCtx_reset(cctx, ZSTD_reset_parameters); assert(!ZSTD_isError(err)); (void)err; @@ -82,8 +113,8 @@ ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem) { ZSTD_STATIC_ASSERT(zcss_init==0); ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1)); - if (!customMem.customAlloc ^ !customMem.customFree) return NULL; - { ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_malloc(sizeof(ZSTD_CCtx), customMem); + if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; + { ZSTD_CCtx* const cctx = (ZSTD_CCtx*)ZSTD_customMalloc(sizeof(ZSTD_CCtx), customMem); if (!cctx) return NULL; ZSTD_initCCtx(cctx, customMem); return cctx; @@ -96,20 +127,20 @@ ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize) ZSTD_CCtx* cctx; if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL; /* minimum size */ if ((size_t)workspace & 7) return NULL; /* must be 8-aligned */ - ZSTD_cwksp_init(&ws, workspace, workspaceSize); + ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc); cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx)); if (cctx == NULL) return NULL; - memset(cctx, 0, sizeof(ZSTD_CCtx)); + ZSTD_memset(cctx, 0, sizeof(ZSTD_CCtx)); ZSTD_cwksp_move(&cctx->workspace, &ws); cctx->staticSize = workspaceSize; /* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */ - if (!ZSTD_cwksp_check_available(&cctx->workspace, HUF_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL; + if (!ZSTD_cwksp_check_available(&cctx->workspace, ENTROPY_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL; cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t)); cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t)); - cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, HUF_WORKSPACE_SIZE); + cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cctx->workspace, ENTROPY_WORKSPACE_SIZE); cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid()); return cctx; } @@ -119,10 +150,10 @@ ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize) */ static void ZSTD_clearAllDicts(ZSTD_CCtx* cctx) { - ZSTD_free(cctx->localDict.dictBuffer, cctx->customMem); + ZSTD_customFree(cctx->localDict.dictBuffer, cctx->customMem); ZSTD_freeCDict(cctx->localDict.cdict); - memset(&cctx->localDict, 0, sizeof(cctx->localDict)); - memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); + ZSTD_memset(&cctx->localDict, 0, sizeof(cctx->localDict)); + ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); cctx->cdict = NULL; } @@ -149,12 +180,9 @@ size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx) if (cctx==NULL) return 0; /* support free on NULL */ RETURN_ERROR_IF(cctx->staticSize, memory_allocation, "not compatible with static CCtx"); - { - int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx); + { int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx); ZSTD_freeCCtxContent(cctx); - if (!cctxInWorkspace) { - ZSTD_free(cctx, cctx->customMem); - } + if (!cctxInWorkspace) ZSTD_customFree(cctx, cctx->customMem); } return 0; } @@ -189,15 +217,116 @@ size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs) /* private API call, for dictBuilder only */ const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); } +/* Returns true if the strategy supports using a row based matchfinder */ +static int ZSTD_rowMatchFinderSupported(const ZSTD_strategy strategy) { + return (strategy >= ZSTD_greedy && strategy <= ZSTD_lazy2); +} + +/* Returns true if the strategy and useRowMatchFinder mode indicate that we will use the row based matchfinder + * for this compression. + */ +static int ZSTD_rowMatchFinderUsed(const ZSTD_strategy strategy, const ZSTD_paramSwitch_e mode) { + assert(mode != ZSTD_ps_auto); + return ZSTD_rowMatchFinderSupported(strategy) && (mode == ZSTD_ps_enable); +} + +/* Returns row matchfinder usage given an initial mode and cParams */ +static ZSTD_paramSwitch_e ZSTD_resolveRowMatchFinderMode(ZSTD_paramSwitch_e mode, + const ZSTD_compressionParameters* const cParams) { +#if defined(ZSTD_ARCH_X86_SSE2) || defined(ZSTD_ARCH_ARM_NEON) + int const kHasSIMD128 = 1; +#else + int const kHasSIMD128 = 0; +#endif + if (mode != ZSTD_ps_auto) return mode; /* if requested enabled, but no SIMD, we still will use row matchfinder */ + mode = ZSTD_ps_disable; + if (!ZSTD_rowMatchFinderSupported(cParams->strategy)) return mode; + if (kHasSIMD128) { + if (cParams->windowLog > 14) mode = ZSTD_ps_enable; + } else { + if (cParams->windowLog > 17) mode = ZSTD_ps_enable; + } + return mode; +} + +/* Returns block splitter usage (generally speaking, when using slower/stronger compression modes) */ +static ZSTD_paramSwitch_e ZSTD_resolveBlockSplitterMode(ZSTD_paramSwitch_e mode, + const ZSTD_compressionParameters* const cParams) { + if (mode != ZSTD_ps_auto) return mode; + return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 17) ? ZSTD_ps_enable : ZSTD_ps_disable; +} + +/* Returns 1 if the arguments indicate that we should allocate a chainTable, 0 otherwise */ +static int ZSTD_allocateChainTable(const ZSTD_strategy strategy, + const ZSTD_paramSwitch_e useRowMatchFinder, + const U32 forDDSDict) { + assert(useRowMatchFinder != ZSTD_ps_auto); + /* We always should allocate a chaintable if we are allocating a matchstate for a DDS dictionary matchstate. + * We do not allocate a chaintable if we are using ZSTD_fast, or are using the row-based matchfinder. + */ + return forDDSDict || ((strategy != ZSTD_fast) && !ZSTD_rowMatchFinderUsed(strategy, useRowMatchFinder)); +} + +/* Returns ZSTD_ps_enable if compression parameters are such that we should + * enable long distance matching (wlog >= 27, strategy >= btopt). + * Returns ZSTD_ps_disable otherwise. + */ +static ZSTD_paramSwitch_e ZSTD_resolveEnableLdm(ZSTD_paramSwitch_e mode, + const ZSTD_compressionParameters* const cParams) { + if (mode != ZSTD_ps_auto) return mode; + return (cParams->strategy >= ZSTD_btopt && cParams->windowLog >= 27) ? ZSTD_ps_enable : ZSTD_ps_disable; +} + +static int ZSTD_resolveExternalSequenceValidation(int mode) { + return mode; +} + +/* Resolves maxBlockSize to the default if no value is present. */ +static size_t ZSTD_resolveMaxBlockSize(size_t maxBlockSize) { + if (maxBlockSize == 0) { + return ZSTD_BLOCKSIZE_MAX; + } else { + return maxBlockSize; + } +} + +static ZSTD_paramSwitch_e ZSTD_resolveExternalRepcodeSearch(ZSTD_paramSwitch_e value, int cLevel) { + if (value != ZSTD_ps_auto) return value; + if (cLevel < 10) { + return ZSTD_ps_disable; + } else { + return ZSTD_ps_enable; + } +} + +/* Returns 1 if compression parameters are such that CDict hashtable and chaintable indices are tagged. + * If so, the tags need to be removed in ZSTD_resetCCtx_byCopyingCDict. */ +static int ZSTD_CDictIndicesAreTagged(const ZSTD_compressionParameters* const cParams) { + return cParams->strategy == ZSTD_fast || cParams->strategy == ZSTD_dfast; +} + static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams( ZSTD_compressionParameters cParams) { ZSTD_CCtx_params cctxParams; - memset(&cctxParams, 0, sizeof(cctxParams)); + /* should not matter, as all cParams are presumed properly defined */ + ZSTD_CCtxParams_init(&cctxParams, ZSTD_CLEVEL_DEFAULT); cctxParams.cParams = cParams; - cctxParams.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */ + + /* Adjust advanced params according to cParams */ + cctxParams.ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams.ldmParams.enableLdm, &cParams); + if (cctxParams.ldmParams.enableLdm == ZSTD_ps_enable) { + ZSTD_ldm_adjustParameters(&cctxParams.ldmParams, &cParams); + assert(cctxParams.ldmParams.hashLog >= cctxParams.ldmParams.bucketSizeLog); + assert(cctxParams.ldmParams.hashRateLog < 32); + } + cctxParams.useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams.useBlockSplitter, &cParams); + cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams); + cctxParams.validateSequences = ZSTD_resolveExternalSequenceValidation(cctxParams.validateSequences); + cctxParams.maxBlockSize = ZSTD_resolveMaxBlockSize(cctxParams.maxBlockSize); + cctxParams.searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(cctxParams.searchForExternalRepcodes, + cctxParams.compressionLevel); assert(!ZSTD_checkCParams(cParams)); - cctxParams.fParams.contentSizeFlag = 1; return cctxParams; } @@ -205,13 +334,12 @@ static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced( ZSTD_customMem customMem) { ZSTD_CCtx_params* params; - if (!customMem.customAlloc ^ !customMem.customFree) return NULL; - params = (ZSTD_CCtx_params*)ZSTD_calloc( + if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; + params = (ZSTD_CCtx_params*)ZSTD_customCalloc( sizeof(ZSTD_CCtx_params), customMem); if (!params) { return NULL; } + ZSTD_CCtxParams_init(params, ZSTD_CLEVEL_DEFAULT); params->customMem = customMem; - params->compressionLevel = ZSTD_CLEVEL_DEFAULT; - params->fParams.contentSizeFlag = 1; return params; } @@ -223,7 +351,7 @@ ZSTD_CCtx_params* ZSTD_createCCtxParams(void) size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params) { if (params == NULL) { return 0; } - ZSTD_free(params, params->customMem); + ZSTD_customFree(params, params->customMem); return 0; } @@ -234,35 +362,63 @@ size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params) size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel) { RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!"); - memset(cctxParams, 0, sizeof(*cctxParams)); + ZSTD_memset(cctxParams, 0, sizeof(*cctxParams)); cctxParams->compressionLevel = compressionLevel; cctxParams->fParams.contentSizeFlag = 1; return 0; } +#define ZSTD_NO_CLEVEL 0 + +/** + * Initializes `cctxParams` from `params` and `compressionLevel`. + * @param compressionLevel If params are derived from a compression level then that compression level, otherwise ZSTD_NO_CLEVEL. + */ +static void +ZSTD_CCtxParams_init_internal(ZSTD_CCtx_params* cctxParams, + const ZSTD_parameters* params, + int compressionLevel) +{ + assert(!ZSTD_checkCParams(params->cParams)); + ZSTD_memset(cctxParams, 0, sizeof(*cctxParams)); + cctxParams->cParams = params->cParams; + cctxParams->fParams = params->fParams; + /* Should not matter, as all cParams are presumed properly defined. + * But, set it for tracing anyway. + */ + cctxParams->compressionLevel = compressionLevel; + cctxParams->useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams->useRowMatchFinder, ¶ms->cParams); + cctxParams->useBlockSplitter = ZSTD_resolveBlockSplitterMode(cctxParams->useBlockSplitter, ¶ms->cParams); + cctxParams->ldmParams.enableLdm = ZSTD_resolveEnableLdm(cctxParams->ldmParams.enableLdm, ¶ms->cParams); + cctxParams->validateSequences = ZSTD_resolveExternalSequenceValidation(cctxParams->validateSequences); + cctxParams->maxBlockSize = ZSTD_resolveMaxBlockSize(cctxParams->maxBlockSize); + cctxParams->searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(cctxParams->searchForExternalRepcodes, compressionLevel); + DEBUGLOG(4, "ZSTD_CCtxParams_init_internal: useRowMatchFinder=%d, useBlockSplitter=%d ldm=%d", + cctxParams->useRowMatchFinder, cctxParams->useBlockSplitter, cctxParams->ldmParams.enableLdm); +} + size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params) { RETURN_ERROR_IF(!cctxParams, GENERIC, "NULL pointer!"); FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , ""); - memset(cctxParams, 0, sizeof(*cctxParams)); - assert(!ZSTD_checkCParams(params.cParams)); - cctxParams->cParams = params.cParams; - cctxParams->fParams = params.fParams; - cctxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */ + ZSTD_CCtxParams_init_internal(cctxParams, ¶ms, ZSTD_NO_CLEVEL); return 0; } -/* ZSTD_assignParamsToCCtxParams() : - * params is presumed valid at this stage */ -static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams( - const ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params) +/** + * Sets cctxParams' cParams and fParams from params, but otherwise leaves them alone. + * @param params Validated zstd parameters. + */ +static void ZSTD_CCtxParams_setZstdParams( + ZSTD_CCtx_params* cctxParams, const ZSTD_parameters* params) { - ZSTD_CCtx_params ret = *cctxParams; assert(!ZSTD_checkCParams(params->cParams)); - ret.cParams = params->cParams; - ret.fParams = params->fParams; - ret.compressionLevel = ZSTD_CLEVEL_DEFAULT; /* should not matter, as all cParams are presumed properly defined */ - return ret; + cctxParams->cParams = params->cParams; + cctxParams->fParams = params->fParams; + /* Should not matter, as all cParams are presumed properly defined. + * But, set it for tracing anyway. + */ + cctxParams->compressionLevel = ZSTD_NO_CLEVEL; } ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param) @@ -354,11 +510,16 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param) #endif return bounds; - case ZSTD_c_enableLongDistanceMatching: + case ZSTD_c_enableDedicatedDictSearch: bounds.lowerBound = 0; bounds.upperBound = 1; return bounds; + case ZSTD_c_enableLongDistanceMatching: + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; + return bounds; + case ZSTD_c_ldmHashLog: bounds.lowerBound = ZSTD_LDM_HASHLOG_MIN; bounds.upperBound = ZSTD_LDM_HASHLOG_MAX; @@ -397,15 +558,15 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param) return bounds; case ZSTD_c_forceAttachDict: - ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceCopy); + ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceLoad); bounds.lowerBound = ZSTD_dictDefaultAttach; bounds.upperBound = ZSTD_dictForceLoad; /* note : how to ensure at compile time that this is the highest value enum ? */ return bounds; case ZSTD_c_literalCompressionMode: - ZSTD_STATIC_ASSERT(ZSTD_lcm_auto < ZSTD_lcm_huffman && ZSTD_lcm_huffman < ZSTD_lcm_uncompressed); - bounds.lowerBound = ZSTD_lcm_auto; - bounds.upperBound = ZSTD_lcm_uncompressed; + ZSTD_STATIC_ASSERT(ZSTD_ps_auto < ZSTD_ps_enable && ZSTD_ps_enable < ZSTD_ps_disable); + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; return bounds; case ZSTD_c_targetCBlockSize: @@ -418,6 +579,57 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param) bounds.upperBound = ZSTD_SRCSIZEHINT_MAX; return bounds; + case ZSTD_c_stableInBuffer: + case ZSTD_c_stableOutBuffer: + bounds.lowerBound = (int)ZSTD_bm_buffered; + bounds.upperBound = (int)ZSTD_bm_stable; + return bounds; + + case ZSTD_c_blockDelimiters: + bounds.lowerBound = (int)ZSTD_sf_noBlockDelimiters; + bounds.upperBound = (int)ZSTD_sf_explicitBlockDelimiters; + return bounds; + + case ZSTD_c_validateSequences: + bounds.lowerBound = 0; + bounds.upperBound = 1; + return bounds; + + case ZSTD_c_useBlockSplitter: + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; + return bounds; + + case ZSTD_c_useRowMatchFinder: + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; + return bounds; + + case ZSTD_c_deterministicRefPrefix: + bounds.lowerBound = 0; + bounds.upperBound = 1; + return bounds; + + case ZSTD_c_prefetchCDictTables: + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; + return bounds; + + case ZSTD_c_enableSeqProducerFallback: + bounds.lowerBound = 0; + bounds.upperBound = 1; + return bounds; + + case ZSTD_c_maxBlockSize: + bounds.lowerBound = ZSTD_BLOCKSIZE_MAX_MIN; + bounds.upperBound = ZSTD_BLOCKSIZE_MAX; + return bounds; + + case ZSTD_c_searchForExternalRepcodes: + bounds.lowerBound = (int)ZSTD_ps_auto; + bounds.upperBound = (int)ZSTD_ps_disable; + return bounds; + default: bounds.error = ERROR(parameter_unsupported); return bounds; @@ -465,6 +677,7 @@ static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param) case ZSTD_c_jobSize: case ZSTD_c_overlapLog: case ZSTD_c_rsyncable: + case ZSTD_c_enableDedicatedDictSearch: case ZSTD_c_enableLongDistanceMatching: case ZSTD_c_ldmHashLog: case ZSTD_c_ldmMinMatch: @@ -474,6 +687,17 @@ static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param) case ZSTD_c_literalCompressionMode: case ZSTD_c_targetCBlockSize: case ZSTD_c_srcSizeHint: + case ZSTD_c_stableInBuffer: + case ZSTD_c_stableOutBuffer: + case ZSTD_c_blockDelimiters: + case ZSTD_c_validateSequences: + case ZSTD_c_useBlockSplitter: + case ZSTD_c_useRowMatchFinder: + case ZSTD_c_deterministicRefPrefix: + case ZSTD_c_prefetchCDictTables: + case ZSTD_c_enableSeqProducerFallback: + case ZSTD_c_maxBlockSize: + case ZSTD_c_searchForExternalRepcodes: default: return 0; } @@ -486,7 +710,7 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value) if (ZSTD_isUpdateAuthorized(param)) { cctx->cParamsChanged = 1; } else { - RETURN_ERROR(stage_wrong, "can only set params in ctx init stage"); + RETURN_ERROR(stage_wrong, "can only set params in cctx init stage"); } } switch(param) @@ -515,12 +739,24 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value) case ZSTD_c_jobSize: case ZSTD_c_overlapLog: case ZSTD_c_rsyncable: + case ZSTD_c_enableDedicatedDictSearch: case ZSTD_c_enableLongDistanceMatching: case ZSTD_c_ldmHashLog: case ZSTD_c_ldmMinMatch: case ZSTD_c_ldmBucketSizeLog: case ZSTD_c_targetCBlockSize: case ZSTD_c_srcSizeHint: + case ZSTD_c_stableInBuffer: + case ZSTD_c_stableOutBuffer: + case ZSTD_c_blockDelimiters: + case ZSTD_c_validateSequences: + case ZSTD_c_useBlockSplitter: + case ZSTD_c_useRowMatchFinder: + case ZSTD_c_deterministicRefPrefix: + case ZSTD_c_prefetchCDictTables: + case ZSTD_c_enableSeqProducerFallback: + case ZSTD_c_maxBlockSize: + case ZSTD_c_searchForExternalRepcodes: break; default: RETURN_ERROR(parameter_unsupported, "unknown parameter"); @@ -541,9 +777,10 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, case ZSTD_c_compressionLevel : { FORWARD_IF_ERROR(ZSTD_cParam_clampBounds(param, &value), ""); - if (value) { /* 0 : does not change current level */ + if (value == 0) + CCtxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT; /* 0 == default */ + else CCtxParams->compressionLevel = value; - } if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel; return 0; /* return type (size_t) cannot represent negative values */ } @@ -575,12 +812,12 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, case ZSTD_c_minMatch : if (value!=0) /* 0 => use default */ BOUNDCHECK(ZSTD_c_minMatch, value); - CCtxParams->cParams.minMatch = value; + CCtxParams->cParams.minMatch = (U32)value; return CCtxParams->cParams.minMatch; case ZSTD_c_targetLength : BOUNDCHECK(ZSTD_c_targetLength, value); - CCtxParams->cParams.targetLength = value; + CCtxParams->cParams.targetLength = (U32)value; return CCtxParams->cParams.targetLength; case ZSTD_c_strategy : @@ -593,12 +830,12 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, /* Content size written in frame header _when known_ (default:1) */ DEBUGLOG(4, "set content size flag = %u", (value!=0)); CCtxParams->fParams.contentSizeFlag = value != 0; - return CCtxParams->fParams.contentSizeFlag; + return (size_t)CCtxParams->fParams.contentSizeFlag; case ZSTD_c_checksumFlag : /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */ CCtxParams->fParams.checksumFlag = value != 0; - return CCtxParams->fParams.checksumFlag; + return (size_t)CCtxParams->fParams.checksumFlag; case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */ DEBUGLOG(4, "set dictIDFlag = %u", (value!=0)); @@ -607,18 +844,18 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, case ZSTD_c_forceMaxWindow : CCtxParams->forceWindow = (value != 0); - return CCtxParams->forceWindow; + return (size_t)CCtxParams->forceWindow; case ZSTD_c_forceAttachDict : { const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value; - BOUNDCHECK(ZSTD_c_forceAttachDict, pref); + BOUNDCHECK(ZSTD_c_forceAttachDict, (int)pref); CCtxParams->attachDictPref = pref; return CCtxParams->attachDictPref; } case ZSTD_c_literalCompressionMode : { - const ZSTD_literalCompressionMode_e lcm = (ZSTD_literalCompressionMode_e)value; - BOUNDCHECK(ZSTD_c_literalCompressionMode, lcm); + const ZSTD_paramSwitch_e lcm = (ZSTD_paramSwitch_e)value; + BOUNDCHECK(ZSTD_c_literalCompressionMode, (int)lcm); CCtxParams->literalCompressionMode = lcm; return CCtxParams->literalCompressionMode; } @@ -667,57 +904,118 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams, return CCtxParams->rsyncable; #endif + case ZSTD_c_enableDedicatedDictSearch : + CCtxParams->enableDedicatedDictSearch = (value!=0); + return (size_t)CCtxParams->enableDedicatedDictSearch; + case ZSTD_c_enableLongDistanceMatching : - CCtxParams->ldmParams.enableLdm = (value!=0); + BOUNDCHECK(ZSTD_c_enableLongDistanceMatching, value); + CCtxParams->ldmParams.enableLdm = (ZSTD_paramSwitch_e)value; return CCtxParams->ldmParams.enableLdm; case ZSTD_c_ldmHashLog : if (value!=0) /* 0 ==> auto */ BOUNDCHECK(ZSTD_c_ldmHashLog, value); - CCtxParams->ldmParams.hashLog = value; + CCtxParams->ldmParams.hashLog = (U32)value; return CCtxParams->ldmParams.hashLog; case ZSTD_c_ldmMinMatch : if (value!=0) /* 0 ==> default */ BOUNDCHECK(ZSTD_c_ldmMinMatch, value); - CCtxParams->ldmParams.minMatchLength = value; + CCtxParams->ldmParams.minMatchLength = (U32)value; return CCtxParams->ldmParams.minMatchLength; case ZSTD_c_ldmBucketSizeLog : if (value!=0) /* 0 ==> default */ BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value); - CCtxParams->ldmParams.bucketSizeLog = value; + CCtxParams->ldmParams.bucketSizeLog = (U32)value; return CCtxParams->ldmParams.bucketSizeLog; case ZSTD_c_ldmHashRateLog : - RETURN_ERROR_IF(value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN, - parameter_outOfBound, "Param out of bounds!"); - CCtxParams->ldmParams.hashRateLog = value; + if (value!=0) /* 0 ==> default */ + BOUNDCHECK(ZSTD_c_ldmHashRateLog, value); + CCtxParams->ldmParams.hashRateLog = (U32)value; return CCtxParams->ldmParams.hashRateLog; case ZSTD_c_targetCBlockSize : if (value!=0) /* 0 ==> default */ BOUNDCHECK(ZSTD_c_targetCBlockSize, value); - CCtxParams->targetCBlockSize = value; + CCtxParams->targetCBlockSize = (U32)value; return CCtxParams->targetCBlockSize; case ZSTD_c_srcSizeHint : if (value!=0) /* 0 ==> default */ BOUNDCHECK(ZSTD_c_srcSizeHint, value); CCtxParams->srcSizeHint = value; - return CCtxParams->srcSizeHint; + return (size_t)CCtxParams->srcSizeHint; + + case ZSTD_c_stableInBuffer: + BOUNDCHECK(ZSTD_c_stableInBuffer, value); + CCtxParams->inBufferMode = (ZSTD_bufferMode_e)value; + return CCtxParams->inBufferMode; + + case ZSTD_c_stableOutBuffer: + BOUNDCHECK(ZSTD_c_stableOutBuffer, value); + CCtxParams->outBufferMode = (ZSTD_bufferMode_e)value; + return CCtxParams->outBufferMode; + + case ZSTD_c_blockDelimiters: + BOUNDCHECK(ZSTD_c_blockDelimiters, value); + CCtxParams->blockDelimiters = (ZSTD_sequenceFormat_e)value; + return CCtxParams->blockDelimiters; + + case ZSTD_c_validateSequences: + BOUNDCHECK(ZSTD_c_validateSequences, value); + CCtxParams->validateSequences = value; + return CCtxParams->validateSequences; + + case ZSTD_c_useBlockSplitter: + BOUNDCHECK(ZSTD_c_useBlockSplitter, value); + CCtxParams->useBlockSplitter = (ZSTD_paramSwitch_e)value; + return CCtxParams->useBlockSplitter; + + case ZSTD_c_useRowMatchFinder: + BOUNDCHECK(ZSTD_c_useRowMatchFinder, value); + CCtxParams->useRowMatchFinder = (ZSTD_paramSwitch_e)value; + return CCtxParams->useRowMatchFinder; + + case ZSTD_c_deterministicRefPrefix: + BOUNDCHECK(ZSTD_c_deterministicRefPrefix, value); + CCtxParams->deterministicRefPrefix = !!value; + return CCtxParams->deterministicRefPrefix; + + case ZSTD_c_prefetchCDictTables: + BOUNDCHECK(ZSTD_c_prefetchCDictTables, value); + CCtxParams->prefetchCDictTables = (ZSTD_paramSwitch_e)value; + return CCtxParams->prefetchCDictTables; + + case ZSTD_c_enableSeqProducerFallback: + BOUNDCHECK(ZSTD_c_enableSeqProducerFallback, value); + CCtxParams->enableMatchFinderFallback = value; + return CCtxParams->enableMatchFinderFallback; + + case ZSTD_c_maxBlockSize: + if (value!=0) /* 0 ==> default */ + BOUNDCHECK(ZSTD_c_maxBlockSize, value); + CCtxParams->maxBlockSize = value; + return CCtxParams->maxBlockSize; + + case ZSTD_c_searchForExternalRepcodes: + BOUNDCHECK(ZSTD_c_searchForExternalRepcodes, value); + CCtxParams->searchForExternalRepcodes = (ZSTD_paramSwitch_e)value; + return CCtxParams->searchForExternalRepcodes; default: RETURN_ERROR(parameter_unsupported, "unknown parameter"); } } -size_t ZSTD_CCtx_getParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int* value) +size_t ZSTD_CCtx_getParameter(ZSTD_CCtx const* cctx, ZSTD_cParameter param, int* value) { return ZSTD_CCtxParams_getParameter(&cctx->requestedParams, param, value); } size_t ZSTD_CCtxParams_getParameter( - ZSTD_CCtx_params* CCtxParams, ZSTD_cParameter param, int* value) + ZSTD_CCtx_params const* CCtxParams, ZSTD_cParameter param, int* value) { switch(param) { @@ -794,6 +1092,9 @@ size_t ZSTD_CCtxParams_getParameter( *value = CCtxParams->rsyncable; break; #endif + case ZSTD_c_enableDedicatedDictSearch : + *value = CCtxParams->enableDedicatedDictSearch; + break; case ZSTD_c_enableLongDistanceMatching : *value = CCtxParams->ldmParams.enableLdm; break; @@ -815,6 +1116,39 @@ size_t ZSTD_CCtxParams_getParameter( case ZSTD_c_srcSizeHint : *value = (int)CCtxParams->srcSizeHint; break; + case ZSTD_c_stableInBuffer : + *value = (int)CCtxParams->inBufferMode; + break; + case ZSTD_c_stableOutBuffer : + *value = (int)CCtxParams->outBufferMode; + break; + case ZSTD_c_blockDelimiters : + *value = (int)CCtxParams->blockDelimiters; + break; + case ZSTD_c_validateSequences : + *value = (int)CCtxParams->validateSequences; + break; + case ZSTD_c_useBlockSplitter : + *value = (int)CCtxParams->useBlockSplitter; + break; + case ZSTD_c_useRowMatchFinder : + *value = (int)CCtxParams->useRowMatchFinder; + break; + case ZSTD_c_deterministicRefPrefix: + *value = (int)CCtxParams->deterministicRefPrefix; + break; + case ZSTD_c_prefetchCDictTables: + *value = (int)CCtxParams->prefetchCDictTables; + break; + case ZSTD_c_enableSeqProducerFallback: + *value = CCtxParams->enableMatchFinderFallback; + break; + case ZSTD_c_maxBlockSize: + *value = (int)CCtxParams->maxBlockSize; + break; + case ZSTD_c_searchForExternalRepcodes: + *value = (int)CCtxParams->searchForExternalRepcodes; + break; default: RETURN_ERROR(parameter_unsupported, "unknown parameter"); } return 0; @@ -841,15 +1175,38 @@ size_t ZSTD_CCtx_setParametersUsingCCtxParams( return 0; } -ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize) +size_t ZSTD_CCtx_setCParams(ZSTD_CCtx* cctx, ZSTD_compressionParameters cparams) { - DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %u bytes", (U32)pledgedSrcSize); + DEBUGLOG(4, "ZSTD_CCtx_setCParams"); + assert(cctx != NULL); + if (cctx->streamStage != zcss_init) { + /* All parameters in @cparams are allowed to be updated during MT compression. + * This must be signaled, so that MT compression picks up the changes */ + cctx->cParamsChanged = 1; + } + /* only update if parameters are valid */ + FORWARD_IF_ERROR(ZSTD_checkCParams(cparams), ""); + cctx->requestedParams.cParams = cparams; + return 0; +} + +size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize) +{ + DEBUGLOG(4, "ZSTD_CCtx_setPledgedSrcSize to %llu bytes", pledgedSrcSize); RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, "Can't set pledgedSrcSize when not in init stage."); cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1; return 0; } +static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams( + int const compressionLevel, + size_t const dictSize); +static int ZSTD_dedicatedDictSearch_isSupported( + const ZSTD_compressionParameters* cParams); +static void ZSTD_dedicatedDictSearch_revertCParams( + ZSTD_compressionParameters* cParams); + /** * Initializes the local dict using the requested parameters. * NOTE: This does not use the pledged src size, because it may be used for more @@ -858,8 +1215,6 @@ ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long lo static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx) { ZSTD_localDict* const dl = &cctx->localDict; - ZSTD_compressionParameters const cParams = ZSTD_getCParamsFromCCtxParams( - &cctx->requestedParams, ZSTD_CONTENTSIZE_UNKNOWN, dl->dictSize); if (dl->dict == NULL) { /* No local dictionary. */ assert(dl->dictBuffer == NULL); @@ -876,12 +1231,12 @@ static size_t ZSTD_initLocalDict(ZSTD_CCtx* cctx) assert(cctx->cdict == NULL); assert(cctx->prefixDict.dict == NULL); - dl->cdict = ZSTD_createCDict_advanced( + dl->cdict = ZSTD_createCDict_advanced2( dl->dict, dl->dictSize, ZSTD_dlm_byRef, dl->dictContentType, - cParams, + &cctx->requestedParams, cctx->customMem); RETURN_ERROR_IF(!dl->cdict, memory_allocation, "ZSTD_createCDict_advanced failed"); cctx->cdict = dl->cdict; @@ -894,8 +1249,6 @@ size_t ZSTD_CCtx_loadDictionary_advanced( { RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, "Can't load a dictionary when ctx is not in init stage."); - RETURN_ERROR_IF(cctx->staticSize, memory_allocation, - "no malloc for static CCtx"); DEBUGLOG(4, "ZSTD_CCtx_loadDictionary_advanced (size: %u)", (U32)dictSize); ZSTD_clearAllDicts(cctx); /* in case one already exists */ if (dict == NULL || dictSize == 0) /* no dictionary mode */ @@ -903,9 +1256,12 @@ size_t ZSTD_CCtx_loadDictionary_advanced( if (dictLoadMethod == ZSTD_dlm_byRef) { cctx->localDict.dict = dict; } else { - void* dictBuffer = ZSTD_malloc(dictSize, cctx->customMem); + void* dictBuffer; + RETURN_ERROR_IF(cctx->staticSize, memory_allocation, + "no malloc for static CCtx"); + dictBuffer = ZSTD_customMalloc(dictSize, cctx->customMem); RETURN_ERROR_IF(!dictBuffer, memory_allocation, "NULL pointer!"); - memcpy(dictBuffer, dict, dictSize); + ZSTD_memcpy(dictBuffer, dict, dictSize); cctx->localDict.dictBuffer = dictBuffer; cctx->localDict.dict = dictBuffer; } @@ -914,14 +1270,14 @@ size_t ZSTD_CCtx_loadDictionary_advanced( return 0; } -ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference( +size_t ZSTD_CCtx_loadDictionary_byReference( ZSTD_CCtx* cctx, const void* dict, size_t dictSize) { return ZSTD_CCtx_loadDictionary_advanced( cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto); } -ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize) +size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize) { return ZSTD_CCtx_loadDictionary_advanced( cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto); @@ -938,6 +1294,14 @@ size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) return 0; } +size_t ZSTD_CCtx_refThreadPool(ZSTD_CCtx* cctx, ZSTD_threadPool* pool) +{ + RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, + "Can't ref a pool when ctx not in init stage."); + cctx->pool = pool; + return 0; +} + size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize) { return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dct_rawContent); @@ -971,6 +1335,7 @@ size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset) RETURN_ERROR_IF(cctx->streamStage != zcss_init, stage_wrong, "Can't reset parameters only when not in init stage."); ZSTD_clearAllDicts(cctx); + ZSTD_memset(&cctx->externalMatchCtx, 0, sizeof(cctx->externalMatchCtx)); return ZSTD_CCtxParams_reset(&cctx->requestedParams); } return 0; @@ -1022,42 +1387,141 @@ U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat) return hashLog - btScale; } +/** ZSTD_dictAndWindowLog() : + * Returns an adjusted window log that is large enough to fit the source and the dictionary. + * The zstd format says that the entire dictionary is valid if one byte of the dictionary + * is within the window. So the hashLog and chainLog should be large enough to reference both + * the dictionary and the window. So we must use this adjusted dictAndWindowLog when downsizing + * the hashLog and windowLog. + * NOTE: srcSize must not be ZSTD_CONTENTSIZE_UNKNOWN. + */ +static U32 ZSTD_dictAndWindowLog(U32 windowLog, U64 srcSize, U64 dictSize) +{ + const U64 maxWindowSize = 1ULL << ZSTD_WINDOWLOG_MAX; + /* No dictionary ==> No change */ + if (dictSize == 0) { + return windowLog; + } + assert(windowLog <= ZSTD_WINDOWLOG_MAX); + assert(srcSize != ZSTD_CONTENTSIZE_UNKNOWN); /* Handled in ZSTD_adjustCParams_internal() */ + { + U64 const windowSize = 1ULL << windowLog; + U64 const dictAndWindowSize = dictSize + windowSize; + /* If the window size is already large enough to fit both the source and the dictionary + * then just use the window size. Otherwise adjust so that it fits the dictionary and + * the window. + */ + if (windowSize >= dictSize + srcSize) { + return windowLog; /* Window size large enough already */ + } else if (dictAndWindowSize >= maxWindowSize) { + return ZSTD_WINDOWLOG_MAX; /* Larger than max window log */ + } else { + return ZSTD_highbit32((U32)dictAndWindowSize - 1) + 1; + } + } +} + /** ZSTD_adjustCParams_internal() : * optimize `cPar` for a specified input (`srcSize` and `dictSize`). * mostly downsize to reduce memory consumption and initialization latency. * `srcSize` can be ZSTD_CONTENTSIZE_UNKNOWN when not known. + * `mode` is the mode for parameter adjustment. See docs for `ZSTD_cParamMode_e`. * note : `srcSize==0` means 0! * condition : cPar is presumed validated (can be checked using ZSTD_checkCParams()). */ static ZSTD_compressionParameters ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar, unsigned long long srcSize, - size_t dictSize) + size_t dictSize, + ZSTD_cParamMode_e mode, + ZSTD_paramSwitch_e useRowMatchFinder) { - static const U64 minSrcSize = 513; /* (1<<9) + 1 */ - static const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1); + const U64 minSrcSize = 513; /* (1<<9) + 1 */ + const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1); assert(ZSTD_checkCParams(cPar)==0); - if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN) - srcSize = minSrcSize; + switch (mode) { + case ZSTD_cpm_unknown: + case ZSTD_cpm_noAttachDict: + /* If we don't know the source size, don't make any + * assumptions about it. We will already have selected + * smaller parameters if a dictionary is in use. + */ + break; + case ZSTD_cpm_createCDict: + /* Assume a small source size when creating a dictionary + * with an unknown source size. + */ + if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN) + srcSize = minSrcSize; + break; + case ZSTD_cpm_attachDict: + /* Dictionary has its own dedicated parameters which have + * already been selected. We are selecting parameters + * for only the source. + */ + dictSize = 0; + break; + default: + assert(0); + break; + } /* resize windowLog if input is small enough, to use less memory */ - if ( (srcSize < maxWindowResize) - && (dictSize < maxWindowResize) ) { + if ( (srcSize <= maxWindowResize) + && (dictSize <= maxWindowResize) ) { U32 const tSize = (U32)(srcSize + dictSize); static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN; U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN : ZSTD_highbit32(tSize-1) + 1; if (cPar.windowLog > srcLog) cPar.windowLog = srcLog; } - if (cPar.hashLog > cPar.windowLog+1) cPar.hashLog = cPar.windowLog+1; - { U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy); - if (cycleLog > cPar.windowLog) - cPar.chainLog -= (cycleLog - cPar.windowLog); + if (srcSize != ZSTD_CONTENTSIZE_UNKNOWN) { + U32 const dictAndWindowLog = ZSTD_dictAndWindowLog(cPar.windowLog, (U64)srcSize, (U64)dictSize); + U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy); + if (cPar.hashLog > dictAndWindowLog+1) cPar.hashLog = dictAndWindowLog+1; + if (cycleLog > dictAndWindowLog) + cPar.chainLog -= (cycleLog - dictAndWindowLog); } if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN) cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */ + /* We can't use more than 32 bits of hash in total, so that means that we require: + * (hashLog + 8) <= 32 && (chainLog + 8) <= 32 + */ + if (mode == ZSTD_cpm_createCDict && ZSTD_CDictIndicesAreTagged(&cPar)) { + U32 const maxShortCacheHashLog = 32 - ZSTD_SHORT_CACHE_TAG_BITS; + if (cPar.hashLog > maxShortCacheHashLog) { + cPar.hashLog = maxShortCacheHashLog; + } + if (cPar.chainLog > maxShortCacheHashLog) { + cPar.chainLog = maxShortCacheHashLog; + } + } + + + /* At this point, we aren't 100% sure if we are using the row match finder. + * Unless it is explicitly disabled, conservatively assume that it is enabled. + * In this case it will only be disabled for small sources, so shrinking the + * hash log a little bit shouldn't result in any ratio loss. + */ + if (useRowMatchFinder == ZSTD_ps_auto) + useRowMatchFinder = ZSTD_ps_enable; + + /* We can't hash more than 32-bits in total. So that means that we require: + * (hashLog - rowLog + 8) <= 32 + */ + if (ZSTD_rowMatchFinderUsed(cPar.strategy, useRowMatchFinder)) { + /* Switch to 32-entry rows if searchLog is 5 (or more) */ + U32 const rowLog = BOUNDED(4, cPar.searchLog, 6); + U32 const maxRowHashLog = 32 - ZSTD_ROW_HASH_TAG_BITS; + U32 const maxHashLog = maxRowHashLog + rowLog; + assert(cPar.hashLog >= rowLog); + if (cPar.hashLog > maxHashLog) { + cPar.hashLog = maxHashLog; + } + } + return cPar; } @@ -1068,38 +1532,50 @@ ZSTD_adjustCParams(ZSTD_compressionParameters cPar, { cPar = ZSTD_clampCParams(cPar); /* resulting cPar is necessarily valid (all parameters within range) */ if (srcSize == 0) srcSize = ZSTD_CONTENTSIZE_UNKNOWN; - return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize); + return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize, ZSTD_cpm_unknown, ZSTD_ps_auto); } -static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize); -static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize); +static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode); +static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode); + +static void ZSTD_overrideCParams( + ZSTD_compressionParameters* cParams, + const ZSTD_compressionParameters* overrides) +{ + if (overrides->windowLog) cParams->windowLog = overrides->windowLog; + if (overrides->hashLog) cParams->hashLog = overrides->hashLog; + if (overrides->chainLog) cParams->chainLog = overrides->chainLog; + if (overrides->searchLog) cParams->searchLog = overrides->searchLog; + if (overrides->minMatch) cParams->minMatch = overrides->minMatch; + if (overrides->targetLength) cParams->targetLength = overrides->targetLength; + if (overrides->strategy) cParams->strategy = overrides->strategy; +} ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams( - const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize) + const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) { ZSTD_compressionParameters cParams; if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) { srcSizeHint = CCtxParams->srcSizeHint; } - cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize); - if (CCtxParams->ldmParams.enableLdm) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG; - if (CCtxParams->cParams.windowLog) cParams.windowLog = CCtxParams->cParams.windowLog; - if (CCtxParams->cParams.hashLog) cParams.hashLog = CCtxParams->cParams.hashLog; - if (CCtxParams->cParams.chainLog) cParams.chainLog = CCtxParams->cParams.chainLog; - if (CCtxParams->cParams.searchLog) cParams.searchLog = CCtxParams->cParams.searchLog; - if (CCtxParams->cParams.minMatch) cParams.minMatch = CCtxParams->cParams.minMatch; - if (CCtxParams->cParams.targetLength) cParams.targetLength = CCtxParams->cParams.targetLength; - if (CCtxParams->cParams.strategy) cParams.strategy = CCtxParams->cParams.strategy; + cParams = ZSTD_getCParams_internal(CCtxParams->compressionLevel, srcSizeHint, dictSize, mode); + if (CCtxParams->ldmParams.enableLdm == ZSTD_ps_enable) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG; + ZSTD_overrideCParams(&cParams, &CCtxParams->cParams); assert(!ZSTD_checkCParams(cParams)); /* srcSizeHint == 0 means 0 */ - return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize); + return ZSTD_adjustCParams_internal(cParams, srcSizeHint, dictSize, mode, CCtxParams->useRowMatchFinder); } static size_t ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams, + const ZSTD_paramSwitch_e useRowMatchFinder, + const U32 enableDedicatedDictSearch, const U32 forCCtx) { - size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog); + /* chain table size should be 0 for fast or row-hash strategies */ + size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, enableDedicatedDictSearch && !forCCtx) + ? ((size_t)1 << cParams->chainLog) + : 0; size_t const hSize = ((size_t)1) << cParams->hashLog; U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0; size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0; @@ -1109,71 +1585,131 @@ ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams, + hSize * sizeof(U32) + h3Size * sizeof(U32); size_t const optPotentialSpace = - ZSTD_cwksp_alloc_size((MaxML+1) * sizeof(U32)) - + ZSTD_cwksp_alloc_size((MaxLL+1) * sizeof(U32)) - + ZSTD_cwksp_alloc_size((MaxOff+1) * sizeof(U32)) - + ZSTD_cwksp_alloc_size((1<<Litbits) * sizeof(U32)) - + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t)) - + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t)); + ZSTD_cwksp_aligned_alloc_size((MaxML+1) * sizeof(U32)) + + ZSTD_cwksp_aligned_alloc_size((MaxLL+1) * sizeof(U32)) + + ZSTD_cwksp_aligned_alloc_size((MaxOff+1) * sizeof(U32)) + + ZSTD_cwksp_aligned_alloc_size((1<<Litbits) * sizeof(U32)) + + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t)) + + ZSTD_cwksp_aligned_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t)); + size_t const lazyAdditionalSpace = ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder) + ? ZSTD_cwksp_aligned_alloc_size(hSize*sizeof(U16)) + : 0; size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt)) ? optPotentialSpace : 0; + size_t const slackSpace = ZSTD_cwksp_slack_space_required(); + + /* tables are guaranteed to be sized in multiples of 64 bytes (or 16 uint32_t) */ + ZSTD_STATIC_ASSERT(ZSTD_HASHLOG_MIN >= 4 && ZSTD_WINDOWLOG_MIN >= 4 && ZSTD_CHAINLOG_MIN >= 4); + assert(useRowMatchFinder != ZSTD_ps_auto); + DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u", (U32)chainSize, (U32)hSize, (U32)h3Size); - return tableSpace + optSpace; + return tableSpace + optSpace + slackSpace + lazyAdditionalSpace; +} + +/* Helper function for calculating memory requirements. + * Gives a tighter bound than ZSTD_sequenceBound() by taking minMatch into account. */ +static size_t ZSTD_maxNbSeq(size_t blockSize, unsigned minMatch, int useSequenceProducer) { + U32 const divider = (minMatch==3 || useSequenceProducer) ? 3 : 4; + return blockSize / divider; +} + +static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal( + const ZSTD_compressionParameters* cParams, + const ldmParams_t* ldmParams, + const int isStatic, + const ZSTD_paramSwitch_e useRowMatchFinder, + const size_t buffInSize, + const size_t buffOutSize, + const U64 pledgedSrcSize, + int useSequenceProducer, + size_t maxBlockSize) +{ + size_t const windowSize = (size_t) BOUNDED(1ULL, 1ULL << cParams->windowLog, pledgedSrcSize); + size_t const blockSize = MIN(ZSTD_resolveMaxBlockSize(maxBlockSize), windowSize); + size_t const maxNbSeq = ZSTD_maxNbSeq(blockSize, cParams->minMatch, useSequenceProducer); + size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize) + + ZSTD_cwksp_aligned_alloc_size(maxNbSeq * sizeof(seqDef)) + + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE)); + size_t const entropySpace = ZSTD_cwksp_alloc_size(ENTROPY_WORKSPACE_SIZE); + size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t)); + size_t const matchStateSize = ZSTD_sizeof_matchState(cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 0, /* forCCtx */ 1); + + size_t const ldmSpace = ZSTD_ldm_getTableSize(*ldmParams); + size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(*ldmParams, blockSize); + size_t const ldmSeqSpace = ldmParams->enableLdm == ZSTD_ps_enable ? + ZSTD_cwksp_aligned_alloc_size(maxNbLdmSeq * sizeof(rawSeq)) : 0; + + + size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + + ZSTD_cwksp_alloc_size(buffOutSize); + + size_t const cctxSpace = isStatic ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0; + + size_t const maxNbExternalSeq = ZSTD_sequenceBound(blockSize); + size_t const externalSeqSpace = useSequenceProducer + ? ZSTD_cwksp_aligned_alloc_size(maxNbExternalSeq * sizeof(ZSTD_Sequence)) + : 0; + + size_t const neededSpace = + cctxSpace + + entropySpace + + blockStateSpace + + ldmSpace + + ldmSeqSpace + + matchStateSize + + tokenSpace + + bufferSpace + + externalSeqSpace; + + DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace); + return neededSpace; } size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params) { - RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only."); - { ZSTD_compressionParameters const cParams = - ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0); - size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog); - U32 const divider = (cParams.minMatch==3) ? 3 : 4; - size_t const maxNbSeq = blockSize / divider; - size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize) - + ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef)) - + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE)); - size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE); - size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t)); - size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1); - - size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams); - size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq)); - - /* estimateCCtxSize is for one-shot compression. So no buffers should - * be needed. However, we still allocate two 0-sized buffers, which can - * take space under ASAN. */ - size_t const bufferSpace = ZSTD_cwksp_alloc_size(0) - + ZSTD_cwksp_alloc_size(0); - - size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)); + ZSTD_compressionParameters const cParams = + ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); + ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, + &cParams); - size_t const neededSpace = - cctxSpace + - entropySpace + - blockStateSpace + - ldmSpace + - ldmSeqSpace + - matchStateSize + - tokenSpace + - bufferSpace; - - DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace); - return neededSpace; - } + RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only."); + /* estimateCCtxSize is for one-shot compression. So no buffers should + * be needed. However, we still allocate two 0-sized buffers, which can + * take space under ASAN. */ + return ZSTD_estimateCCtxSize_usingCCtxParams_internal( + &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, 0, 0, ZSTD_CONTENTSIZE_UNKNOWN, params->useSequenceProducer, params->maxBlockSize); } size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams) { - ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams); - return ZSTD_estimateCCtxSize_usingCCtxParams(¶ms); + ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams); + if (ZSTD_rowMatchFinderSupported(cParams.strategy)) { + /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */ + size_t noRowCCtxSize; + size_t rowCCtxSize; + initialParams.useRowMatchFinder = ZSTD_ps_disable; + noRowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); + initialParams.useRowMatchFinder = ZSTD_ps_enable; + rowCCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); + return MAX(noRowCCtxSize, rowCCtxSize); + } else { + return ZSTD_estimateCCtxSize_usingCCtxParams(&initialParams); + } } static size_t ZSTD_estimateCCtxSize_internal(int compressionLevel) { - ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0); - return ZSTD_estimateCCtxSize_usingCParams(cParams); + int tier = 0; + size_t largestSize = 0; + static const unsigned long long srcSizeTiers[4] = {16 KB, 128 KB, 256 KB, ZSTD_CONTENTSIZE_UNKNOWN}; + for (; tier < 4; ++tier) { + /* Choose the set of cParams for a given level across all srcSizes that give the largest cctxSize */ + ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeTiers[tier], 0, ZSTD_cpm_noAttachDict); + largestSize = MAX(ZSTD_estimateCCtxSize_usingCParams(cParams), largestSize); + } + return largestSize; } size_t ZSTD_estimateCCtxSize(int compressionLevel) @@ -1181,6 +1717,7 @@ size_t ZSTD_estimateCCtxSize(int compressionLevel) int level; size_t memBudget = 0; for (level=MIN(compressionLevel, 1); level<=compressionLevel; level++) { + /* Ensure monotonically increasing memory usage as compression level increases */ size_t const newMB = ZSTD_estimateCCtxSize_internal(level); if (newMB > memBudget) memBudget = newMB; } @@ -1191,27 +1728,42 @@ size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params) { RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only."); { ZSTD_compressionParameters const cParams = - ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0); - size_t const CCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(params); - size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog); - size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize; - size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1; - size_t const streamingSize = ZSTD_cwksp_alloc_size(inBuffSize) - + ZSTD_cwksp_alloc_size(outBuffSize); - - return CCtxSize + streamingSize; + ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); + size_t const blockSize = MIN(ZSTD_resolveMaxBlockSize(params->maxBlockSize), (size_t)1 << cParams.windowLog); + size_t const inBuffSize = (params->inBufferMode == ZSTD_bm_buffered) + ? ((size_t)1 << cParams.windowLog) + blockSize + : 0; + size_t const outBuffSize = (params->outBufferMode == ZSTD_bm_buffered) + ? ZSTD_compressBound(blockSize) + 1 + : 0; + ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params->useRowMatchFinder, ¶ms->cParams); + + return ZSTD_estimateCCtxSize_usingCCtxParams_internal( + &cParams, ¶ms->ldmParams, 1, useRowMatchFinder, inBuffSize, outBuffSize, + ZSTD_CONTENTSIZE_UNKNOWN, params->useSequenceProducer, params->maxBlockSize); } } size_t ZSTD_estimateCStreamSize_usingCParams(ZSTD_compressionParameters cParams) { - ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams); - return ZSTD_estimateCStreamSize_usingCCtxParams(¶ms); + ZSTD_CCtx_params initialParams = ZSTD_makeCCtxParamsFromCParams(cParams); + if (ZSTD_rowMatchFinderSupported(cParams.strategy)) { + /* Pick bigger of not using and using row-based matchfinder for greedy and lazy strategies */ + size_t noRowCCtxSize; + size_t rowCCtxSize; + initialParams.useRowMatchFinder = ZSTD_ps_disable; + noRowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); + initialParams.useRowMatchFinder = ZSTD_ps_enable; + rowCCtxSize = ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); + return MAX(noRowCCtxSize, rowCCtxSize); + } else { + return ZSTD_estimateCStreamSize_usingCCtxParams(&initialParams); + } } static size_t ZSTD_estimateCStreamSize_internal(int compressionLevel) { - ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0); + ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_cpm_noAttachDict); return ZSTD_estimateCStreamSize_usingCParams(cParams); } @@ -1305,16 +1857,6 @@ static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms) } /** - * Indicates whether this compression proceeds directly from user-provided - * source buffer to user-provided destination buffer (ZSTDb_not_buffered), or - * whether the context needs to buffer the input/output (ZSTDb_buffered). - */ -typedef enum { - ZSTDb_not_buffered, - ZSTDb_buffered -} ZSTD_buffered_policy_e; - -/** * Controls, for this matchState reset, whether the tables need to be cleared / * prepared for the coming compression (ZSTDcrp_makeClean), or whether the * tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a @@ -1341,20 +1883,27 @@ typedef enum { ZSTD_resetTarget_CCtx } ZSTD_resetTarget_e; + static size_t ZSTD_reset_matchState(ZSTD_matchState_t* ms, ZSTD_cwksp* ws, const ZSTD_compressionParameters* cParams, + const ZSTD_paramSwitch_e useRowMatchFinder, const ZSTD_compResetPolicy_e crp, const ZSTD_indexResetPolicy_e forceResetIndex, const ZSTD_resetTarget_e forWho) { - size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog); + /* disable chain table allocation for fast or row-based strategies */ + size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, + ms->dedicatedDictSearch && (forWho == ZSTD_resetTarget_CDict)) + ? ((size_t)1 << cParams->chainLog) + : 0; size_t const hSize = ((size_t)1) << cParams->hashLog; U32 const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0; size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0; DEBUGLOG(4, "reset indices : %u", forceResetIndex == ZSTDirp_reset); + assert(useRowMatchFinder != ZSTD_ps_auto); if (forceResetIndex == ZSTDirp_reset) { ZSTD_window_init(&ms->window); ZSTD_cwksp_mark_tables_dirty(ws); @@ -1393,11 +1942,23 @@ ZSTD_reset_matchState(ZSTD_matchState_t* ms, ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t)); } + if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) { + { /* Row match finder needs an additional table of hashes ("tags") */ + size_t const tagTableSize = hSize*sizeof(U16); + ms->tagTable = (U16*)ZSTD_cwksp_reserve_aligned(ws, tagTableSize); + if (ms->tagTable) ZSTD_memset(ms->tagTable, 0, tagTableSize); + } + { /* Switch to 32-entry rows if searchLog is 5 (or more) */ + U32 const rowLog = BOUNDED(4, cParams->searchLog, 6); + assert(cParams->hashLog >= rowLog); + ms->rowHashLog = cParams->hashLog - rowLog; + } + } + ms->cParams = *cParams; RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation, "failed a workspace allocation in ZSTD_reset_matchState"); - return 0; } @@ -1414,75 +1975,87 @@ static int ZSTD_indexTooCloseToMax(ZSTD_window_t w) return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN); } +/** ZSTD_dictTooBig(): + * When dictionaries are larger than ZSTD_CHUNKSIZE_MAX they can't be loaded in + * one go generically. So we ensure that in that case we reset the tables to zero, + * so that we can load as much of the dictionary as possible. + */ +static int ZSTD_dictTooBig(size_t const loadedDictSize) +{ + return loadedDictSize > ZSTD_CHUNKSIZE_MAX; +} + /*! ZSTD_resetCCtx_internal() : - note : `params` are assumed fully validated at this stage */ + * @param loadedDictSize The size of the dictionary to be loaded + * into the context, if any. If no dictionary is used, or the + * dictionary is being attached / copied, then pass 0. + * note : `params` are assumed fully validated at this stage. + */ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, - ZSTD_CCtx_params params, + ZSTD_CCtx_params const* params, U64 const pledgedSrcSize, + size_t const loadedDictSize, ZSTD_compResetPolicy_e const crp, ZSTD_buffered_policy_e const zbuff) { ZSTD_cwksp* const ws = &zc->workspace; - DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u", - (U32)pledgedSrcSize, params.cParams.windowLog); - assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams))); + DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u, useRowMatchFinder=%d useBlockSplitter=%d", + (U32)pledgedSrcSize, params->cParams.windowLog, (int)params->useRowMatchFinder, (int)params->useBlockSplitter); + assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams))); zc->isFirstBlock = 1; - if (params.ldmParams.enableLdm) { + /* Set applied params early so we can modify them for LDM, + * and point params at the applied params. + */ + zc->appliedParams = *params; + params = &zc->appliedParams; + + assert(params->useRowMatchFinder != ZSTD_ps_auto); + assert(params->useBlockSplitter != ZSTD_ps_auto); + assert(params->ldmParams.enableLdm != ZSTD_ps_auto); + assert(params->maxBlockSize != 0); + if (params->ldmParams.enableLdm == ZSTD_ps_enable) { /* Adjust long distance matching parameters */ - ZSTD_ldm_adjustParameters(¶ms.ldmParams, ¶ms.cParams); - assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog); - assert(params.ldmParams.hashRateLog < 32); - zc->ldmState.hashPower = ZSTD_rollingHash_primePower(params.ldmParams.minMatchLength); - } - - { size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params.cParams.windowLog), pledgedSrcSize)); - size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize); - U32 const divider = (params.cParams.minMatch==3) ? 3 : 4; - size_t const maxNbSeq = blockSize / divider; - size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize) - + ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef)) - + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE)); - size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0; - size_t const buffInSize = (zbuff==ZSTDb_buffered) ? windowSize + blockSize : 0; - size_t const matchStateSize = ZSTD_sizeof_matchState(¶ms.cParams, /* forCCtx */ 1); - size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params.ldmParams, blockSize); - - ZSTD_indexResetPolicy_e needsIndexReset = zc->initialized ? ZSTDirp_continue : ZSTDirp_reset; - - if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) { - needsIndexReset = ZSTDirp_reset; - } + ZSTD_ldm_adjustParameters(&zc->appliedParams.ldmParams, ¶ms->cParams); + assert(params->ldmParams.hashLog >= params->ldmParams.bucketSizeLog); + assert(params->ldmParams.hashRateLog < 32); + } - if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0); + { size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params->cParams.windowLog), pledgedSrcSize)); + size_t const blockSize = MIN(params->maxBlockSize, windowSize); + size_t const maxNbSeq = ZSTD_maxNbSeq(blockSize, params->cParams.minMatch, params->useSequenceProducer); + size_t const buffOutSize = (zbuff == ZSTDb_buffered && params->outBufferMode == ZSTD_bm_buffered) + ? ZSTD_compressBound(blockSize) + 1 + : 0; + size_t const buffInSize = (zbuff == ZSTDb_buffered && params->inBufferMode == ZSTD_bm_buffered) + ? windowSize + blockSize + : 0; + size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize); + + int const indexTooClose = ZSTD_indexTooCloseToMax(zc->blockState.matchState.window); + int const dictTooBig = ZSTD_dictTooBig(loadedDictSize); + ZSTD_indexResetPolicy_e needsIndexReset = + (indexTooClose || dictTooBig || !zc->initialized) ? ZSTDirp_reset : ZSTDirp_continue; + + size_t const neededSpace = + ZSTD_estimateCCtxSize_usingCCtxParams_internal( + ¶ms->cParams, ¶ms->ldmParams, zc->staticSize != 0, params->useRowMatchFinder, + buffInSize, buffOutSize, pledgedSrcSize, params->useSequenceProducer, params->maxBlockSize); + int resizeWorkspace; + + FORWARD_IF_ERROR(neededSpace, "cctx size estimate failed!"); - /* Check if workspace is large enough, alloc a new one if needed */ - { size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0; - size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE); - size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t)); - size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + ZSTD_cwksp_alloc_size(buffOutSize); - size_t const ldmSpace = ZSTD_ldm_getTableSize(params.ldmParams); - size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq)); - - size_t const neededSpace = - cctxSpace + - entropySpace + - blockStateSpace + - ldmSpace + - ldmSeqSpace + - matchStateSize + - tokenSpace + - bufferSpace; + if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0); + { /* Check if workspace is large enough, alloc a new one if needed */ int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace; int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace); - - DEBUGLOG(4, "Need %zuKB workspace, including %zuKB for match state, and %zuKB for buffers", - neededSpace>>10, matchStateSize>>10, bufferSpace>>10); + resizeWorkspace = workspaceTooSmall || workspaceWasteful; + DEBUGLOG(4, "Need %zu B workspace", neededSpace); DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize); - if (workspaceTooSmall || workspaceWasteful) { + if (resizeWorkspace) { DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB", ZSTD_cwksp_sizeof(ws) >> 10, neededSpace >> 10); @@ -1503,15 +2076,15 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock"); zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t)); RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock"); - zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, HUF_WORKSPACE_SIZE); - RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate entropyWorkspace"); + zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, ENTROPY_WORKSPACE_SIZE); + RETURN_ERROR_IF(zc->entropyWorkspace == NULL, memory_allocation, "couldn't allocate entropyWorkspace"); } } ZSTD_cwksp_clear(ws); /* init params */ - zc->appliedParams = params; - zc->blockState.matchState.cParams = params.cParams; + zc->blockState.matchState.cParams = params->cParams; + zc->blockState.matchState.prefetchCDictTables = params->prefetchCDictTables == ZSTD_ps_enable; zc->pledgedSrcSizePlusOne = pledgedSrcSize+1; zc->consumedSrcSize = 0; zc->producedCSize = 0; @@ -1524,6 +2097,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, XXH64_reset(&zc->xxhState, 0); zc->stage = ZSTDcs_init; zc->dictID = 0; + zc->dictContentSize = 0; ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock); @@ -1534,19 +2108,20 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, zc->seqStore.maxNbLit = blockSize; /* buffers */ + zc->bufferedPolicy = zbuff; zc->inBuffSize = buffInSize; zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize); zc->outBuffSize = buffOutSize; zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize); /* ldm bucketOffsets table */ - if (params.ldmParams.enableLdm) { + if (params->ldmParams.enableLdm == ZSTD_ps_enable) { /* TODO: avoid memset? */ - size_t const ldmBucketSize = - ((size_t)1) << (params.ldmParams.hashLog - - params.ldmParams.bucketSizeLog); - zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, ldmBucketSize); - memset(zc->ldmState.bucketOffsets, 0, ldmBucketSize); + size_t const numBuckets = + ((size_t)1) << (params->ldmParams.hashLog - + params->ldmParams.bucketSizeLog); + zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets); + ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets); } /* sequences storage */ @@ -1560,26 +2135,36 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc, FORWARD_IF_ERROR(ZSTD_reset_matchState( &zc->blockState.matchState, ws, - ¶ms.cParams, + ¶ms->cParams, + params->useRowMatchFinder, crp, needsIndexReset, ZSTD_resetTarget_CCtx), ""); /* ldm hash table */ - if (params.ldmParams.enableLdm) { + if (params->ldmParams.enableLdm == ZSTD_ps_enable) { /* TODO: avoid memset? */ - size_t const ldmHSize = ((size_t)1) << params.ldmParams.hashLog; + size_t const ldmHSize = ((size_t)1) << params->ldmParams.hashLog; zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t)); - memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t)); + ZSTD_memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t)); zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq)); zc->maxNbLdmSequences = maxNbLdmSeq; ZSTD_window_init(&zc->ldmState.window); - ZSTD_window_clear(&zc->ldmState.window); zc->ldmState.loadedDictEnd = 0; } + /* reserve space for block-level external sequences */ + if (params->useSequenceProducer) { + size_t const maxNbExternalSeq = ZSTD_sequenceBound(blockSize); + zc->externalMatchCtx.seqBufferCapacity = maxNbExternalSeq; + zc->externalMatchCtx.seqBuffer = + (ZSTD_Sequence*)ZSTD_cwksp_reserve_aligned(ws, maxNbExternalSeq * sizeof(ZSTD_Sequence)); + } + DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws)); + assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace, resizeWorkspace)); + zc->initialized = 1; return 0; @@ -1618,12 +2203,14 @@ static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict, U64 pledgedSrcSize) { size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy]; - return ( pledgedSrcSize <= cutoff - || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN - || params->attachDictPref == ZSTD_dictForceAttach ) - && params->attachDictPref != ZSTD_dictForceCopy - && !params->forceWindow; /* dictMatchState isn't correctly - * handled in _enforceMaxDist */ + int const dedicatedDictSearch = cdict->matchState.dedicatedDictSearch; + return dedicatedDictSearch + || ( ( pledgedSrcSize <= cutoff + || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN + || params->attachDictPref == ZSTD_dictForceAttach ) + && params->attachDictPref != ZSTD_dictForceCopy + && !params->forceWindow ); /* dictMatchState isn't correctly + * handled in _enforceMaxDist */ } static size_t @@ -1633,17 +2220,29 @@ ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx, U64 pledgedSrcSize, ZSTD_buffered_policy_e zbuff) { - { const ZSTD_compressionParameters* const cdict_cParams = &cdict->matchState.cParams; + DEBUGLOG(4, "ZSTD_resetCCtx_byAttachingCDict() pledgedSrcSize=%llu", + (unsigned long long)pledgedSrcSize); + { + ZSTD_compressionParameters adjusted_cdict_cParams = cdict->matchState.cParams; unsigned const windowLog = params.cParams.windowLog; assert(windowLog != 0); /* Resize working context table params for input only, since the dict * has its own tables. */ - /* pledgeSrcSize == 0 means 0! */ - params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0); + /* pledgedSrcSize == 0 means 0! */ + + if (cdict->matchState.dedicatedDictSearch) { + ZSTD_dedicatedDictSearch_revertCParams(&adjusted_cdict_cParams); + } + + params.cParams = ZSTD_adjustCParams_internal(adjusted_cdict_cParams, pledgedSrcSize, + cdict->dictContentSize, ZSTD_cpm_attachDict, + params.useRowMatchFinder); params.cParams.windowLog = windowLog; - FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize, + params.useRowMatchFinder = cdict->useRowMatchFinder; /* cdict overrides */ + FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize, + /* loadedDictSize */ 0, ZSTDcrp_makeClean, zbuff), ""); - assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy); + assert(cctx->appliedParams.cParams.strategy == adjusted_cdict_cParams.strategy); } { const U32 cdictEnd = (U32)( cdict->matchState.window.nextSrc @@ -1668,13 +2267,30 @@ ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx, } } cctx->dictID = cdict->dictID; + cctx->dictContentSize = cdict->dictContentSize; /* copy block state */ - memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); + ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); return 0; } +static void ZSTD_copyCDictTableIntoCCtx(U32* dst, U32 const* src, size_t tableSize, + ZSTD_compressionParameters const* cParams) { + if (ZSTD_CDictIndicesAreTagged(cParams)){ + /* Remove tags from the CDict table if they are present. + * See docs on "short cache" in zstd_compress_internal.h for context. */ + size_t i; + for (i = 0; i < tableSize; i++) { + U32 const taggedIndex = src[i]; + U32 const index = taggedIndex >> ZSTD_SHORT_CACHE_TAG_BITS; + dst[i] = index; + } + } else { + ZSTD_memcpy(dst, src, tableSize * sizeof(U32)); + } +} + static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict, ZSTD_CCtx_params params, @@ -1683,14 +2299,18 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx, { const ZSTD_compressionParameters *cdict_cParams = &cdict->matchState.cParams; - DEBUGLOG(4, "copying dictionary into context"); + assert(!cdict->matchState.dedicatedDictSearch); + DEBUGLOG(4, "ZSTD_resetCCtx_byCopyingCDict() pledgedSrcSize=%llu", + (unsigned long long)pledgedSrcSize); { unsigned const windowLog = params.cParams.windowLog; assert(windowLog != 0); /* Copy only compression parameters related to tables. */ params.cParams = *cdict_cParams; params.cParams.windowLog = windowLog; - FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize, + params.useRowMatchFinder = cdict->useRowMatchFinder; + FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, ¶ms, pledgedSrcSize, + /* loadedDictSize */ 0, ZSTDcrp_leaveDirty, zbuff), ""); assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy); assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog); @@ -1698,24 +2318,38 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx, } ZSTD_cwksp_mark_tables_dirty(&cctx->workspace); + assert(params.useRowMatchFinder != ZSTD_ps_auto); /* copy tables */ - { size_t const chainSize = (cdict_cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cdict_cParams->chainLog); + { size_t const chainSize = ZSTD_allocateChainTable(cdict_cParams->strategy, cdict->useRowMatchFinder, 0 /* DDS guaranteed disabled */) + ? ((size_t)1 << cdict_cParams->chainLog) + : 0; size_t const hSize = (size_t)1 << cdict_cParams->hashLog; - memcpy(cctx->blockState.matchState.hashTable, - cdict->matchState.hashTable, - hSize * sizeof(U32)); - memcpy(cctx->blockState.matchState.chainTable, - cdict->matchState.chainTable, - chainSize * sizeof(U32)); + ZSTD_copyCDictTableIntoCCtx(cctx->blockState.matchState.hashTable, + cdict->matchState.hashTable, + hSize, cdict_cParams); + + /* Do not copy cdict's chainTable if cctx has parameters such that it would not use chainTable */ + if (ZSTD_allocateChainTable(cctx->appliedParams.cParams.strategy, cctx->appliedParams.useRowMatchFinder, 0 /* forDDSDict */)) { + ZSTD_copyCDictTableIntoCCtx(cctx->blockState.matchState.chainTable, + cdict->matchState.chainTable, + chainSize, cdict_cParams); + } + /* copy tag table */ + if (ZSTD_rowMatchFinderUsed(cdict_cParams->strategy, cdict->useRowMatchFinder)) { + size_t const tagTableSize = hSize*sizeof(U16); + ZSTD_memcpy(cctx->blockState.matchState.tagTable, + cdict->matchState.tagTable, + tagTableSize); + } } /* Zero the hashTable3, since the cdict never fills it */ { int const h3log = cctx->blockState.matchState.hashLog3; size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0; assert(cdict->matchState.hashLog3 == 0); - memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32)); + ZSTD_memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32)); } ZSTD_cwksp_mark_tables_clean(&cctx->workspace); @@ -1729,9 +2363,10 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx, } cctx->dictID = cdict->dictID; + cctx->dictContentSize = cdict->dictContentSize; /* copy block state */ - memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); + ZSTD_memcpy(cctx->blockState.prevCBlock, &cdict->cBlockState, sizeof(cdict->cBlockState)); return 0; } @@ -1771,16 +2406,23 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, U64 pledgedSrcSize, ZSTD_buffered_policy_e zbuff) { - DEBUGLOG(5, "ZSTD_copyCCtx_internal"); RETURN_ERROR_IF(srcCCtx->stage!=ZSTDcs_init, stage_wrong, "Can't copy a ctx that's not in init stage."); - - memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem)); + DEBUGLOG(5, "ZSTD_copyCCtx_internal"); + ZSTD_memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem)); { ZSTD_CCtx_params params = dstCCtx->requestedParams; /* Copy only compression parameters related to tables. */ params.cParams = srcCCtx->appliedParams.cParams; + assert(srcCCtx->appliedParams.useRowMatchFinder != ZSTD_ps_auto); + assert(srcCCtx->appliedParams.useBlockSplitter != ZSTD_ps_auto); + assert(srcCCtx->appliedParams.ldmParams.enableLdm != ZSTD_ps_auto); + params.useRowMatchFinder = srcCCtx->appliedParams.useRowMatchFinder; + params.useBlockSplitter = srcCCtx->appliedParams.useBlockSplitter; + params.ldmParams = srcCCtx->appliedParams.ldmParams; params.fParams = fParams; - ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize, + params.maxBlockSize = srcCCtx->appliedParams.maxBlockSize; + ZSTD_resetCCtx_internal(dstCCtx, ¶ms, pledgedSrcSize, + /* loadedDictSize */ 0, ZSTDcrp_leaveDirty, zbuff); assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog); assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy); @@ -1792,18 +2434,22 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace); /* copy tables */ - { size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog); + { size_t const chainSize = ZSTD_allocateChainTable(srcCCtx->appliedParams.cParams.strategy, + srcCCtx->appliedParams.useRowMatchFinder, + 0 /* forDDSDict */) + ? ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog) + : 0; size_t const hSize = (size_t)1 << srcCCtx->appliedParams.cParams.hashLog; int const h3log = srcCCtx->blockState.matchState.hashLog3; size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0; - memcpy(dstCCtx->blockState.matchState.hashTable, + ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable, srcCCtx->blockState.matchState.hashTable, hSize * sizeof(U32)); - memcpy(dstCCtx->blockState.matchState.chainTable, + ZSTD_memcpy(dstCCtx->blockState.matchState.chainTable, srcCCtx->blockState.matchState.chainTable, chainSize * sizeof(U32)); - memcpy(dstCCtx->blockState.matchState.hashTable3, + ZSTD_memcpy(dstCCtx->blockState.matchState.hashTable3, srcCCtx->blockState.matchState.hashTable3, h3Size * sizeof(U32)); } @@ -1819,9 +2465,10 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, dstMatchState->loadedDictEnd= srcMatchState->loadedDictEnd; } dstCCtx->dictID = srcCCtx->dictID; + dstCCtx->dictContentSize = srcCCtx->dictContentSize; /* copy block state */ - memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock)); + ZSTD_memcpy(dstCCtx->blockState.prevCBlock, srcCCtx->blockState.prevCBlock, sizeof(*srcCCtx->blockState.prevCBlock)); return 0; } @@ -1834,7 +2481,7 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize) { ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; - ZSTD_buffered_policy_e const zbuff = (ZSTD_buffered_policy_e)(srcCCtx->inBuffSize>0); + ZSTD_buffered_policy_e const zbuff = srcCCtx->bufferedPolicy; ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1); if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN; fParams.contentSizeFlag = (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN); @@ -1858,10 +2505,12 @@ ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerVa int const nbRows = (int)size / ZSTD_ROWSIZE; int cellNb = 0; int rowNb; + /* Protect special index values < ZSTD_WINDOW_START_INDEX. */ + U32 const reducerThreshold = reducerValue + ZSTD_WINDOW_START_INDEX; assert((size & (ZSTD_ROWSIZE-1)) == 0); /* multiple of ZSTD_ROWSIZE */ assert(size < (1U<<31)); /* can be casted to int */ -#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE) +#if ZSTD_MEMORY_SANITIZER && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE) /* To validate that the table re-use logic is sound, and that we don't * access table space that we haven't cleaned, we re-"poison" the table * space every time we mark it dirty. @@ -1877,12 +2526,17 @@ ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerVa for (rowNb=0 ; rowNb < nbRows ; rowNb++) { int column; for (column=0; column<ZSTD_ROWSIZE; column++) { - if (preserveMark) { - U32 const adder = (table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) ? reducerValue : 0; - table[cellNb] += adder; + U32 newVal; + if (preserveMark && table[cellNb] == ZSTD_DUBT_UNSORTED_MARK) { + /* This write is pointless, but is required(?) for the compiler + * to auto-vectorize the loop. */ + newVal = ZSTD_DUBT_UNSORTED_MARK; + } else if (table[cellNb] < reducerThreshold) { + newVal = 0; + } else { + newVal = table[cellNb] - reducerValue; } - if (table[cellNb] < reducerValue) table[cellNb] = 0; - else table[cellNb] -= reducerValue; + table[cellNb] = newVal; cellNb++; } } } @@ -1905,7 +2559,7 @@ static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* par ZSTD_reduceTable(ms->hashTable, hSize, reducerValue); } - if (params->cParams.strategy != ZSTD_fast) { + if (ZSTD_allocateChainTable(params->cParams.strategy, params->useRowMatchFinder, (U32)ms->dedicatedDictSearch)) { U32 const chainSize = (U32)1 << params->cParams.chainLog; if (params->cParams.strategy == ZSTD_btlazy2) ZSTD_reduceTable_btlazy2(ms->chainTable, chainSize, reducerValue); @@ -1926,7 +2580,7 @@ static void ZSTD_reduceIndex (ZSTD_matchState_t* ms, ZSTD_CCtx_params const* par /* See doc/zstd_compression_format.md for detailed format description */ -void ZSTD_seqToCodes(const seqStore_t* seqStorePtr) +int ZSTD_seqToCodes(const seqStore_t* seqStorePtr) { const seqDef* const sequences = seqStorePtr->sequencesStart; BYTE* const llCodeTable = seqStorePtr->llCode; @@ -1934,18 +2588,24 @@ void ZSTD_seqToCodes(const seqStore_t* seqStorePtr) BYTE* const mlCodeTable = seqStorePtr->mlCode; U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart); U32 u; + int longOffsets = 0; assert(nbSeq <= seqStorePtr->maxNbSeq); for (u=0; u<nbSeq; u++) { U32 const llv = sequences[u].litLength; - U32 const mlv = sequences[u].matchLength; + U32 const ofCode = ZSTD_highbit32(sequences[u].offBase); + U32 const mlv = sequences[u].mlBase; llCodeTable[u] = (BYTE)ZSTD_LLcode(llv); - ofCodeTable[u] = (BYTE)ZSTD_highbit32(sequences[u].offset); + ofCodeTable[u] = (BYTE)ofCode; mlCodeTable[u] = (BYTE)ZSTD_MLcode(mlv); + assert(!(MEM_64bits() && ofCode >= STREAM_ACCUMULATOR_MIN)); + if (MEM_32bits() && ofCode >= STREAM_ACCUMULATOR_MIN) + longOffsets = 1; } - if (seqStorePtr->longLengthID==1) + if (seqStorePtr->longLengthType==ZSTD_llt_literalLength) llCodeTable[seqStorePtr->longLengthPos] = MaxLL; - if (seqStorePtr->longLengthID==2) + if (seqStorePtr->longLengthType==ZSTD_llt_matchLength) mlCodeTable[seqStorePtr->longLengthPos] = MaxML; + return longOffsets; } /* ZSTD_useTargetCBlockSize(): @@ -1958,163 +2618,250 @@ static int ZSTD_useTargetCBlockSize(const ZSTD_CCtx_params* cctxParams) return (cctxParams->targetCBlockSize != 0); } -/* ZSTD_compressSequences_internal(): - * actually compresses both literals and sequences */ -MEM_STATIC size_t -ZSTD_compressSequences_internal(seqStore_t* seqStorePtr, - const ZSTD_entropyCTables_t* prevEntropy, - ZSTD_entropyCTables_t* nextEntropy, - const ZSTD_CCtx_params* cctxParams, - void* dst, size_t dstCapacity, - void* entropyWorkspace, size_t entropyWkspSize, - const int bmi2) +/* ZSTD_blockSplitterEnabled(): + * Returns if block splitting param is being used + * If used, compression will do best effort to split a block in order to improve compression ratio. + * At the time this function is called, the parameter must be finalized. + * Returns 1 if true, 0 otherwise. */ +static int ZSTD_blockSplitterEnabled(ZSTD_CCtx_params* cctxParams) { - const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN; - ZSTD_strategy const strategy = cctxParams->cParams.strategy; - unsigned count[MaxSeq+1]; - FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable; - FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable; - FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable; - U32 LLtype, Offtype, MLtype; /* compressed, raw or rle */ - const seqDef* const sequences = seqStorePtr->sequencesStart; + DEBUGLOG(5, "ZSTD_blockSplitterEnabled (useBlockSplitter=%d)", cctxParams->useBlockSplitter); + assert(cctxParams->useBlockSplitter != ZSTD_ps_auto); + return (cctxParams->useBlockSplitter == ZSTD_ps_enable); +} + +/* Type returned by ZSTD_buildSequencesStatistics containing finalized symbol encoding types + * and size of the sequences statistics + */ +typedef struct { + U32 LLtype; + U32 Offtype; + U32 MLtype; + size_t size; + size_t lastCountSize; /* Accounts for bug in 1.3.4. More detail in ZSTD_entropyCompressSeqStore_internal() */ + int longOffsets; +} ZSTD_symbolEncodingTypeStats_t; + +/* ZSTD_buildSequencesStatistics(): + * Returns a ZSTD_symbolEncodingTypeStats_t, or a zstd error code in the `size` field. + * Modifies `nextEntropy` to have the appropriate values as a side effect. + * nbSeq must be greater than 0. + * + * entropyWkspSize must be of size at least ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1)*sizeof(U32) + */ +static ZSTD_symbolEncodingTypeStats_t +ZSTD_buildSequencesStatistics( + const seqStore_t* seqStorePtr, size_t nbSeq, + const ZSTD_fseCTables_t* prevEntropy, ZSTD_fseCTables_t* nextEntropy, + BYTE* dst, const BYTE* const dstEnd, + ZSTD_strategy strategy, unsigned* countWorkspace, + void* entropyWorkspace, size_t entropyWkspSize) +{ + BYTE* const ostart = dst; + const BYTE* const oend = dstEnd; + BYTE* op = ostart; + FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable; + FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable; + FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable; const BYTE* const ofCodeTable = seqStorePtr->ofCode; const BYTE* const llCodeTable = seqStorePtr->llCode; const BYTE* const mlCodeTable = seqStorePtr->mlCode; - BYTE* const ostart = (BYTE*)dst; - BYTE* const oend = ostart + dstCapacity; - BYTE* op = ostart; - size_t const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); - BYTE* seqHead; - BYTE* lastNCount = NULL; - - DEBUGLOG(5, "ZSTD_compressSequences_internal (nbSeq=%zu)", nbSeq); - ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog))); - - /* Compress literals */ - { const BYTE* const literals = seqStorePtr->litStart; - size_t const litSize = (size_t)(seqStorePtr->lit - literals); - size_t const cSize = ZSTD_compressLiterals( - &prevEntropy->huf, &nextEntropy->huf, - cctxParams->cParams.strategy, - ZSTD_disableLiteralsCompression(cctxParams), - op, dstCapacity, - literals, litSize, - entropyWorkspace, entropyWkspSize, - bmi2); - FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed"); - assert(cSize <= dstCapacity); - op += cSize; - } - - /* Sequences Header */ - RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/, - dstSize_tooSmall, "Can't fit seq hdr in output buf!"); - if (nbSeq < 128) { - *op++ = (BYTE)nbSeq; - } else if (nbSeq < LONGNBSEQ) { - op[0] = (BYTE)((nbSeq>>8) + 0x80); - op[1] = (BYTE)nbSeq; - op+=2; - } else { - op[0]=0xFF; - MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)); - op+=3; - } - assert(op <= oend); - if (nbSeq==0) { - /* Copy the old tables over as if we repeated them */ - memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse)); - return (size_t)(op - ostart); - } - - /* seqHead : flags for FSE encoding type */ - seqHead = op++; - assert(op <= oend); + ZSTD_symbolEncodingTypeStats_t stats; + stats.lastCountSize = 0; /* convert length/distances into codes */ - ZSTD_seqToCodes(seqStorePtr); + stats.longOffsets = ZSTD_seqToCodes(seqStorePtr); + assert(op <= oend); + assert(nbSeq != 0); /* ZSTD_selectEncodingType() divides by nbSeq */ /* build CTable for Literal Lengths */ { unsigned max = MaxLL; - size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ + size_t const mostFrequent = HIST_countFast_wksp(countWorkspace, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ DEBUGLOG(5, "Building LL table"); - nextEntropy->fse.litlength_repeatMode = prevEntropy->fse.litlength_repeatMode; - LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode, - count, max, mostFrequent, nbSeq, - LLFSELog, prevEntropy->fse.litlengthCTable, + nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode; + stats.LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode, + countWorkspace, max, mostFrequent, nbSeq, + LLFSELog, prevEntropy->litlengthCTable, LL_defaultNorm, LL_defaultNormLog, ZSTD_defaultAllowed, strategy); assert(set_basic < set_compressed && set_rle < set_compressed); - assert(!(LLtype < set_compressed && nextEntropy->fse.litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ + assert(!(stats.LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ { size_t const countSize = ZSTD_buildCTable( op, (size_t)(oend - op), - CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype, - count, max, llCodeTable, nbSeq, + CTable_LitLength, LLFSELog, (symbolEncodingType_e)stats.LLtype, + countWorkspace, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL, - prevEntropy->fse.litlengthCTable, - sizeof(prevEntropy->fse.litlengthCTable), + prevEntropy->litlengthCTable, + sizeof(prevEntropy->litlengthCTable), entropyWorkspace, entropyWkspSize); - FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed"); - if (LLtype == set_compressed) - lastNCount = op; + if (ZSTD_isError(countSize)) { + DEBUGLOG(3, "ZSTD_buildCTable for LitLens failed"); + stats.size = countSize; + return stats; + } + if (stats.LLtype == set_compressed) + stats.lastCountSize = countSize; op += countSize; assert(op <= oend); } } /* build CTable for Offsets */ { unsigned max = MaxOff; size_t const mostFrequent = HIST_countFast_wksp( - count, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ + countWorkspace, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */ ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed; DEBUGLOG(5, "Building OF table"); - nextEntropy->fse.offcode_repeatMode = prevEntropy->fse.offcode_repeatMode; - Offtype = ZSTD_selectEncodingType(&nextEntropy->fse.offcode_repeatMode, - count, max, mostFrequent, nbSeq, - OffFSELog, prevEntropy->fse.offcodeCTable, + nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode; + stats.Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode, + countWorkspace, max, mostFrequent, nbSeq, + OffFSELog, prevEntropy->offcodeCTable, OF_defaultNorm, OF_defaultNormLog, defaultPolicy, strategy); - assert(!(Offtype < set_compressed && nextEntropy->fse.offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */ + assert(!(stats.Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */ { size_t const countSize = ZSTD_buildCTable( op, (size_t)(oend - op), - CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype, - count, max, ofCodeTable, nbSeq, + CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)stats.Offtype, + countWorkspace, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff, - prevEntropy->fse.offcodeCTable, - sizeof(prevEntropy->fse.offcodeCTable), + prevEntropy->offcodeCTable, + sizeof(prevEntropy->offcodeCTable), entropyWorkspace, entropyWkspSize); - FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed"); - if (Offtype == set_compressed) - lastNCount = op; + if (ZSTD_isError(countSize)) { + DEBUGLOG(3, "ZSTD_buildCTable for Offsets failed"); + stats.size = countSize; + return stats; + } + if (stats.Offtype == set_compressed) + stats.lastCountSize = countSize; op += countSize; assert(op <= oend); } } /* build CTable for MatchLengths */ { unsigned max = MaxML; size_t const mostFrequent = HIST_countFast_wksp( - count, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ + countWorkspace, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize); /* can't fail */ DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op)); - nextEntropy->fse.matchlength_repeatMode = prevEntropy->fse.matchlength_repeatMode; - MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode, - count, max, mostFrequent, nbSeq, - MLFSELog, prevEntropy->fse.matchlengthCTable, + nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode; + stats.MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode, + countWorkspace, max, mostFrequent, nbSeq, + MLFSELog, prevEntropy->matchlengthCTable, ML_defaultNorm, ML_defaultNormLog, ZSTD_defaultAllowed, strategy); - assert(!(MLtype < set_compressed && nextEntropy->fse.matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ + assert(!(stats.MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */ { size_t const countSize = ZSTD_buildCTable( op, (size_t)(oend - op), - CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype, - count, max, mlCodeTable, nbSeq, + CTable_MatchLength, MLFSELog, (symbolEncodingType_e)stats.MLtype, + countWorkspace, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML, - prevEntropy->fse.matchlengthCTable, - sizeof(prevEntropy->fse.matchlengthCTable), + prevEntropy->matchlengthCTable, + sizeof(prevEntropy->matchlengthCTable), entropyWorkspace, entropyWkspSize); - FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed"); - if (MLtype == set_compressed) - lastNCount = op; + if (ZSTD_isError(countSize)) { + DEBUGLOG(3, "ZSTD_buildCTable for MatchLengths failed"); + stats.size = countSize; + return stats; + } + if (stats.MLtype == set_compressed) + stats.lastCountSize = countSize; op += countSize; assert(op <= oend); } } + stats.size = (size_t)(op-ostart); + return stats; +} + +/* ZSTD_entropyCompressSeqStore_internal(): + * compresses both literals and sequences + * Returns compressed size of block, or a zstd error. + */ +#define SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO 20 +MEM_STATIC size_t +ZSTD_entropyCompressSeqStore_internal( + const seqStore_t* seqStorePtr, + const ZSTD_entropyCTables_t* prevEntropy, + ZSTD_entropyCTables_t* nextEntropy, + const ZSTD_CCtx_params* cctxParams, + void* dst, size_t dstCapacity, + void* entropyWorkspace, size_t entropyWkspSize, + const int bmi2) +{ + ZSTD_strategy const strategy = cctxParams->cParams.strategy; + unsigned* count = (unsigned*)entropyWorkspace; + FSE_CTable* CTable_LitLength = nextEntropy->fse.litlengthCTable; + FSE_CTable* CTable_OffsetBits = nextEntropy->fse.offcodeCTable; + FSE_CTable* CTable_MatchLength = nextEntropy->fse.matchlengthCTable; + const seqDef* const sequences = seqStorePtr->sequencesStart; + const size_t nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); + const BYTE* const ofCodeTable = seqStorePtr->ofCode; + const BYTE* const llCodeTable = seqStorePtr->llCode; + const BYTE* const mlCodeTable = seqStorePtr->mlCode; + BYTE* const ostart = (BYTE*)dst; + BYTE* const oend = ostart + dstCapacity; + BYTE* op = ostart; + size_t lastCountSize; + int longOffsets = 0; + + entropyWorkspace = count + (MaxSeq + 1); + entropyWkspSize -= (MaxSeq + 1) * sizeof(*count); + + DEBUGLOG(5, "ZSTD_entropyCompressSeqStore_internal (nbSeq=%zu, dstCapacity=%zu)", nbSeq, dstCapacity); + ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog))); + assert(entropyWkspSize >= HUF_WORKSPACE_SIZE); + + /* Compress literals */ + { const BYTE* const literals = seqStorePtr->litStart; + size_t const numSequences = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); + size_t const numLiterals = (size_t)(seqStorePtr->lit - seqStorePtr->litStart); + /* Base suspicion of uncompressibility on ratio of literals to sequences */ + unsigned const suspectUncompressible = (numSequences == 0) || (numLiterals / numSequences >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO); + size_t const litSize = (size_t)(seqStorePtr->lit - literals); + + size_t const cSize = ZSTD_compressLiterals( + op, dstCapacity, + literals, litSize, + entropyWorkspace, entropyWkspSize, + &prevEntropy->huf, &nextEntropy->huf, + cctxParams->cParams.strategy, + ZSTD_literalsCompressionIsDisabled(cctxParams), + suspectUncompressible, bmi2); + FORWARD_IF_ERROR(cSize, "ZSTD_compressLiterals failed"); + assert(cSize <= dstCapacity); + op += cSize; + } - *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2)); + /* Sequences Header */ + RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/, + dstSize_tooSmall, "Can't fit seq hdr in output buf!"); + if (nbSeq < 128) { + *op++ = (BYTE)nbSeq; + } else if (nbSeq < LONGNBSEQ) { + op[0] = (BYTE)((nbSeq>>8) + 0x80); + op[1] = (BYTE)nbSeq; + op+=2; + } else { + op[0]=0xFF; + MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)); + op+=3; + } + assert(op <= oend); + if (nbSeq==0) { + /* Copy the old tables over as if we repeated them */ + ZSTD_memcpy(&nextEntropy->fse, &prevEntropy->fse, sizeof(prevEntropy->fse)); + return (size_t)(op - ostart); + } + { BYTE* const seqHead = op++; + /* build stats for sequences */ + const ZSTD_symbolEncodingTypeStats_t stats = + ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq, + &prevEntropy->fse, &nextEntropy->fse, + op, oend, + strategy, count, + entropyWorkspace, entropyWkspSize); + FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!"); + *seqHead = (BYTE)((stats.LLtype<<6) + (stats.Offtype<<4) + (stats.MLtype<<2)); + lastCountSize = stats.lastCountSize; + op += stats.size; + longOffsets = stats.longOffsets; + } { size_t const bitstreamSize = ZSTD_encodeSequences( op, (size_t)(oend - op), @@ -2134,9 +2881,9 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr, * In this exceedingly rare case, we will simply emit an uncompressed * block, since it isn't worth optimizing. */ - if (lastNCount && (op - lastNCount) < 4) { - /* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */ - assert(op - lastNCount == 3); + if (lastCountSize && (lastCountSize + bitstreamSize) < 4) { + /* lastCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */ + assert(lastCountSize + bitstreamSize == 3); DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by " "emitting an uncompressed block."); return 0; @@ -2148,16 +2895,17 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr, } MEM_STATIC size_t -ZSTD_compressSequences(seqStore_t* seqStorePtr, - const ZSTD_entropyCTables_t* prevEntropy, - ZSTD_entropyCTables_t* nextEntropy, - const ZSTD_CCtx_params* cctxParams, - void* dst, size_t dstCapacity, - size_t srcSize, - void* entropyWorkspace, size_t entropyWkspSize, - int bmi2) -{ - size_t const cSize = ZSTD_compressSequences_internal( +ZSTD_entropyCompressSeqStore( + const seqStore_t* seqStorePtr, + const ZSTD_entropyCTables_t* prevEntropy, + ZSTD_entropyCTables_t* nextEntropy, + const ZSTD_CCtx_params* cctxParams, + void* dst, size_t dstCapacity, + size_t srcSize, + void* entropyWorkspace, size_t entropyWkspSize, + int bmi2) +{ + size_t const cSize = ZSTD_entropyCompressSeqStore_internal( seqStorePtr, prevEntropy, nextEntropy, cctxParams, dst, dstCapacity, entropyWorkspace, entropyWkspSize, bmi2); @@ -2165,24 +2913,30 @@ ZSTD_compressSequences(seqStore_t* seqStorePtr, /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block. * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block. */ - if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity)) + if ((cSize == ERROR(dstSize_tooSmall)) & (srcSize <= dstCapacity)) { + DEBUGLOG(4, "not enough dstCapacity (%zu) for ZSTD_entropyCompressSeqStore_internal()=> do not compress block", dstCapacity); return 0; /* block not compressed */ - FORWARD_IF_ERROR(cSize, "ZSTD_compressSequences_internal failed"); + } + FORWARD_IF_ERROR(cSize, "ZSTD_entropyCompressSeqStore_internal failed"); /* Check compressibility */ { size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, cctxParams->cParams.strategy); if (cSize >= maxCSize) return 0; /* block not compressed */ } - + DEBUGLOG(5, "ZSTD_entropyCompressSeqStore() cSize: %zu", cSize); + /* libzstd decoder before > v1.5.4 is not compatible with compressed blocks of size ZSTD_BLOCKSIZE_MAX exactly. + * This restriction is indirectly already fulfilled by respecting ZSTD_minGain() condition above. + */ + assert(cSize < ZSTD_BLOCKSIZE_MAX); return cSize; } /* ZSTD_selectBlockCompressor() : * Not static, but internal use only (used by long distance matcher) * assumption : strat is a valid strategy */ -ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMode_e dictMode) +ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_paramSwitch_e useRowMatchFinder, ZSTD_dictMode_e dictMode) { - static const ZSTD_blockCompressor blockCompressor[3][ZSTD_STRATEGY_MAX+1] = { + static const ZSTD_blockCompressor blockCompressor[4][ZSTD_STRATEGY_MAX+1] = { { ZSTD_compressBlock_fast /* default for 0 */, ZSTD_compressBlock_fast, ZSTD_compressBlock_doubleFast, @@ -2212,13 +2966,44 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo ZSTD_compressBlock_btlazy2_dictMatchState, ZSTD_compressBlock_btopt_dictMatchState, ZSTD_compressBlock_btultra_dictMatchState, - ZSTD_compressBlock_btultra_dictMatchState } + ZSTD_compressBlock_btultra_dictMatchState }, + { NULL /* default for 0 */, + NULL, + NULL, + ZSTD_compressBlock_greedy_dedicatedDictSearch, + ZSTD_compressBlock_lazy_dedicatedDictSearch, + ZSTD_compressBlock_lazy2_dedicatedDictSearch, + NULL, + NULL, + NULL, + NULL } }; ZSTD_blockCompressor selectedCompressor; ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1); assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat)); - selectedCompressor = blockCompressor[(int)dictMode][(int)strat]; + DEBUGLOG(4, "Selected block compressor: dictMode=%d strat=%d rowMatchfinder=%d", (int)dictMode, (int)strat, (int)useRowMatchFinder); + if (ZSTD_rowMatchFinderUsed(strat, useRowMatchFinder)) { + static const ZSTD_blockCompressor rowBasedBlockCompressors[4][3] = { + { ZSTD_compressBlock_greedy_row, + ZSTD_compressBlock_lazy_row, + ZSTD_compressBlock_lazy2_row }, + { ZSTD_compressBlock_greedy_extDict_row, + ZSTD_compressBlock_lazy_extDict_row, + ZSTD_compressBlock_lazy2_extDict_row }, + { ZSTD_compressBlock_greedy_dictMatchState_row, + ZSTD_compressBlock_lazy_dictMatchState_row, + ZSTD_compressBlock_lazy2_dictMatchState_row }, + { ZSTD_compressBlock_greedy_dedicatedDictSearch_row, + ZSTD_compressBlock_lazy_dedicatedDictSearch_row, + ZSTD_compressBlock_lazy2_dedicatedDictSearch_row } + }; + DEBUGLOG(4, "Selecting a row-based matchfinder"); + assert(useRowMatchFinder != ZSTD_ps_auto); + selectedCompressor = rowBasedBlockCompressors[(int)dictMode][(int)strat - (int)ZSTD_greedy]; + } else { + selectedCompressor = blockCompressor[(int)dictMode][(int)strat]; + } assert(selectedCompressor != NULL); return selectedCompressor; } @@ -2226,7 +3011,7 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr, const BYTE* anchor, size_t lastLLSize) { - memcpy(seqStorePtr->lit, anchor, lastLLSize); + ZSTD_memcpy(seqStorePtr->lit, anchor, lastLLSize); seqStorePtr->lit += lastLLSize; } @@ -2234,7 +3019,73 @@ void ZSTD_resetSeqStore(seqStore_t* ssPtr) { ssPtr->lit = ssPtr->litStart; ssPtr->sequences = ssPtr->sequencesStart; - ssPtr->longLengthID = 0; + ssPtr->longLengthType = ZSTD_llt_none; +} + +/* ZSTD_postProcessSequenceProducerResult() : + * Validates and post-processes sequences obtained through the external matchfinder API: + * - Checks whether nbExternalSeqs represents an error condition. + * - Appends a block delimiter to outSeqs if one is not already present. + * See zstd.h for context regarding block delimiters. + * Returns the number of sequences after post-processing, or an error code. */ +static size_t ZSTD_postProcessSequenceProducerResult( + ZSTD_Sequence* outSeqs, size_t nbExternalSeqs, size_t outSeqsCapacity, size_t srcSize +) { + RETURN_ERROR_IF( + nbExternalSeqs > outSeqsCapacity, + sequenceProducer_failed, + "External sequence producer returned error code %lu", + (unsigned long)nbExternalSeqs + ); + + RETURN_ERROR_IF( + nbExternalSeqs == 0 && srcSize > 0, + sequenceProducer_failed, + "Got zero sequences from external sequence producer for a non-empty src buffer!" + ); + + if (srcSize == 0) { + ZSTD_memset(&outSeqs[0], 0, sizeof(ZSTD_Sequence)); + return 1; + } + + { + ZSTD_Sequence const lastSeq = outSeqs[nbExternalSeqs - 1]; + + /* We can return early if lastSeq is already a block delimiter. */ + if (lastSeq.offset == 0 && lastSeq.matchLength == 0) { + return nbExternalSeqs; + } + + /* This error condition is only possible if the external matchfinder + * produced an invalid parse, by definition of ZSTD_sequenceBound(). */ + RETURN_ERROR_IF( + nbExternalSeqs == outSeqsCapacity, + sequenceProducer_failed, + "nbExternalSeqs == outSeqsCapacity but lastSeq is not a block delimiter!" + ); + + /* lastSeq is not a block delimiter, so we need to append one. */ + ZSTD_memset(&outSeqs[nbExternalSeqs], 0, sizeof(ZSTD_Sequence)); + return nbExternalSeqs + 1; + } +} + +/* ZSTD_fastSequenceLengthSum() : + * Returns sum(litLen) + sum(matchLen) + lastLits for *seqBuf*. + * Similar to another function in zstd_compress.c (determine_blockSize), + * except it doesn't check for a block delimiter to end summation. + * Removing the early exit allows the compiler to auto-vectorize (https://godbolt.org/z/cY1cajz9P). + * This function can be deleted and replaced by determine_blockSize after we resolve issue #3456. */ +static size_t ZSTD_fastSequenceLengthSum(ZSTD_Sequence const* seqBuf, size_t seqBufSize) { + size_t matchLenSum, litLenSum, i; + matchLenSum = 0; + litLenSum = 0; + for (i = 0; i < seqBufSize; i++) { + litLenSum += seqBuf[i].litLength; + matchLenSum += seqBuf[i].matchLength; + } + return litLenSum + matchLenSum; } typedef enum { ZSTDbss_compress, ZSTDbss_noCompress } ZSTD_buildSeqStore_e; @@ -2246,8 +3097,14 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) assert(srcSize <= ZSTD_BLOCKSIZE_MAX); /* Assert that we have correctly flushed the ctx params into the ms's copy */ ZSTD_assertEqualCParams(zc->appliedParams.cParams, ms->cParams); - if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) { - ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch); + /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding + * additional 1. We need to revisit and change this logic to be more consistent */ + if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1+1) { + if (zc->appliedParams.cParams.strategy >= ZSTD_btopt) { + ZSTD_ldm_skipRawSeqStoreBytes(&zc->externSeqStore, srcSize); + } else { + ZSTD_ldm_skipSequences(&zc->externSeqStore, srcSize, zc->appliedParams.cParams.minMatch); + } return ZSTDbss_noCompress; /* don't even attempt compression below a certain srcSize */ } ZSTD_resetSeqStore(&(zc->seqStore)); @@ -2263,10 +3120,10 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) /* limited update after a very long match */ { const BYTE* const base = ms->window.base; const BYTE* const istart = (const BYTE*)src; - const U32 current = (U32)(istart-base); + const U32 curr = (U32)(istart-base); if (sizeof(ptrdiff_t)==8) assert(istart - base < (ptrdiff_t)(U32)(-1)); /* ensure no overflow */ - if (current > ms->nextToUpdate + 384) - ms->nextToUpdate = current - MIN(192, (U32)(current - ms->nextToUpdate - 384)); + if (curr > ms->nextToUpdate + 384) + ms->nextToUpdate = curr - MIN(192, (U32)(curr - ms->nextToUpdate - 384)); } /* select and store sequences */ @@ -2277,16 +3134,34 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) zc->blockState.nextCBlock->rep[i] = zc->blockState.prevCBlock->rep[i]; } if (zc->externSeqStore.pos < zc->externSeqStore.size) { - assert(!zc->appliedParams.ldmParams.enableLdm); + assert(zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_disable); + + /* External matchfinder + LDM is technically possible, just not implemented yet. + * We need to revisit soon and implement it. */ + RETURN_ERROR_IF( + zc->appliedParams.useSequenceProducer, + parameter_combination_unsupported, + "Long-distance matching with external sequence producer enabled is not currently supported." + ); + /* Updates ldmSeqStore.pos */ lastLLSize = ZSTD_ldm_blockCompress(&zc->externSeqStore, ms, &zc->seqStore, zc->blockState.nextCBlock->rep, + zc->appliedParams.useRowMatchFinder, src, srcSize); assert(zc->externSeqStore.pos <= zc->externSeqStore.size); - } else if (zc->appliedParams.ldmParams.enableLdm) { - rawSeqStore_t ldmSeqStore = {NULL, 0, 0, 0}; + } else if (zc->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) { + rawSeqStore_t ldmSeqStore = kNullRawSeqStore; + + /* External matchfinder + LDM is technically possible, just not implemented yet. + * We need to revisit soon and implement it. */ + RETURN_ERROR_IF( + zc->appliedParams.useSequenceProducer, + parameter_combination_unsupported, + "Long-distance matching with external sequence producer enabled is not currently supported." + ); ldmSeqStore.seq = zc->ldmSequences; ldmSeqStore.capacity = zc->maxNbLdmSequences; @@ -2299,10 +3174,75 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) ZSTD_ldm_blockCompress(&ldmSeqStore, ms, &zc->seqStore, zc->blockState.nextCBlock->rep, + zc->appliedParams.useRowMatchFinder, src, srcSize); assert(ldmSeqStore.pos == ldmSeqStore.size); - } else { /* not long range mode */ - ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, dictMode); + } else if (zc->appliedParams.useSequenceProducer) { + assert( + zc->externalMatchCtx.seqBufferCapacity >= ZSTD_sequenceBound(srcSize) + ); + assert(zc->externalMatchCtx.mFinder != NULL); + + { U32 const windowSize = (U32)1 << zc->appliedParams.cParams.windowLog; + + size_t const nbExternalSeqs = (zc->externalMatchCtx.mFinder)( + zc->externalMatchCtx.mState, + zc->externalMatchCtx.seqBuffer, + zc->externalMatchCtx.seqBufferCapacity, + src, srcSize, + NULL, 0, /* dict and dictSize, currently not supported */ + zc->appliedParams.compressionLevel, + windowSize + ); + + size_t const nbPostProcessedSeqs = ZSTD_postProcessSequenceProducerResult( + zc->externalMatchCtx.seqBuffer, + nbExternalSeqs, + zc->externalMatchCtx.seqBufferCapacity, + srcSize + ); + + /* Return early if there is no error, since we don't need to worry about last literals */ + if (!ZSTD_isError(nbPostProcessedSeqs)) { + ZSTD_sequencePosition seqPos = {0,0,0}; + size_t const seqLenSum = ZSTD_fastSequenceLengthSum(zc->externalMatchCtx.seqBuffer, nbPostProcessedSeqs); + RETURN_ERROR_IF(seqLenSum > srcSize, externalSequences_invalid, "External sequences imply too large a block!"); + FORWARD_IF_ERROR( + ZSTD_copySequencesToSeqStoreExplicitBlockDelim( + zc, &seqPos, + zc->externalMatchCtx.seqBuffer, nbPostProcessedSeqs, + src, srcSize, + zc->appliedParams.searchForExternalRepcodes + ), + "Failed to copy external sequences to seqStore!" + ); + ms->ldmSeqStore = NULL; + DEBUGLOG(5, "Copied %lu sequences from external sequence producer to internal seqStore.", (unsigned long)nbExternalSeqs); + return ZSTDbss_compress; + } + + /* Propagate the error if fallback is disabled */ + if (!zc->appliedParams.enableMatchFinderFallback) { + return nbPostProcessedSeqs; + } + + /* Fallback to software matchfinder */ + { ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, + zc->appliedParams.useRowMatchFinder, + dictMode); + ms->ldmSeqStore = NULL; + DEBUGLOG( + 5, + "External sequence producer returned error code %lu. Falling back to internal parser.", + (unsigned long)nbExternalSeqs + ); + lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize); + } } + } else { /* not long range mode and no external matchfinder */ + ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, + zc->appliedParams.useRowMatchFinder, + dictMode); + ms->ldmSeqStore = NULL; lastLLSize = blockCompressor(ms, &zc->seqStore, zc->blockState.nextCBlock->rep, src, srcSize); } { const BYTE* const lastLiterals = (const BYTE*)src + srcSize - lastLLSize; @@ -2314,59 +3254,76 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize) static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc) { const seqStore_t* seqStore = ZSTD_getSeqStore(zc); - const seqDef* seqs = seqStore->sequencesStart; - size_t seqsSize = seqStore->sequences - seqs; + const seqDef* seqStoreSeqs = seqStore->sequencesStart; + size_t seqStoreSeqSize = seqStore->sequences - seqStoreSeqs; + size_t seqStoreLiteralsSize = (size_t)(seqStore->lit - seqStore->litStart); + size_t literalsRead = 0; + size_t lastLLSize; ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex]; - size_t i; size_t position; int repIdx; + size_t i; + repcodes_t updatedRepcodes; assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences); - for (i = 0, position = 0; i < seqsSize; ++i) { - outSeqs[i].offset = seqs[i].offset; - outSeqs[i].litLength = seqs[i].litLength; - outSeqs[i].matchLength = seqs[i].matchLength + MINMATCH; + /* Ensure we have enough space for last literals "sequence" */ + assert(zc->seqCollector.maxSequences >= seqStoreSeqSize + 1); + ZSTD_memcpy(updatedRepcodes.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); + for (i = 0; i < seqStoreSeqSize; ++i) { + U32 rawOffset = seqStoreSeqs[i].offBase - ZSTD_REP_NUM; + outSeqs[i].litLength = seqStoreSeqs[i].litLength; + outSeqs[i].matchLength = seqStoreSeqs[i].mlBase + MINMATCH; + outSeqs[i].rep = 0; if (i == seqStore->longLengthPos) { - if (seqStore->longLengthID == 1) { + if (seqStore->longLengthType == ZSTD_llt_literalLength) { outSeqs[i].litLength += 0x10000; - } else if (seqStore->longLengthID == 2) { + } else if (seqStore->longLengthType == ZSTD_llt_matchLength) { outSeqs[i].matchLength += 0x10000; } } - if (outSeqs[i].offset <= ZSTD_REP_NUM) { - outSeqs[i].rep = outSeqs[i].offset; - repIdx = (unsigned int)i - outSeqs[i].offset; - - if (outSeqs[i].litLength == 0) { - if (outSeqs[i].offset < 3) { - --repIdx; + if (seqStoreSeqs[i].offBase <= ZSTD_REP_NUM) { + /* Derive the correct offset corresponding to a repcode */ + outSeqs[i].rep = seqStoreSeqs[i].offBase; + if (outSeqs[i].litLength != 0) { + rawOffset = updatedRepcodes.rep[outSeqs[i].rep - 1]; + } else { + if (outSeqs[i].rep == 3) { + rawOffset = updatedRepcodes.rep[0] - 1; } else { - repIdx = (unsigned int)i - 1; + rawOffset = updatedRepcodes.rep[outSeqs[i].rep]; } - ++outSeqs[i].rep; } - assert(repIdx >= -3); - outSeqs[i].offset = repIdx >= 0 ? outSeqs[repIdx].offset : repStartValue[-repIdx - 1]; - if (outSeqs[i].rep == 4) { - --outSeqs[i].offset; - } - } else { - outSeqs[i].offset -= ZSTD_REP_NUM; } - - position += outSeqs[i].litLength; - outSeqs[i].matchPos = (unsigned int)position; - position += outSeqs[i].matchLength; + outSeqs[i].offset = rawOffset; + /* seqStoreSeqs[i].offset == offCode+1, and ZSTD_updateRep() expects offCode + so we provide seqStoreSeqs[i].offset - 1 */ + ZSTD_updateRep(updatedRepcodes.rep, + seqStoreSeqs[i].offBase, + seqStoreSeqs[i].litLength == 0); + literalsRead += outSeqs[i].litLength; } - zc->seqCollector.seqIndex += seqsSize; + /* Insert last literals (if any exist) in the block as a sequence with ml == off == 0. + * If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker + * for the block boundary, according to the API. + */ + assert(seqStoreLiteralsSize >= literalsRead); + lastLLSize = seqStoreLiteralsSize - literalsRead; + outSeqs[i].litLength = (U32)lastLLSize; + outSeqs[i].matchLength = outSeqs[i].offset = outSeqs[i].rep = 0; + seqStoreSeqSize++; + zc->seqCollector.seqIndex += seqStoreSeqSize; } -size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs, - size_t outSeqsSize, const void* src, size_t srcSize) +size_t ZSTD_sequenceBound(size_t srcSize) { + return (srcSize / ZSTD_MINMATCH_MIN) + 1; +} + +size_t ZSTD_generateSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs, + size_t outSeqsSize, const void* src, size_t srcSize) { const size_t dstCapacity = ZSTD_compressBound(srcSize); - void* dst = ZSTD_malloc(dstCapacity, ZSTD_defaultCMem); + void* dst = ZSTD_customMalloc(dstCapacity, ZSTD_defaultCMem); SeqCollector seqCollector; RETURN_ERROR_IF(dst == NULL, memory_allocation, "NULL pointer!"); @@ -2378,17 +3335,46 @@ size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs, zc->seqCollector = seqCollector; ZSTD_compress2(zc, dst, dstCapacity, src, srcSize); - ZSTD_free(dst, ZSTD_defaultCMem); + ZSTD_customFree(dst, ZSTD_defaultCMem); return zc->seqCollector.seqIndex; } -/* Returns true if the given block is a RLE block */ -static int ZSTD_isRLE(const BYTE *ip, size_t length) { +size_t ZSTD_mergeBlockDelimiters(ZSTD_Sequence* sequences, size_t seqsSize) { + size_t in = 0; + size_t out = 0; + for (; in < seqsSize; ++in) { + if (sequences[in].offset == 0 && sequences[in].matchLength == 0) { + if (in != seqsSize - 1) { + sequences[in+1].litLength += sequences[in].litLength; + } + } else { + sequences[out] = sequences[in]; + ++out; + } + } + return out; +} + +/* Unrolled loop to read four size_ts of input at a time. Returns 1 if is RLE, 0 if not. */ +static int ZSTD_isRLE(const BYTE* src, size_t length) { + const BYTE* ip = src; + const BYTE value = ip[0]; + const size_t valueST = (size_t)((U64)value * 0x0101010101010101ULL); + const size_t unrollSize = sizeof(size_t) * 4; + const size_t unrollMask = unrollSize - 1; + const size_t prefixLength = length & unrollMask; size_t i; - if (length < 2) return 1; - for (i = 1; i < length; ++i) { - if (ip[0] != ip[i]) return 0; + if (length == 1) return 1; + /* Check if prefix is RLE first before using unrolled loop */ + if (prefixLength && ZSTD_count(ip+1, ip, ip+prefixLength) != prefixLength-1) { + return 0; } + for (i = prefixLength; i != length; i += unrollSize) { + size_t u; + for (u = 0; u < unrollSize; u += sizeof(size_t)) { + if (MEM_readST(ip + i + u) != valueST) { + return 0; + } } } return 1; } @@ -2404,20 +3390,787 @@ static int ZSTD_maybeRLE(seqStore_t const* seqStore) return nbSeqs < 4 && nbLits < 10; } -static void ZSTD_confirmRepcodesAndEntropyTables(ZSTD_CCtx* zc) +static void +ZSTD_blockState_confirmRepcodesAndEntropyTables(ZSTD_blockState_t* const bs) +{ + ZSTD_compressedBlockState_t* const tmp = bs->prevCBlock; + bs->prevCBlock = bs->nextCBlock; + bs->nextCBlock = tmp; +} + +/* Writes the block header */ +static void +writeBlockHeader(void* op, size_t cSize, size_t blockSize, U32 lastBlock) +{ + U32 const cBlockHeader = cSize == 1 ? + lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) : + lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3); + MEM_writeLE24(op, cBlockHeader); + DEBUGLOG(3, "writeBlockHeader: cSize: %zu blockSize: %zu lastBlock: %u", cSize, blockSize, lastBlock); +} + +/** ZSTD_buildBlockEntropyStats_literals() : + * Builds entropy for the literals. + * Stores literals block type (raw, rle, compressed, repeat) and + * huffman description table to hufMetadata. + * Requires ENTROPY_WORKSPACE_SIZE workspace + * @return : size of huffman description table, or an error code + */ +static size_t +ZSTD_buildBlockEntropyStats_literals(void* const src, size_t srcSize, + const ZSTD_hufCTables_t* prevHuf, + ZSTD_hufCTables_t* nextHuf, + ZSTD_hufCTablesMetadata_t* hufMetadata, + const int literalsCompressionIsDisabled, + void* workspace, size_t wkspSize, + int hufFlags) +{ + BYTE* const wkspStart = (BYTE*)workspace; + BYTE* const wkspEnd = wkspStart + wkspSize; + BYTE* const countWkspStart = wkspStart; + unsigned* const countWksp = (unsigned*)workspace; + const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned); + BYTE* const nodeWksp = countWkspStart + countWkspSize; + const size_t nodeWkspSize = (size_t)(wkspEnd - nodeWksp); + unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX; + unsigned huffLog = LitHufLog; + HUF_repeat repeat = prevHuf->repeatMode; + DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_literals (srcSize=%zu)", srcSize); + + /* Prepare nextEntropy assuming reusing the existing table */ + ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); + + if (literalsCompressionIsDisabled) { + DEBUGLOG(5, "set_basic - disabled"); + hufMetadata->hType = set_basic; + return 0; + } + + /* small ? don't even attempt compression (speed opt) */ +#ifndef COMPRESS_LITERALS_SIZE_MIN +# define COMPRESS_LITERALS_SIZE_MIN 63 /* heuristic */ +#endif + { size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN; + if (srcSize <= minLitSize) { + DEBUGLOG(5, "set_basic - too small"); + hufMetadata->hType = set_basic; + return 0; + } } + + /* Scan input and build symbol stats */ + { size_t const largest = + HIST_count_wksp (countWksp, &maxSymbolValue, + (const BYTE*)src, srcSize, + workspace, wkspSize); + FORWARD_IF_ERROR(largest, "HIST_count_wksp failed"); + if (largest == srcSize) { + /* only one literal symbol */ + DEBUGLOG(5, "set_rle"); + hufMetadata->hType = set_rle; + return 0; + } + if (largest <= (srcSize >> 7)+4) { + /* heuristic: likely not compressible */ + DEBUGLOG(5, "set_basic - no gain"); + hufMetadata->hType = set_basic; + return 0; + } } + + /* Validate the previous Huffman table */ + if (repeat == HUF_repeat_check + && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) { + repeat = HUF_repeat_none; + } + + /* Build Huffman Tree */ + ZSTD_memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable)); + huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue, nodeWksp, nodeWkspSize, nextHuf->CTable, countWksp, hufFlags); + assert(huffLog <= LitHufLog); + { size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp, + maxSymbolValue, huffLog, + nodeWksp, nodeWkspSize); + FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp"); + huffLog = (U32)maxBits; + } + { /* Build and write the CTable */ + size_t const newCSize = HUF_estimateCompressedSize( + (HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue); + size_t const hSize = HUF_writeCTable_wksp( + hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer), + (HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog, + nodeWksp, nodeWkspSize); + /* Check against repeating the previous CTable */ + if (repeat != HUF_repeat_none) { + size_t const oldCSize = HUF_estimateCompressedSize( + (HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue); + if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) { + DEBUGLOG(5, "set_repeat - smaller"); + ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); + hufMetadata->hType = set_repeat; + return 0; + } } + if (newCSize + hSize >= srcSize) { + DEBUGLOG(5, "set_basic - no gains"); + ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); + hufMetadata->hType = set_basic; + return 0; + } + DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize); + hufMetadata->hType = set_compressed; + nextHuf->repeatMode = HUF_repeat_check; + return hSize; + } +} + + +/* ZSTD_buildDummySequencesStatistics(): + * Returns a ZSTD_symbolEncodingTypeStats_t with all encoding types as set_basic, + * and updates nextEntropy to the appropriate repeatMode. + */ +static ZSTD_symbolEncodingTypeStats_t +ZSTD_buildDummySequencesStatistics(ZSTD_fseCTables_t* nextEntropy) +{ + ZSTD_symbolEncodingTypeStats_t stats = {set_basic, set_basic, set_basic, 0, 0, 0}; + nextEntropy->litlength_repeatMode = FSE_repeat_none; + nextEntropy->offcode_repeatMode = FSE_repeat_none; + nextEntropy->matchlength_repeatMode = FSE_repeat_none; + return stats; +} + +/** ZSTD_buildBlockEntropyStats_sequences() : + * Builds entropy for the sequences. + * Stores symbol compression modes and fse table to fseMetadata. + * Requires ENTROPY_WORKSPACE_SIZE wksp. + * @return : size of fse tables or error code */ +static size_t +ZSTD_buildBlockEntropyStats_sequences( + const seqStore_t* seqStorePtr, + const ZSTD_fseCTables_t* prevEntropy, + ZSTD_fseCTables_t* nextEntropy, + const ZSTD_CCtx_params* cctxParams, + ZSTD_fseCTablesMetadata_t* fseMetadata, + void* workspace, size_t wkspSize) +{ + ZSTD_strategy const strategy = cctxParams->cParams.strategy; + size_t const nbSeq = (size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart); + BYTE* const ostart = fseMetadata->fseTablesBuffer; + BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer); + BYTE* op = ostart; + unsigned* countWorkspace = (unsigned*)workspace; + unsigned* entropyWorkspace = countWorkspace + (MaxSeq + 1); + size_t entropyWorkspaceSize = wkspSize - (MaxSeq + 1) * sizeof(*countWorkspace); + ZSTD_symbolEncodingTypeStats_t stats; + + DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_sequences (nbSeq=%zu)", nbSeq); + stats = nbSeq != 0 ? ZSTD_buildSequencesStatistics(seqStorePtr, nbSeq, + prevEntropy, nextEntropy, op, oend, + strategy, countWorkspace, + entropyWorkspace, entropyWorkspaceSize) + : ZSTD_buildDummySequencesStatistics(nextEntropy); + FORWARD_IF_ERROR(stats.size, "ZSTD_buildSequencesStatistics failed!"); + fseMetadata->llType = (symbolEncodingType_e) stats.LLtype; + fseMetadata->ofType = (symbolEncodingType_e) stats.Offtype; + fseMetadata->mlType = (symbolEncodingType_e) stats.MLtype; + fseMetadata->lastCountSize = stats.lastCountSize; + return stats.size; +} + + +/** ZSTD_buildBlockEntropyStats() : + * Builds entropy for the block. + * Requires workspace size ENTROPY_WORKSPACE_SIZE + * @return : 0 on success, or an error code + * Note : also employed in superblock + */ +size_t ZSTD_buildBlockEntropyStats( + const seqStore_t* seqStorePtr, + const ZSTD_entropyCTables_t* prevEntropy, + ZSTD_entropyCTables_t* nextEntropy, + const ZSTD_CCtx_params* cctxParams, + ZSTD_entropyCTablesMetadata_t* entropyMetadata, + void* workspace, size_t wkspSize) +{ + size_t const litSize = (size_t)(seqStorePtr->lit - seqStorePtr->litStart); + int const huf_useOptDepth = (cctxParams->cParams.strategy >= HUF_OPTIMAL_DEPTH_THRESHOLD); + int const hufFlags = huf_useOptDepth ? HUF_flags_optimalDepth : 0; + + entropyMetadata->hufMetadata.hufDesSize = + ZSTD_buildBlockEntropyStats_literals(seqStorePtr->litStart, litSize, + &prevEntropy->huf, &nextEntropy->huf, + &entropyMetadata->hufMetadata, + ZSTD_literalsCompressionIsDisabled(cctxParams), + workspace, wkspSize, hufFlags); + + FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildBlockEntropyStats_literals failed"); + entropyMetadata->fseMetadata.fseTablesSize = + ZSTD_buildBlockEntropyStats_sequences(seqStorePtr, + &prevEntropy->fse, &nextEntropy->fse, + cctxParams, + &entropyMetadata->fseMetadata, + workspace, wkspSize); + FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildBlockEntropyStats_sequences failed"); + return 0; +} + +/* Returns the size estimate for the literals section (header + content) of a block */ +static size_t +ZSTD_estimateBlockSize_literal(const BYTE* literals, size_t litSize, + const ZSTD_hufCTables_t* huf, + const ZSTD_hufCTablesMetadata_t* hufMetadata, + void* workspace, size_t wkspSize, + int writeEntropy) +{ + unsigned* const countWksp = (unsigned*)workspace; + unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX; + size_t literalSectionHeaderSize = 3 + (litSize >= 1 KB) + (litSize >= 16 KB); + U32 singleStream = litSize < 256; + + if (hufMetadata->hType == set_basic) return litSize; + else if (hufMetadata->hType == set_rle) return 1; + else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) { + size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize); + if (ZSTD_isError(largest)) return litSize; + { size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue); + if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize; + if (!singleStream) cLitSizeEstimate += 6; /* multi-stream huffman uses 6-byte jump table */ + return cLitSizeEstimate + literalSectionHeaderSize; + } } + assert(0); /* impossible */ + return 0; +} + +/* Returns the size estimate for the FSE-compressed symbols (of, ml, ll) of a block */ +static size_t +ZSTD_estimateBlockSize_symbolType(symbolEncodingType_e type, + const BYTE* codeTable, size_t nbSeq, unsigned maxCode, + const FSE_CTable* fseCTable, + const U8* additionalBits, + short const* defaultNorm, U32 defaultNormLog, U32 defaultMax, + void* workspace, size_t wkspSize) +{ + unsigned* const countWksp = (unsigned*)workspace; + const BYTE* ctp = codeTable; + const BYTE* const ctStart = ctp; + const BYTE* const ctEnd = ctStart + nbSeq; + size_t cSymbolTypeSizeEstimateInBits = 0; + unsigned max = maxCode; + + HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */ + if (type == set_basic) { + /* We selected this encoding type, so it must be valid. */ + assert(max <= defaultMax); + (void)defaultMax; + cSymbolTypeSizeEstimateInBits = ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max); + } else if (type == set_rle) { + cSymbolTypeSizeEstimateInBits = 0; + } else if (type == set_compressed || type == set_repeat) { + cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max); + } + if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) { + return nbSeq * 10; + } + while (ctp < ctEnd) { + if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp]; + else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */ + ctp++; + } + return cSymbolTypeSizeEstimateInBits >> 3; +} + +/* Returns the size estimate for the sequences section (header + content) of a block */ +static size_t +ZSTD_estimateBlockSize_sequences(const BYTE* ofCodeTable, + const BYTE* llCodeTable, + const BYTE* mlCodeTable, + size_t nbSeq, + const ZSTD_fseCTables_t* fseTables, + const ZSTD_fseCTablesMetadata_t* fseMetadata, + void* workspace, size_t wkspSize, + int writeEntropy) +{ + size_t sequencesSectionHeaderSize = 1 /* seqHead */ + 1 /* min seqSize size */ + (nbSeq >= 128) + (nbSeq >= LONGNBSEQ); + size_t cSeqSizeEstimate = 0; + cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, nbSeq, MaxOff, + fseTables->offcodeCTable, NULL, + OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff, + workspace, wkspSize); + cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->llType, llCodeTable, nbSeq, MaxLL, + fseTables->litlengthCTable, LL_bits, + LL_defaultNorm, LL_defaultNormLog, MaxLL, + workspace, wkspSize); + cSeqSizeEstimate += ZSTD_estimateBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, nbSeq, MaxML, + fseTables->matchlengthCTable, ML_bits, + ML_defaultNorm, ML_defaultNormLog, MaxML, + workspace, wkspSize); + if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize; + return cSeqSizeEstimate + sequencesSectionHeaderSize; +} + +/* Returns the size estimate for a given stream of literals, of, ll, ml */ +static size_t +ZSTD_estimateBlockSize(const BYTE* literals, size_t litSize, + const BYTE* ofCodeTable, + const BYTE* llCodeTable, + const BYTE* mlCodeTable, + size_t nbSeq, + const ZSTD_entropyCTables_t* entropy, + const ZSTD_entropyCTablesMetadata_t* entropyMetadata, + void* workspace, size_t wkspSize, + int writeLitEntropy, int writeSeqEntropy) +{ + size_t const literalsSize = ZSTD_estimateBlockSize_literal(literals, litSize, + &entropy->huf, &entropyMetadata->hufMetadata, + workspace, wkspSize, writeLitEntropy); + size_t const seqSize = ZSTD_estimateBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable, + nbSeq, &entropy->fse, &entropyMetadata->fseMetadata, + workspace, wkspSize, writeSeqEntropy); + return seqSize + literalsSize + ZSTD_blockHeaderSize; +} + +/* Builds entropy statistics and uses them for blocksize estimation. + * + * @return: estimated compressed size of the seqStore, or a zstd error. + */ +static size_t +ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(seqStore_t* seqStore, ZSTD_CCtx* zc) +{ + ZSTD_entropyCTablesMetadata_t* const entropyMetadata = &zc->blockSplitCtx.entropyMetadata; + DEBUGLOG(6, "ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize()"); + FORWARD_IF_ERROR(ZSTD_buildBlockEntropyStats(seqStore, + &zc->blockState.prevCBlock->entropy, + &zc->blockState.nextCBlock->entropy, + &zc->appliedParams, + entropyMetadata, + zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE), ""); + return ZSTD_estimateBlockSize( + seqStore->litStart, (size_t)(seqStore->lit - seqStore->litStart), + seqStore->ofCode, seqStore->llCode, seqStore->mlCode, + (size_t)(seqStore->sequences - seqStore->sequencesStart), + &zc->blockState.nextCBlock->entropy, + entropyMetadata, + zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE, + (int)(entropyMetadata->hufMetadata.hType == set_compressed), 1); +} + +/* Returns literals bytes represented in a seqStore */ +static size_t ZSTD_countSeqStoreLiteralsBytes(const seqStore_t* const seqStore) +{ + size_t literalsBytes = 0; + size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart); + size_t i; + for (i = 0; i < nbSeqs; ++i) { + seqDef const seq = seqStore->sequencesStart[i]; + literalsBytes += seq.litLength; + if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_literalLength) { + literalsBytes += 0x10000; + } } + return literalsBytes; +} + +/* Returns match bytes represented in a seqStore */ +static size_t ZSTD_countSeqStoreMatchBytes(const seqStore_t* const seqStore) +{ + size_t matchBytes = 0; + size_t const nbSeqs = (size_t)(seqStore->sequences - seqStore->sequencesStart); + size_t i; + for (i = 0; i < nbSeqs; ++i) { + seqDef seq = seqStore->sequencesStart[i]; + matchBytes += seq.mlBase + MINMATCH; + if (i == seqStore->longLengthPos && seqStore->longLengthType == ZSTD_llt_matchLength) { + matchBytes += 0x10000; + } } + return matchBytes; +} + +/* Derives the seqStore that is a chunk of the originalSeqStore from [startIdx, endIdx). + * Stores the result in resultSeqStore. + */ +static void ZSTD_deriveSeqStoreChunk(seqStore_t* resultSeqStore, + const seqStore_t* originalSeqStore, + size_t startIdx, size_t endIdx) +{ + *resultSeqStore = *originalSeqStore; + if (startIdx > 0) { + resultSeqStore->sequences = originalSeqStore->sequencesStart + startIdx; + resultSeqStore->litStart += ZSTD_countSeqStoreLiteralsBytes(resultSeqStore); + } + + /* Move longLengthPos into the correct position if necessary */ + if (originalSeqStore->longLengthType != ZSTD_llt_none) { + if (originalSeqStore->longLengthPos < startIdx || originalSeqStore->longLengthPos > endIdx) { + resultSeqStore->longLengthType = ZSTD_llt_none; + } else { + resultSeqStore->longLengthPos -= (U32)startIdx; + } + } + resultSeqStore->sequencesStart = originalSeqStore->sequencesStart + startIdx; + resultSeqStore->sequences = originalSeqStore->sequencesStart + endIdx; + if (endIdx == (size_t)(originalSeqStore->sequences - originalSeqStore->sequencesStart)) { + /* This accounts for possible last literals if the derived chunk reaches the end of the block */ + assert(resultSeqStore->lit == originalSeqStore->lit); + } else { + size_t const literalsBytes = ZSTD_countSeqStoreLiteralsBytes(resultSeqStore); + resultSeqStore->lit = resultSeqStore->litStart + literalsBytes; + } + resultSeqStore->llCode += startIdx; + resultSeqStore->mlCode += startIdx; + resultSeqStore->ofCode += startIdx; +} + +/** + * Returns the raw offset represented by the combination of offBase, ll0, and repcode history. + * offBase must represent a repcode in the numeric representation of ZSTD_storeSeq(). + */ +static U32 +ZSTD_resolveRepcodeToRawOffset(const U32 rep[ZSTD_REP_NUM], const U32 offBase, const U32 ll0) +{ + U32 const adjustedRepCode = OFFBASE_TO_REPCODE(offBase) - 1 + ll0; /* [ 0 - 3 ] */ + assert(OFFBASE_IS_REPCODE(offBase)); + if (adjustedRepCode == ZSTD_REP_NUM) { + assert(ll0); + /* litlength == 0 and offCode == 2 implies selection of first repcode - 1 + * This is only valid if it results in a valid offset value, aka > 0. + * Note : it may happen that `rep[0]==1` in exceptional circumstances. + * In which case this function will return 0, which is an invalid offset. + * It's not an issue though, since this value will be + * compared and discarded within ZSTD_seqStore_resolveOffCodes(). + */ + return rep[0] - 1; + } + return rep[adjustedRepCode]; +} + +/** + * ZSTD_seqStore_resolveOffCodes() reconciles any possible divergences in offset history that may arise + * due to emission of RLE/raw blocks that disturb the offset history, + * and replaces any repcodes within the seqStore that may be invalid. + * + * dRepcodes are updated as would be on the decompression side. + * cRepcodes are updated exactly in accordance with the seqStore. + * + * Note : this function assumes seq->offBase respects the following numbering scheme : + * 0 : invalid + * 1-3 : repcode 1-3 + * 4+ : real_offset+3 + */ +static void +ZSTD_seqStore_resolveOffCodes(repcodes_t* const dRepcodes, repcodes_t* const cRepcodes, + const seqStore_t* const seqStore, U32 const nbSeq) +{ + U32 idx = 0; + for (; idx < nbSeq; ++idx) { + seqDef* const seq = seqStore->sequencesStart + idx; + U32 const ll0 = (seq->litLength == 0); + U32 const offBase = seq->offBase; + assert(offBase > 0); + if (OFFBASE_IS_REPCODE(offBase)) { + U32 const dRawOffset = ZSTD_resolveRepcodeToRawOffset(dRepcodes->rep, offBase, ll0); + U32 const cRawOffset = ZSTD_resolveRepcodeToRawOffset(cRepcodes->rep, offBase, ll0); + /* Adjust simulated decompression repcode history if we come across a mismatch. Replace + * the repcode with the offset it actually references, determined by the compression + * repcode history. + */ + if (dRawOffset != cRawOffset) { + seq->offBase = OFFSET_TO_OFFBASE(cRawOffset); + } + } + /* Compression repcode history is always updated with values directly from the unmodified seqStore. + * Decompression repcode history may use modified seq->offset value taken from compression repcode history. + */ + ZSTD_updateRep(dRepcodes->rep, seq->offBase, ll0); + ZSTD_updateRep(cRepcodes->rep, offBase, ll0); + } +} + +/* ZSTD_compressSeqStore_singleBlock(): + * Compresses a seqStore into a block with a block header, into the buffer dst. + * + * Returns the total size of that block (including header) or a ZSTD error code. + */ +static size_t +ZSTD_compressSeqStore_singleBlock(ZSTD_CCtx* zc, + const seqStore_t* const seqStore, + repcodes_t* const dRep, repcodes_t* const cRep, + void* dst, size_t dstCapacity, + const void* src, size_t srcSize, + U32 lastBlock, U32 isPartition) +{ + const U32 rleMaxLength = 25; + BYTE* op = (BYTE*)dst; + const BYTE* ip = (const BYTE*)src; + size_t cSize; + size_t cSeqsSize; + + /* In case of an RLE or raw block, the simulated decompression repcode history must be reset */ + repcodes_t const dRepOriginal = *dRep; + DEBUGLOG(5, "ZSTD_compressSeqStore_singleBlock"); + if (isPartition) + ZSTD_seqStore_resolveOffCodes(dRep, cRep, seqStore, (U32)(seqStore->sequences - seqStore->sequencesStart)); + + RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "Block header doesn't fit"); + cSeqsSize = ZSTD_entropyCompressSeqStore(seqStore, + &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy, + &zc->appliedParams, + op + ZSTD_blockHeaderSize, dstCapacity - ZSTD_blockHeaderSize, + srcSize, + zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, + zc->bmi2); + FORWARD_IF_ERROR(cSeqsSize, "ZSTD_entropyCompressSeqStore failed!"); + + if (!zc->isFirstBlock && + cSeqsSize < rleMaxLength && + ZSTD_isRLE((BYTE const*)src, srcSize)) { + /* We don't want to emit our first block as a RLE even if it qualifies because + * doing so will cause the decoder (cli only) to throw a "should consume all input error." + * This is only an issue for zstd <= v1.4.3 + */ + cSeqsSize = 1; + } + + if (zc->seqCollector.collectSequences) { + ZSTD_copyBlockSequences(zc); + ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); + return 0; + } + + if (cSeqsSize == 0) { + cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, srcSize, lastBlock); + FORWARD_IF_ERROR(cSize, "Nocompress block failed"); + DEBUGLOG(4, "Writing out nocompress block, size: %zu", cSize); + *dRep = dRepOriginal; /* reset simulated decompression repcode history */ + } else if (cSeqsSize == 1) { + cSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, srcSize, lastBlock); + FORWARD_IF_ERROR(cSize, "RLE compress block failed"); + DEBUGLOG(4, "Writing out RLE block, size: %zu", cSize); + *dRep = dRepOriginal; /* reset simulated decompression repcode history */ + } else { + ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); + writeBlockHeader(op, cSeqsSize, srcSize, lastBlock); + cSize = ZSTD_blockHeaderSize + cSeqsSize; + DEBUGLOG(4, "Writing out compressed block, size: %zu", cSize); + } + + if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) + zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; + + return cSize; +} + +/* Struct to keep track of where we are in our recursive calls. */ +typedef struct { + U32* splitLocations; /* Array of split indices */ + size_t idx; /* The current index within splitLocations being worked on */ +} seqStoreSplits; + +#define MIN_SEQUENCES_BLOCK_SPLITTING 300 + +/* Helper function to perform the recursive search for block splits. + * Estimates the cost of seqStore prior to split, and estimates the cost of splitting the sequences in half. + * If advantageous to split, then we recurse down the two sub-blocks. + * If not, or if an error occurred in estimation, then we do not recurse. + * + * Note: The recursion depth is capped by a heuristic minimum number of sequences, + * defined by MIN_SEQUENCES_BLOCK_SPLITTING. + * In theory, this means the absolute largest recursion depth is 10 == log2(maxNbSeqInBlock/MIN_SEQUENCES_BLOCK_SPLITTING). + * In practice, recursion depth usually doesn't go beyond 4. + * + * Furthermore, the number of splits is capped by ZSTD_MAX_NB_BLOCK_SPLITS. + * At ZSTD_MAX_NB_BLOCK_SPLITS == 196 with the current existing blockSize + * maximum of 128 KB, this value is actually impossible to reach. + */ +static void +ZSTD_deriveBlockSplitsHelper(seqStoreSplits* splits, size_t startIdx, size_t endIdx, + ZSTD_CCtx* zc, const seqStore_t* origSeqStore) +{ + seqStore_t* const fullSeqStoreChunk = &zc->blockSplitCtx.fullSeqStoreChunk; + seqStore_t* const firstHalfSeqStore = &zc->blockSplitCtx.firstHalfSeqStore; + seqStore_t* const secondHalfSeqStore = &zc->blockSplitCtx.secondHalfSeqStore; + size_t estimatedOriginalSize; + size_t estimatedFirstHalfSize; + size_t estimatedSecondHalfSize; + size_t midIdx = (startIdx + endIdx)/2; + + DEBUGLOG(5, "ZSTD_deriveBlockSplitsHelper: startIdx=%zu endIdx=%zu", startIdx, endIdx); + assert(endIdx >= startIdx); + if (endIdx - startIdx < MIN_SEQUENCES_BLOCK_SPLITTING || splits->idx >= ZSTD_MAX_NB_BLOCK_SPLITS) { + DEBUGLOG(6, "ZSTD_deriveBlockSplitsHelper: Too few sequences (%zu)", endIdx - startIdx); + return; + } + ZSTD_deriveSeqStoreChunk(fullSeqStoreChunk, origSeqStore, startIdx, endIdx); + ZSTD_deriveSeqStoreChunk(firstHalfSeqStore, origSeqStore, startIdx, midIdx); + ZSTD_deriveSeqStoreChunk(secondHalfSeqStore, origSeqStore, midIdx, endIdx); + estimatedOriginalSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(fullSeqStoreChunk, zc); + estimatedFirstHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(firstHalfSeqStore, zc); + estimatedSecondHalfSize = ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(secondHalfSeqStore, zc); + DEBUGLOG(5, "Estimated original block size: %zu -- First half split: %zu -- Second half split: %zu", + estimatedOriginalSize, estimatedFirstHalfSize, estimatedSecondHalfSize); + if (ZSTD_isError(estimatedOriginalSize) || ZSTD_isError(estimatedFirstHalfSize) || ZSTD_isError(estimatedSecondHalfSize)) { + return; + } + if (estimatedFirstHalfSize + estimatedSecondHalfSize < estimatedOriginalSize) { + DEBUGLOG(5, "split decided at seqNb:%zu", midIdx); + ZSTD_deriveBlockSplitsHelper(splits, startIdx, midIdx, zc, origSeqStore); + splits->splitLocations[splits->idx] = (U32)midIdx; + splits->idx++; + ZSTD_deriveBlockSplitsHelper(splits, midIdx, endIdx, zc, origSeqStore); + } +} + +/* Base recursive function. + * Populates a table with intra-block partition indices that can improve compression ratio. + * + * @return: number of splits made (which equals the size of the partition table - 1). + */ +static size_t ZSTD_deriveBlockSplits(ZSTD_CCtx* zc, U32 partitions[], U32 nbSeq) +{ + seqStoreSplits splits; + splits.splitLocations = partitions; + splits.idx = 0; + if (nbSeq <= 4) { + DEBUGLOG(5, "ZSTD_deriveBlockSplits: Too few sequences to split (%u <= 4)", nbSeq); + /* Refuse to try and split anything with less than 4 sequences */ + return 0; + } + ZSTD_deriveBlockSplitsHelper(&splits, 0, nbSeq, zc, &zc->seqStore); + splits.splitLocations[splits.idx] = nbSeq; + DEBUGLOG(5, "ZSTD_deriveBlockSplits: final nb partitions: %zu", splits.idx+1); + return splits.idx; +} + +/* ZSTD_compressBlock_splitBlock(): + * Attempts to split a given block into multiple blocks to improve compression ratio. + * + * Returns combined size of all blocks (which includes headers), or a ZSTD error code. + */ +static size_t +ZSTD_compressBlock_splitBlock_internal(ZSTD_CCtx* zc, + void* dst, size_t dstCapacity, + const void* src, size_t blockSize, + U32 lastBlock, U32 nbSeq) +{ + size_t cSize = 0; + const BYTE* ip = (const BYTE*)src; + BYTE* op = (BYTE*)dst; + size_t i = 0; + size_t srcBytesTotal = 0; + U32* const partitions = zc->blockSplitCtx.partitions; /* size == ZSTD_MAX_NB_BLOCK_SPLITS */ + seqStore_t* const nextSeqStore = &zc->blockSplitCtx.nextSeqStore; + seqStore_t* const currSeqStore = &zc->blockSplitCtx.currSeqStore; + size_t const numSplits = ZSTD_deriveBlockSplits(zc, partitions, nbSeq); + + /* If a block is split and some partitions are emitted as RLE/uncompressed, then repcode history + * may become invalid. In order to reconcile potentially invalid repcodes, we keep track of two + * separate repcode histories that simulate repcode history on compression and decompression side, + * and use the histories to determine whether we must replace a particular repcode with its raw offset. + * + * 1) cRep gets updated for each partition, regardless of whether the block was emitted as uncompressed + * or RLE. This allows us to retrieve the offset value that an invalid repcode references within + * a nocompress/RLE block. + * 2) dRep gets updated only for compressed partitions, and when a repcode gets replaced, will use + * the replacement offset value rather than the original repcode to update the repcode history. + * dRep also will be the final repcode history sent to the next block. + * + * See ZSTD_seqStore_resolveOffCodes() for more details. + */ + repcodes_t dRep; + repcodes_t cRep; + ZSTD_memcpy(dRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); + ZSTD_memcpy(cRep.rep, zc->blockState.prevCBlock->rep, sizeof(repcodes_t)); + ZSTD_memset(nextSeqStore, 0, sizeof(seqStore_t)); + + DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)", + (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit, + (unsigned)zc->blockState.matchState.nextToUpdate); + + if (numSplits == 0) { + size_t cSizeSingleBlock = + ZSTD_compressSeqStore_singleBlock(zc, &zc->seqStore, + &dRep, &cRep, + op, dstCapacity, + ip, blockSize, + lastBlock, 0 /* isPartition */); + FORWARD_IF_ERROR(cSizeSingleBlock, "Compressing single block from splitBlock_internal() failed!"); + DEBUGLOG(5, "ZSTD_compressBlock_splitBlock_internal: No splits"); + assert(zc->blockSize <= ZSTD_BLOCKSIZE_MAX); + assert(cSizeSingleBlock <= zc->blockSize + ZSTD_blockHeaderSize); + return cSizeSingleBlock; + } + + ZSTD_deriveSeqStoreChunk(currSeqStore, &zc->seqStore, 0, partitions[0]); + for (i = 0; i <= numSplits; ++i) { + size_t cSizeChunk; + U32 const lastPartition = (i == numSplits); + U32 lastBlockEntireSrc = 0; + + size_t srcBytes = ZSTD_countSeqStoreLiteralsBytes(currSeqStore) + ZSTD_countSeqStoreMatchBytes(currSeqStore); + srcBytesTotal += srcBytes; + if (lastPartition) { + /* This is the final partition, need to account for possible last literals */ + srcBytes += blockSize - srcBytesTotal; + lastBlockEntireSrc = lastBlock; + } else { + ZSTD_deriveSeqStoreChunk(nextSeqStore, &zc->seqStore, partitions[i], partitions[i+1]); + } + + cSizeChunk = ZSTD_compressSeqStore_singleBlock(zc, currSeqStore, + &dRep, &cRep, + op, dstCapacity, + ip, srcBytes, + lastBlockEntireSrc, 1 /* isPartition */); + DEBUGLOG(5, "Estimated size: %zu vs %zu : actual size", + ZSTD_buildEntropyStatisticsAndEstimateSubBlockSize(currSeqStore, zc), cSizeChunk); + FORWARD_IF_ERROR(cSizeChunk, "Compressing chunk failed!"); + + ip += srcBytes; + op += cSizeChunk; + dstCapacity -= cSizeChunk; + cSize += cSizeChunk; + *currSeqStore = *nextSeqStore; + assert(cSizeChunk <= zc->blockSize + ZSTD_blockHeaderSize); + } + /* cRep and dRep may have diverged during the compression. + * If so, we use the dRep repcodes for the next block. + */ + ZSTD_memcpy(zc->blockState.prevCBlock->rep, dRep.rep, sizeof(repcodes_t)); + return cSize; +} + +static size_t +ZSTD_compressBlock_splitBlock(ZSTD_CCtx* zc, + void* dst, size_t dstCapacity, + const void* src, size_t srcSize, U32 lastBlock) { - ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock; - zc->blockState.prevCBlock = zc->blockState.nextCBlock; - zc->blockState.nextCBlock = tmp; + U32 nbSeq; + size_t cSize; + DEBUGLOG(4, "ZSTD_compressBlock_splitBlock"); + assert(zc->appliedParams.useBlockSplitter == ZSTD_ps_enable); + + { const size_t bss = ZSTD_buildSeqStore(zc, src, srcSize); + FORWARD_IF_ERROR(bss, "ZSTD_buildSeqStore failed"); + if (bss == ZSTDbss_noCompress) { + if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) + zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; + cSize = ZSTD_noCompressBlock(dst, dstCapacity, src, srcSize, lastBlock); + FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed"); + DEBUGLOG(4, "ZSTD_compressBlock_splitBlock: Nocompress block"); + return cSize; + } + nbSeq = (U32)(zc->seqStore.sequences - zc->seqStore.sequencesStart); + } + + cSize = ZSTD_compressBlock_splitBlock_internal(zc, dst, dstCapacity, src, srcSize, lastBlock, nbSeq); + FORWARD_IF_ERROR(cSize, "Splitting blocks failed!"); + return cSize; } -static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc, - void* dst, size_t dstCapacity, - const void* src, size_t srcSize, U32 frame) +static size_t +ZSTD_compressBlock_internal(ZSTD_CCtx* zc, + void* dst, size_t dstCapacity, + const void* src, size_t srcSize, U32 frame) { - /* This the upper bound for the length of an rle block. - * This isn't the actual upper bound. Finding the real threshold - * needs further investigation. + /* This is an estimated upper bound for the length of an rle block. + * This isn't the actual upper bound. + * Finding the real threshold needs further investigation. */ const U32 rleMaxLength = 25; size_t cSize; @@ -2434,16 +4187,17 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc, if (zc->seqCollector.collectSequences) { ZSTD_copyBlockSequences(zc); + ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); return 0; } /* encode sequences and literals */ - cSize = ZSTD_compressSequences(&zc->seqStore, + cSize = ZSTD_entropyCompressSeqStore(&zc->seqStore, &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy, &zc->appliedParams, dst, dstCapacity, srcSize, - zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */, + zc->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, zc->bmi2); if (frame && @@ -2461,7 +4215,7 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc, out: if (!ZSTD_isError(cSize) && cSize > 1) { - ZSTD_confirmRepcodesAndEntropyTables(zc); + ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); } /* We check that dictionaries have offset codes available for the first * block. After the first block, the offcode table might not have large @@ -2508,18 +4262,19 @@ static size_t ZSTD_compressBlock_targetCBlockSize_body(ZSTD_CCtx* zc, * * cSize >= blockBound(srcSize): We have expanded the block too much so * emit an uncompressed block. */ - { - size_t const cSize = ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock); + { size_t const cSize = + ZSTD_compressSuperBlock(zc, dst, dstCapacity, src, srcSize, lastBlock); if (cSize != ERROR(dstSize_tooSmall)) { - size_t const maxCSize = srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy); + size_t const maxCSize = + srcSize - ZSTD_minGain(srcSize, zc->appliedParams.cParams.strategy); FORWARD_IF_ERROR(cSize, "ZSTD_compressSuperBlock failed"); if (cSize != 0 && cSize < maxCSize + ZSTD_blockHeaderSize) { - ZSTD_confirmRepcodesAndEntropyTables(zc); + ZSTD_blockState_confirmRepcodesAndEntropyTables(&zc->blockState); return cSize; } } } - } + } /* if (bss == ZSTDbss_compress)*/ DEBUGLOG(6, "Resorting to ZSTD_noCompressBlock()"); /* Superblock compression failed, attempt to emit a single no compress block. @@ -2554,9 +4309,9 @@ static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, void const* ip, void const* iend) { - if (ZSTD_window_needOverflowCorrection(ms->window, iend)) { - U32 const maxDist = (U32)1 << params->cParams.windowLog; - U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy); + U32 const cycleLog = ZSTD_cycleLog(params->cParams.chainLog, params->cParams.strategy); + U32 const maxDist = (U32)1 << params->cParams.windowLog; + if (ZSTD_window_needOverflowCorrection(ms->window, cycleLog, maxDist, ms->loadedDictEnd, ip, iend)) { U32 const correction = ZSTD_window_correctOverflow(&ms->window, cycleLog, maxDist, ip); ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30); ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30); @@ -2577,9 +4332,9 @@ static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, * All blocks will be terminated, all input will be consumed. * Function will issue an error if there is not enough `dstCapacity` to hold the compressed content. * Frame is supposed already started (header already produced) -* @return : compressed size, or an error code +* @return : compressed size, or an error code */ -static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, +static size_t ZSTD_compress_frameChunk(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, U32 lastFrameChunk) @@ -2593,7 +4348,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, assert(cctx->appliedParams.cParams.windowLog <= ZSTD_WINDOWLOG_MAX); - DEBUGLOG(5, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize); + DEBUGLOG(4, "ZSTD_compress_frameChunk (blockSize=%u)", (unsigned)blockSize); if (cctx->appliedParams.fParams.checksumFlag && srcSize) XXH64_update(&cctx->xxhState, src, srcSize); @@ -2601,7 +4356,9 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, ZSTD_matchState_t* const ms = &cctx->blockState.matchState; U32 const lastBlock = lastFrameChunk & (blockSize >= remaining); - RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE, + /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding + * additional 1. We need to revisit and change this logic to be more consistent */ + RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize + MIN_CBLOCK_SIZE + 1, dstSize_tooSmall, "not enough space to store compressed block"); if (remaining < blockSize) blockSize = remaining; @@ -2609,6 +4366,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, ZSTD_overflowCorrectIfNeeded( ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize); ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState); + ZSTD_window_enforceMaxDist(&ms->window, ip, maxDist, &ms->loadedDictEnd, &ms->dictMatchState); /* Ensure hash/chain table insertion resumes no sooner than lowlimit */ if (ms->nextToUpdate < ms->window.lowLimit) ms->nextToUpdate = ms->window.lowLimit; @@ -2619,6 +4377,10 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_targetCBlockSize failed"); assert(cSize > 0); assert(cSize <= blockSize + ZSTD_blockHeaderSize); + } else if (ZSTD_blockSplitterEnabled(&cctx->appliedParams)) { + cSize = ZSTD_compressBlock_splitBlock(cctx, op, dstCapacity, ip, blockSize, lastBlock); + FORWARD_IF_ERROR(cSize, "ZSTD_compressBlock_splitBlock failed"); + assert(cSize > 0 || cctx->seqCollector.collectSequences == 1); } else { cSize = ZSTD_compressBlock_internal(cctx, op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize, @@ -2635,7 +4397,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx, MEM_writeLE24(op, cBlockHeader); cSize += ZSTD_blockHeaderSize; } - } + } /* if (ZSTD_useTargetCBlockSize(&cctx->appliedParams))*/ ip += blockSize; @@ -2673,7 +4435,6 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity, "dst buf is too small to fit worst-case frame header size."); DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u", !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode); - if (params->format == ZSTD_f_zstd1) { MEM_writeLE32(dst, ZSTD_MAGICNUMBER); pos = 4; @@ -2682,7 +4443,9 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity, if (!singleSegment) op[pos++] = windowLogByte; switch(dictIDSizeCode) { - default: assert(0); /* impossible */ + default: + assert(0); /* impossible */ + ZSTD_FALLTHROUGH; case 0 : break; case 1 : op[pos] = (BYTE)(dictID); pos++; break; case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break; @@ -2690,7 +4453,9 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity, } switch(fcsCode) { - default: assert(0); /* impossible */ + default: + assert(0); /* impossible */ + ZSTD_FALLTHROUGH; case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break; case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break; case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break; @@ -2699,6 +4464,26 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity, return pos; } +/* ZSTD_writeSkippableFrame_advanced() : + * Writes out a skippable frame with the specified magic number variant (16 are supported), + * from ZSTD_MAGIC_SKIPPABLE_START to ZSTD_MAGIC_SKIPPABLE_START+15, and the desired source data. + * + * Returns the total number of bytes written, or a ZSTD error code. + */ +size_t ZSTD_writeSkippableFrame(void* dst, size_t dstCapacity, + const void* src, size_t srcSize, unsigned magicVariant) { + BYTE* op = (BYTE*)dst; + RETURN_ERROR_IF(dstCapacity < srcSize + ZSTD_SKIPPABLEHEADERSIZE /* Skippable frame overhead */, + dstSize_tooSmall, "Not enough room for skippable frame"); + RETURN_ERROR_IF(srcSize > (unsigned)0xFFFFFFFF, srcSize_wrong, "Src size too large for skippable frame"); + RETURN_ERROR_IF(magicVariant > 15, parameter_outOfBound, "Skippable frame magic number variant not supported"); + + MEM_writeLE32(op, (U32)(ZSTD_MAGIC_SKIPPABLE_START + magicVariant)); + MEM_writeLE32(op+4, (U32)srcSize); + ZSTD_memcpy(op+8, src, srcSize); + return srcSize + ZSTD_SKIPPABLEHEADERSIZE; +} + /* ZSTD_writeLastEmptyBlock() : * output an empty Block with end-of-frame mark to complete a frame * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h)) @@ -2718,13 +4503,14 @@ size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSe { RETURN_ERROR_IF(cctx->stage != ZSTDcs_init, stage_wrong, "wrong cctx stage"); - RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm, + RETURN_ERROR_IF(cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable, parameter_unsupported, "incompatible with ldm"); cctx->externSeqStore.seq = seq; cctx->externSeqStore.size = nbSeq; cctx->externSeqStore.capacity = nbSeq; cctx->externSeqStore.pos = 0; + cctx->externSeqStore.posInSequence = 0; return 0; } @@ -2754,11 +4540,12 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx, if (!srcSize) return fhSize; /* do not generate an empty block if no input */ - if (!ZSTD_window_update(&ms->window, src, srcSize)) { + if (!ZSTD_window_update(&ms->window, src, srcSize, ms->forceNonContiguous)) { + ms->forceNonContiguous = 0; ms->nextToUpdate = ms->window.dictLimit; } - if (cctx->appliedParams.ldmParams.enableLdm) { - ZSTD_window_update(&cctx->ldmState.window, src, srcSize); + if (cctx->appliedParams.ldmParams.enableLdm == ZSTD_ps_enable) { + ZSTD_window_update(&cctx->ldmState.window, src, srcSize, /* forceNonContiguous */ 0); } if (!frame) { @@ -2802,7 +4589,7 @@ size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx) { ZSTD_compressionParameters const cParams = cctx->appliedParams.cParams; assert(!ZSTD_checkCParams(cParams)); - return MIN (ZSTD_BLOCKSIZE_MAX, (U32)1 << cParams.windowLog); + return MIN(cctx->appliedParams.maxBlockSize, (size_t)1 << cParams.windowLog); } size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) @@ -2822,63 +4609,106 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms, ZSTD_cwksp* ws, ZSTD_CCtx_params const* params, const void* src, size_t srcSize, - ZSTD_dictTableLoadMethod_e dtlm) + ZSTD_dictTableLoadMethod_e dtlm, + ZSTD_tableFillPurpose_e tfp) { const BYTE* ip = (const BYTE*) src; const BYTE* const iend = ip + srcSize; + int const loadLdmDict = params->ldmParams.enableLdm == ZSTD_ps_enable && ls != NULL; - ZSTD_window_update(&ms->window, src, srcSize); - ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base); + /* Assert that the ms params match the params we're being given */ + ZSTD_assertEqualCParams(params->cParams, ms->cParams); - if (params->ldmParams.enableLdm && ls != NULL) { - ZSTD_window_update(&ls->window, src, srcSize); - ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base); - } + { /* Ensure large dictionaries can't cause index overflow */ - /* Assert that we the ms params match the params we're being given */ - ZSTD_assertEqualCParams(params->cParams, ms->cParams); + /* Allow the dictionary to set indices up to exactly ZSTD_CURRENT_MAX. + * Dictionaries right at the edge will immediately trigger overflow + * correction, but I don't want to insert extra constraints here. + */ + U32 maxDictSize = ZSTD_CURRENT_MAX - ZSTD_WINDOW_START_INDEX; + + int const CDictTaggedIndices = ZSTD_CDictIndicesAreTagged(¶ms->cParams); + if (CDictTaggedIndices && tfp == ZSTD_tfp_forCDict) { + /* Some dictionary matchfinders in zstd use "short cache", + * which treats the lower ZSTD_SHORT_CACHE_TAG_BITS of each + * CDict hashtable entry as a tag rather than as part of an index. + * When short cache is used, we need to truncate the dictionary + * so that its indices don't overlap with the tag. */ + U32 const shortCacheMaxDictSize = (1u << (32 - ZSTD_SHORT_CACHE_TAG_BITS)) - ZSTD_WINDOW_START_INDEX; + maxDictSize = MIN(maxDictSize, shortCacheMaxDictSize); + assert(!loadLdmDict); + } - if (srcSize <= HASH_READ_SIZE) return 0; + /* If the dictionary is too large, only load the suffix of the dictionary. */ + if (srcSize > maxDictSize) { + ip = iend - maxDictSize; + src = ip; + srcSize = maxDictSize; + } } - while (iend - ip > HASH_READ_SIZE) { - size_t const remaining = (size_t)(iend - ip); - size_t const chunk = MIN(remaining, ZSTD_CHUNKSIZE_MAX); - const BYTE* const ichunk = ip + chunk; + if (srcSize > ZSTD_CHUNKSIZE_MAX) { + /* We must have cleared our windows when our source is this large. */ + assert(ZSTD_window_isEmpty(ms->window)); + if (loadLdmDict) assert(ZSTD_window_isEmpty(ls->window)); + } - ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, ichunk); + DEBUGLOG(4, "ZSTD_loadDictionaryContent(): useRowMatchFinder=%d", (int)params->useRowMatchFinder); + ZSTD_window_update(&ms->window, src, srcSize, /* forceNonContiguous */ 0); + ms->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ms->window.base); + ms->forceNonContiguous = params->deterministicRefPrefix; - if (params->ldmParams.enableLdm && ls != NULL) - ZSTD_ldm_fillHashTable(ls, (const BYTE*)src, (const BYTE*)src + srcSize, ¶ms->ldmParams); + if (loadLdmDict) { + ZSTD_window_update(&ls->window, src, srcSize, /* forceNonContiguous */ 0); + ls->loadedDictEnd = params->forceWindow ? 0 : (U32)(iend - ls->window.base); + } - switch(params->cParams.strategy) - { - case ZSTD_fast: - ZSTD_fillHashTable(ms, ichunk, dtlm); - break; - case ZSTD_dfast: - ZSTD_fillDoubleHashTable(ms, ichunk, dtlm); - break; + if (srcSize <= HASH_READ_SIZE) return 0; - case ZSTD_greedy: - case ZSTD_lazy: - case ZSTD_lazy2: - if (chunk >= HASH_READ_SIZE) - ZSTD_insertAndFindFirstIndex(ms, ichunk-HASH_READ_SIZE); - break; + ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, iend); - case ZSTD_btlazy2: /* we want the dictionary table fully sorted */ - case ZSTD_btopt: - case ZSTD_btultra: - case ZSTD_btultra2: - if (chunk >= HASH_READ_SIZE) - ZSTD_updateTree(ms, ichunk-HASH_READ_SIZE, ichunk); - break; + if (loadLdmDict) + ZSTD_ldm_fillHashTable(ls, ip, iend, ¶ms->ldmParams); - default: - assert(0); /* not possible : not a valid strategy id */ + switch(params->cParams.strategy) + { + case ZSTD_fast: + ZSTD_fillHashTable(ms, iend, dtlm, tfp); + break; + case ZSTD_dfast: + ZSTD_fillDoubleHashTable(ms, iend, dtlm, tfp); + break; + + case ZSTD_greedy: + case ZSTD_lazy: + case ZSTD_lazy2: + assert(srcSize >= HASH_READ_SIZE); + if (ms->dedicatedDictSearch) { + assert(ms->chainTable != NULL); + ZSTD_dedicatedDictSearch_lazy_loadDictionary(ms, iend-HASH_READ_SIZE); + } else { + assert(params->useRowMatchFinder != ZSTD_ps_auto); + if (params->useRowMatchFinder == ZSTD_ps_enable) { + size_t const tagTableSize = ((size_t)1 << params->cParams.hashLog) * sizeof(U16); + ZSTD_memset(ms->tagTable, 0, tagTableSize); + ZSTD_row_update(ms, iend-HASH_READ_SIZE); + DEBUGLOG(4, "Using row-based hash table for lazy dict"); + } else { + ZSTD_insertAndFindFirstIndex(ms, iend-HASH_READ_SIZE); + DEBUGLOG(4, "Using chain-based hash table for lazy dict"); + } } + break; - ip = ichunk; + case ZSTD_btlazy2: /* we want the dictionary table fully sorted */ + case ZSTD_btopt: + case ZSTD_btultra: + case ZSTD_btultra2: + assert(srcSize >= HASH_READ_SIZE); + ZSTD_updateTree(ms, iend-HASH_READ_SIZE, iend); + break; + + default: + assert(0); /* not possible : not a valid strategy id */ } ms->nextToUpdate = (U32)(iend - ms->window.base); @@ -2887,22 +4717,28 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms, /* Dictionaries that assign zero probability to symbols that show up causes problems - when FSE encoding. Refuse dictionaries that assign zero probability to symbols - that we may encounter during compression. - NOTE: This behavior is not standard and could be improved in the future. */ -static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) { + * when FSE encoding. Mark dictionaries with zero probability symbols as FSE_repeat_check + * and only dictionaries with 100% valid symbols can be assumed valid. + */ +static FSE_repeat ZSTD_dictNCountRepeat(short* normalizedCounter, unsigned dictMaxSymbolValue, unsigned maxSymbolValue) +{ U32 s; - RETURN_ERROR_IF(dictMaxSymbolValue < maxSymbolValue, dictionary_corrupted, "dict fse tables don't have all symbols"); + if (dictMaxSymbolValue < maxSymbolValue) { + return FSE_repeat_check; + } for (s = 0; s <= maxSymbolValue; ++s) { - RETURN_ERROR_IF(normalizedCounter[s] == 0, dictionary_corrupted, "dict fse tables don't have all symbols"); + if (normalizedCounter[s] == 0) { + return FSE_repeat_check; + } } - return 0; + return FSE_repeat_valid; } size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, - short* offcodeNCount, unsigned* offcodeMaxValue, const void* const dict, size_t dictSize) { + short offcodeNCount[MaxOff+1]; + unsigned offcodeMaxValue = MaxOff; const BYTE* dictPtr = (const BYTE*)dict; /* skip magic num and dict ID */ const BYTE* const dictEnd = dictPtr + dictSize; dictPtr += 8; @@ -2924,16 +4760,16 @@ size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, } { unsigned offcodeLog; - size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr); + size_t const offcodeHeaderSize = FSE_readNCount(offcodeNCount, &offcodeMaxValue, &offcodeLog, dictPtr, dictEnd-dictPtr); RETURN_ERROR_IF(FSE_isError(offcodeHeaderSize), dictionary_corrupted, ""); RETURN_ERROR_IF(offcodeLog > OffFSELog, dictionary_corrupted, ""); - /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */ /* fill all offset symbols to avoid garbage at end of table */ RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( bs->entropy.fse.offcodeCTable, offcodeNCount, MaxOff, offcodeLog, workspace, HUF_WORKSPACE_SIZE)), dictionary_corrupted, ""); + /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */ dictPtr += offcodeHeaderSize; } @@ -2942,13 +4778,12 @@ size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, size_t const matchlengthHeaderSize = FSE_readNCount(matchlengthNCount, &matchlengthMaxValue, &matchlengthLog, dictPtr, dictEnd-dictPtr); RETURN_ERROR_IF(FSE_isError(matchlengthHeaderSize), dictionary_corrupted, ""); RETURN_ERROR_IF(matchlengthLog > MLFSELog, dictionary_corrupted, ""); - /* Every match length code must have non-zero probability */ - FORWARD_IF_ERROR( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML), ""); RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( bs->entropy.fse.matchlengthCTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog, workspace, HUF_WORKSPACE_SIZE)), dictionary_corrupted, ""); + bs->entropy.fse.matchlength_repeatMode = ZSTD_dictNCountRepeat(matchlengthNCount, matchlengthMaxValue, MaxML); dictPtr += matchlengthHeaderSize; } @@ -2957,13 +4792,12 @@ size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, size_t const litlengthHeaderSize = FSE_readNCount(litlengthNCount, &litlengthMaxValue, &litlengthLog, dictPtr, dictEnd-dictPtr); RETURN_ERROR_IF(FSE_isError(litlengthHeaderSize), dictionary_corrupted, ""); RETURN_ERROR_IF(litlengthLog > LLFSELog, dictionary_corrupted, ""); - /* Every literal length code must have non-zero probability */ - FORWARD_IF_ERROR( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL), ""); RETURN_ERROR_IF(FSE_isError(FSE_buildCTable_wksp( bs->entropy.fse.litlengthCTable, litlengthNCount, litlengthMaxValue, litlengthLog, workspace, HUF_WORKSPACE_SIZE)), dictionary_corrupted, ""); + bs->entropy.fse.litlength_repeatMode = ZSTD_dictNCountRepeat(litlengthNCount, litlengthMaxValue, MaxLL); dictPtr += litlengthHeaderSize; } @@ -2973,12 +4807,28 @@ size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace, bs->rep[2] = MEM_readLE32(dictPtr+8); dictPtr += 12; + { size_t const dictContentSize = (size_t)(dictEnd - dictPtr); + U32 offcodeMax = MaxOff; + if (dictContentSize <= ((U32)-1) - 128 KB) { + U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */ + offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */ + } + /* All offset values <= dictContentSize + 128 KB must be representable for a valid table */ + bs->entropy.fse.offcode_repeatMode = ZSTD_dictNCountRepeat(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)); + + /* All repCodes must be <= dictContentSize and != 0 */ + { U32 u; + for (u=0; u<3; u++) { + RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, ""); + RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, ""); + } } } + return dictPtr - (const BYTE*)dict; } /* Dictionary format : * See : - * https://github.com/facebook/zstd/blob/master/doc/zstd_compression_format.md#dictionary-format + * https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#dictionary-format */ /*! ZSTD_loadZstdDictionary() : * @return : dictID, or an error code @@ -2991,46 +4841,28 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs, ZSTD_CCtx_params const* params, const void* dict, size_t dictSize, ZSTD_dictTableLoadMethod_e dtlm, + ZSTD_tableFillPurpose_e tfp, void* workspace) { const BYTE* dictPtr = (const BYTE*)dict; const BYTE* const dictEnd = dictPtr + dictSize; - short offcodeNCount[MaxOff+1]; - unsigned offcodeMaxValue = MaxOff; size_t dictID; size_t eSize; - ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog))); assert(dictSize >= 8); assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY); dictID = params->fParams.noDictIDFlag ? 0 : MEM_readLE32(dictPtr + 4 /* skip magic number */ ); - eSize = ZSTD_loadCEntropy(bs, workspace, offcodeNCount, &offcodeMaxValue, dict, dictSize); + eSize = ZSTD_loadCEntropy(bs, workspace, dict, dictSize); FORWARD_IF_ERROR(eSize, "ZSTD_loadCEntropy failed"); dictPtr += eSize; - { size_t const dictContentSize = (size_t)(dictEnd - dictPtr); - U32 offcodeMax = MaxOff; - if (dictContentSize <= ((U32)-1) - 128 KB) { - U32 const maxOffset = (U32)dictContentSize + 128 KB; /* The maximum offset that must be supported */ - offcodeMax = ZSTD_highbit32(maxOffset); /* Calculate minimum offset code required to represent maxOffset */ - } - /* All offset values <= dictContentSize + 128 KB must be representable */ - FORWARD_IF_ERROR(ZSTD_checkDictNCount(offcodeNCount, offcodeMaxValue, MIN(offcodeMax, MaxOff)), ""); - /* All repCodes must be <= dictContentSize and != 0*/ - { U32 u; - for (u=0; u<3; u++) { - RETURN_ERROR_IF(bs->rep[u] == 0, dictionary_corrupted, ""); - RETURN_ERROR_IF(bs->rep[u] > dictContentSize, dictionary_corrupted, ""); - } } - - bs->entropy.fse.offcode_repeatMode = FSE_repeat_valid; - bs->entropy.fse.matchlength_repeatMode = FSE_repeat_valid; - bs->entropy.fse.litlength_repeatMode = FSE_repeat_valid; + { + size_t const dictContentSize = (size_t)(dictEnd - dictPtr); FORWARD_IF_ERROR(ZSTD_loadDictionaryContent( - ms, NULL, ws, params, dictPtr, dictContentSize, dtlm), ""); - return dictID; + ms, NULL, ws, params, dictPtr, dictContentSize, dtlm, tfp), ""); } + return dictID; } /** ZSTD_compress_insertDictionary() : @@ -3044,6 +4876,7 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs, const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType, ZSTD_dictTableLoadMethod_e dtlm, + ZSTD_tableFillPurpose_e tfp, void* workspace) { DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize); @@ -3056,13 +4889,13 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs, /* dict restricted modes */ if (dictContentType == ZSTD_dct_rawContent) - return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm); + return ZSTD_loadDictionaryContent(ms, ls, ws, params, dict, dictSize, dtlm, tfp); if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) { if (dictContentType == ZSTD_dct_auto) { DEBUGLOG(4, "raw content dictionary detected"); return ZSTD_loadDictionaryContent( - ms, ls, ws, params, dict, dictSize, dtlm); + ms, ls, ws, params, dict, dictSize, dtlm, tfp); } RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong, ""); assert(0); /* impossible */ @@ -3070,13 +4903,14 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs, /* dict as full zstd dictionary */ return ZSTD_loadZstdDictionary( - bs, ms, ws, params, dict, dictSize, dtlm, workspace); + bs, ms, ws, params, dict, dictSize, dtlm, tfp, workspace); } #define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB) -#define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6) +#define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6ULL) /*! ZSTD_compressBegin_internal() : + * Assumption : either @dict OR @cdict (or none) is non-NULL, never both * @return : 0, or an error code */ static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, @@ -3086,6 +4920,10 @@ static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params, U64 pledgedSrcSize, ZSTD_buffered_policy_e zbuff) { + size_t const dictContentSize = cdict ? cdict->dictContentSize : dictSize; +#if ZSTD_TRACE + cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0; +#endif DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params->cParams.windowLog); /* params are supposed to be fully validated at this point */ assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams))); @@ -3100,21 +4938,23 @@ static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx, return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff); } - FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, *params, pledgedSrcSize, + FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize, + dictContentSize, ZSTDcrp_makeClean, zbuff) , ""); { size_t const dictID = cdict ? ZSTD_compress_insertDictionary( cctx->blockState.prevCBlock, &cctx->blockState.matchState, &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, cdict->dictContent, - cdict->dictContentSize, dictContentType, dtlm, - cctx->entropyWorkspace) + cdict->dictContentSize, cdict->dictContentType, dtlm, + ZSTD_tfp_forCCtx, cctx->entropyWorkspace) : ZSTD_compress_insertDictionary( cctx->blockState.prevCBlock, &cctx->blockState.matchState, &cctx->ldmState, &cctx->workspace, &cctx->appliedParams, dict, dictSize, - dictContentType, dtlm, cctx->entropyWorkspace); + dictContentType, dtlm, ZSTD_tfp_forCCtx, cctx->entropyWorkspace); FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed"); assert(dictID <= UINT_MAX); cctx->dictID = (U32)dictID; + cctx->dictContentSize = dictContentSize; } return 0; } @@ -3143,19 +4983,21 @@ size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_parameters params, unsigned long long pledgedSrcSize) { - ZSTD_CCtx_params const cctxParams = - ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms); + ZSTD_CCtx_params cctxParams; + ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, ZSTD_NO_CLEVEL); return ZSTD_compressBegin_advanced_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL /*cdict*/, &cctxParams, pledgedSrcSize); } -size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel) +size_t +ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel) { - ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize); - ZSTD_CCtx_params const cctxParams = - ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms); + ZSTD_CCtx_params cctxParams; + { ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_noAttachDict); + ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel); + } DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize); return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL, &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered); @@ -3209,6 +5051,30 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity) return op-ostart; } +void ZSTD_CCtx_trace(ZSTD_CCtx* cctx, size_t extraCSize) +{ +#if ZSTD_TRACE + if (cctx->traceCtx && ZSTD_trace_compress_end != NULL) { + int const streaming = cctx->inBuffSize > 0 || cctx->outBuffSize > 0 || cctx->appliedParams.nbWorkers > 0; + ZSTD_Trace trace; + ZSTD_memset(&trace, 0, sizeof(trace)); + trace.version = ZSTD_VERSION_NUMBER; + trace.streaming = streaming; + trace.dictionaryID = cctx->dictID; + trace.dictionarySize = cctx->dictContentSize; + trace.uncompressedSize = cctx->consumedSrcSize; + trace.compressedSize = cctx->producedCSize + extraCSize; + trace.params = &cctx->appliedParams; + trace.cctx = cctx; + ZSTD_trace_compress_end(cctx->traceCtx, &trace); + } + cctx->traceCtx = 0; +#else + (void)cctx; + (void)extraCSize; +#endif +} + size_t ZSTD_compressEnd (ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) @@ -3231,26 +5097,10 @@ size_t ZSTD_compressEnd (ZSTD_CCtx* cctx, (unsigned)cctx->pledgedSrcSizePlusOne-1, (unsigned)cctx->consumedSrcSize); } + ZSTD_CCtx_trace(cctx, endResult); return cSize + endResult; } - -static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx, - void* dst, size_t dstCapacity, - const void* src, size_t srcSize, - const void* dict,size_t dictSize, - const ZSTD_parameters* params) -{ - ZSTD_CCtx_params const cctxParams = - ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params); - DEBUGLOG(4, "ZSTD_compress_internal"); - return ZSTD_compress_advanced_internal(cctx, - dst, dstCapacity, - src, srcSize, - dict, dictSize, - &cctxParams); -} - size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, @@ -3259,11 +5109,12 @@ size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx, { DEBUGLOG(4, "ZSTD_compress_advanced"); FORWARD_IF_ERROR(ZSTD_checkCParams(params.cParams), ""); - return ZSTD_compress_internal(cctx, - dst, dstCapacity, - src, srcSize, - dict, dictSize, - ¶ms); + ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, ZSTD_NO_CLEVEL); + return ZSTD_compress_advanced_internal(cctx, + dst, dstCapacity, + src, srcSize, + dict, dictSize, + &cctx->simpleApiParams); } /* Internal */ @@ -3287,11 +5138,13 @@ size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel) { - ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0); - ZSTD_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, ¶ms); + { + ZSTD_parameters const params = ZSTD_getParams_internal(compressionLevel, srcSize, dict ? dictSize : 0, ZSTD_cpm_noAttachDict); + assert(params.fParams.contentSizeFlag == 1); + ZSTD_CCtxParams_init_internal(&cctx->simpleApiParams, ¶ms, (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT: compressionLevel); + } DEBUGLOG(4, "ZSTD_compress_usingDict (srcSize=%u)", (unsigned)srcSize); - assert(params.fParams.contentSizeFlag == 1); - return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctxParams); + return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctx->simpleApiParams); } size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx, @@ -3309,10 +5162,17 @@ size_t ZSTD_compress(void* dst, size_t dstCapacity, int compressionLevel) { size_t result; +#if ZSTD_COMPRESS_HEAPMODE + ZSTD_CCtx* cctx = ZSTD_createCCtx(); + RETURN_ERROR_IF(!cctx, memory_allocation, "ZSTD_createCCtx failed"); + result = ZSTD_compressCCtx(cctx, dst, dstCapacity, src, srcSize, compressionLevel); + ZSTD_freeCCtx(cctx); +#else ZSTD_CCtx ctxBody; ZSTD_initCCtx(&ctxBody, ZSTD_defaultCMem); result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel); ZSTD_freeCCtxContent(&ctxBody); /* can't free ctxBody itself, as it's on stack; free only heap content */ +#endif return result; } @@ -3328,14 +5188,17 @@ size_t ZSTD_estimateCDictSize_advanced( DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict)); return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) - + ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0) + /* enableDedicatedDictSearch == 1 ensures that CDict estimation will not be too small + * in case we are using DDS with row-hash. */ + + ZSTD_sizeof_matchState(&cParams, ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams), + /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *)))); } size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel) { - ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize); + ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy); } @@ -3353,20 +5216,22 @@ static size_t ZSTD_initCDict_internal( const void* dictBuffer, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType, - ZSTD_compressionParameters cParams) + ZSTD_CCtx_params params) { DEBUGLOG(3, "ZSTD_initCDict_internal (dictContentType:%u)", (unsigned)dictContentType); - assert(!ZSTD_checkCParams(cParams)); - cdict->matchState.cParams = cParams; + assert(!ZSTD_checkCParams(params.cParams)); + cdict->matchState.cParams = params.cParams; + cdict->matchState.dedicatedDictSearch = params.enableDedicatedDictSearch; if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) { cdict->dictContent = dictBuffer; } else { void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*))); RETURN_ERROR_IF(!internalBuffer, memory_allocation, "NULL pointer!"); cdict->dictContent = internalBuffer; - memcpy(internalBuffer, dictBuffer, dictSize); + ZSTD_memcpy(internalBuffer, dictBuffer, dictSize); } cdict->dictContentSize = dictSize; + cdict->dictContentType = dictContentType; cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE); @@ -3376,22 +5241,20 @@ static size_t ZSTD_initCDict_internal( FORWARD_IF_ERROR(ZSTD_reset_matchState( &cdict->matchState, &cdict->workspace, - &cParams, + ¶ms.cParams, + params.useRowMatchFinder, ZSTDcrp_makeClean, ZSTDirp_reset, ZSTD_resetTarget_CDict), ""); /* (Maybe) load the dictionary * Skips loading the dictionary if it is < 8 bytes. */ - { ZSTD_CCtx_params params; - memset(¶ms, 0, sizeof(params)); - params.compressionLevel = ZSTD_CLEVEL_DEFAULT; + { params.compressionLevel = ZSTD_CLEVEL_DEFAULT; params.fParams.contentSizeFlag = 1; - params.cParams = cParams; { size_t const dictID = ZSTD_compress_insertDictionary( &cdict->cBlockState, &cdict->matchState, NULL, &cdict->workspace, ¶ms, cdict->dictContent, cdict->dictContentSize, - dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace); + dictContentType, ZSTD_dtlm_full, ZSTD_tfp_forCDict, cdict->entropyWorkspace); FORWARD_IF_ERROR(dictID, "ZSTD_compress_insertDictionary failed"); assert(dictID <= (size_t)(U32)-1); cdict->dictID = (U32)dictID; @@ -3401,66 +5264,129 @@ static size_t ZSTD_initCDict_internal( return 0; } -ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize, +static ZSTD_CDict* ZSTD_createCDict_advanced_internal(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, - ZSTD_dictContentType_e dictContentType, - ZSTD_compressionParameters cParams, ZSTD_customMem customMem) + ZSTD_compressionParameters cParams, + ZSTD_paramSwitch_e useRowMatchFinder, + U32 enableDedicatedDictSearch, + ZSTD_customMem customMem) { - DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (unsigned)dictContentType); - if (!customMem.customAlloc ^ !customMem.customFree) return NULL; + if ((!customMem.customAlloc) ^ (!customMem.customFree)) return NULL; { size_t const workspaceSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) + - ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0) + + ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, enableDedicatedDictSearch, /* forCCtx */ 0) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*)))); - void* const workspace = ZSTD_malloc(workspaceSize, customMem); + void* const workspace = ZSTD_customMalloc(workspaceSize, customMem); ZSTD_cwksp ws; ZSTD_CDict* cdict; if (!workspace) { - ZSTD_free(workspace, customMem); + ZSTD_customFree(workspace, customMem); return NULL; } - ZSTD_cwksp_init(&ws, workspace, workspaceSize); + ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_dynamic_alloc); cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict)); assert(cdict != NULL); ZSTD_cwksp_move(&cdict->workspace, &ws); cdict->customMem = customMem; - cdict->compressionLevel = 0; /* signals advanced API usage */ + cdict->compressionLevel = ZSTD_NO_CLEVEL; /* signals advanced API usage */ + cdict->useRowMatchFinder = useRowMatchFinder; + return cdict; + } +} - if (ZSTD_isError( ZSTD_initCDict_internal(cdict, - dictBuffer, dictSize, - dictLoadMethod, dictContentType, - cParams) )) { - ZSTD_freeCDict(cdict); - return NULL; - } +ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize, + ZSTD_dictLoadMethod_e dictLoadMethod, + ZSTD_dictContentType_e dictContentType, + ZSTD_compressionParameters cParams, + ZSTD_customMem customMem) +{ + ZSTD_CCtx_params cctxParams; + ZSTD_memset(&cctxParams, 0, sizeof(cctxParams)); + ZSTD_CCtxParams_init(&cctxParams, 0); + cctxParams.cParams = cParams; + cctxParams.customMem = customMem; + return ZSTD_createCDict_advanced2( + dictBuffer, dictSize, + dictLoadMethod, dictContentType, + &cctxParams, customMem); +} + +ZSTD_CDict* ZSTD_createCDict_advanced2( + const void* dict, size_t dictSize, + ZSTD_dictLoadMethod_e dictLoadMethod, + ZSTD_dictContentType_e dictContentType, + const ZSTD_CCtx_params* originalCctxParams, + ZSTD_customMem customMem) +{ + ZSTD_CCtx_params cctxParams = *originalCctxParams; + ZSTD_compressionParameters cParams; + ZSTD_CDict* cdict; - return cdict; + DEBUGLOG(3, "ZSTD_createCDict_advanced2, mode %u", (unsigned)dictContentType); + if (!customMem.customAlloc ^ !customMem.customFree) return NULL; + + if (cctxParams.enableDedicatedDictSearch) { + cParams = ZSTD_dedicatedDictSearch_getCParams( + cctxParams.compressionLevel, dictSize); + ZSTD_overrideCParams(&cParams, &cctxParams.cParams); + } else { + cParams = ZSTD_getCParamsFromCCtxParams( + &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); + } + + if (!ZSTD_dedicatedDictSearch_isSupported(&cParams)) { + /* Fall back to non-DDSS params */ + cctxParams.enableDedicatedDictSearch = 0; + cParams = ZSTD_getCParamsFromCCtxParams( + &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); } + + DEBUGLOG(3, "ZSTD_createCDict_advanced2: DDS: %u", cctxParams.enableDedicatedDictSearch); + cctxParams.cParams = cParams; + cctxParams.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(cctxParams.useRowMatchFinder, &cParams); + + cdict = ZSTD_createCDict_advanced_internal(dictSize, + dictLoadMethod, cctxParams.cParams, + cctxParams.useRowMatchFinder, cctxParams.enableDedicatedDictSearch, + customMem); + + if (ZSTD_isError( ZSTD_initCDict_internal(cdict, + dict, dictSize, + dictLoadMethod, dictContentType, + cctxParams) )) { + ZSTD_freeCDict(cdict); + return NULL; + } + + return cdict; } ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel) { - ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize); - ZSTD_CDict* cdict = ZSTD_createCDict_advanced(dict, dictSize, + ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); + ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dct_auto, cParams, ZSTD_defaultCMem); if (cdict) - cdict->compressionLevel = compressionLevel == 0 ? ZSTD_CLEVEL_DEFAULT : compressionLevel; + cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel; return cdict; } ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel) { - ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize); - return ZSTD_createCDict_advanced(dict, dictSize, + ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize, ZSTD_cpm_createCDict); + ZSTD_CDict* const cdict = ZSTD_createCDict_advanced(dict, dictSize, ZSTD_dlm_byRef, ZSTD_dct_auto, cParams, ZSTD_defaultCMem); + if (cdict) + cdict->compressionLevel = (compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : compressionLevel; + return cdict; } size_t ZSTD_freeCDict(ZSTD_CDict* cdict) @@ -3470,7 +5396,7 @@ size_t ZSTD_freeCDict(ZSTD_CDict* cdict) int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict); ZSTD_cwksp_free(&cdict->workspace, cMem); if (!cdictInWorkspace) { - ZSTD_free(cdict, cMem); + ZSTD_customFree(cdict, cMem); } return 0; } @@ -3496,19 +5422,22 @@ const ZSTD_CDict* ZSTD_initStaticCDict( ZSTD_dictContentType_e dictContentType, ZSTD_compressionParameters cParams) { - size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0); + ZSTD_paramSwitch_e const useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams); + /* enableDedicatedDictSearch == 1 ensures matchstate is not too small in case this CDict will be used for DDS + row hash */ + size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, useRowMatchFinder, /* enableDedicatedDictSearch */ 1, /* forCCtx */ 0); size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*)))) + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) + matchStateSize; ZSTD_CDict* cdict; + ZSTD_CCtx_params params; if ((size_t)workspace & 7) return NULL; /* 8-aligned */ { ZSTD_cwksp ws; - ZSTD_cwksp_init(&ws, workspace, workspaceSize); + ZSTD_cwksp_init(&ws, workspace, workspaceSize, ZSTD_cwksp_static_alloc); cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict)); if (cdict == NULL) return NULL; ZSTD_cwksp_move(&cdict->workspace, &ws); @@ -3518,10 +5447,15 @@ const ZSTD_CDict* ZSTD_initStaticCDict( (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize)); if (workspaceSize < neededSize) return NULL; + ZSTD_CCtxParams_init(¶ms, 0); + params.cParams = cParams; + params.useRowMatchFinder = useRowMatchFinder; + cdict->useRowMatchFinder = useRowMatchFinder; + if (ZSTD_isError( ZSTD_initCDict_internal(cdict, dict, dictSize, dictLoadMethod, dictContentType, - cParams) )) + params) )) return NULL; return cdict; @@ -3533,61 +5467,98 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict) return cdict->matchState.cParams; } -/* ZSTD_compressBegin_usingCDict_advanced() : - * cdict must be != NULL */ -size_t ZSTD_compressBegin_usingCDict_advanced( +/*! ZSTD_getDictID_fromCDict() : + * Provides the dictID of the dictionary loaded into `cdict`. + * If @return == 0, the dictionary is not conformant to Zstandard specification, or empty. + * Non-conformant dictionaries can still be loaded, but as content-only dictionaries. */ +unsigned ZSTD_getDictID_fromCDict(const ZSTD_CDict* cdict) +{ + if (cdict==NULL) return 0; + return cdict->dictID; +} + +/* ZSTD_compressBegin_usingCDict_internal() : + * Implementation of various ZSTD_compressBegin_usingCDict* functions. + */ +static size_t ZSTD_compressBegin_usingCDict_internal( ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize) { - DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_advanced"); + ZSTD_CCtx_params cctxParams; + DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_internal"); RETURN_ERROR_IF(cdict==NULL, dictionary_wrong, "NULL pointer!"); - { ZSTD_CCtx_params params = cctx->requestedParams; + /* Initialize the cctxParams from the cdict */ + { + ZSTD_parameters params; + params.fParams = fParams; params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN - || cdict->compressionLevel == 0 ) - && (params.attachDictPref != ZSTD_dictForceLoad) ? + || cdict->compressionLevel == 0 ) ? ZSTD_getCParamsFromCDict(cdict) : ZSTD_getCParams(cdict->compressionLevel, pledgedSrcSize, cdict->dictContentSize); - /* Increase window log to fit the entire dictionary and source if the - * source size is known. Limit the increase to 19, which is the - * window log for compression level 1 with the largest source size. - */ - if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) { - U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19); - U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1; - params.cParams.windowLog = MAX(params.cParams.windowLog, limitedSrcLog); - } - params.fParams = fParams; - return ZSTD_compressBegin_internal(cctx, - NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, - cdict, - ¶ms, pledgedSrcSize, - ZSTDb_not_buffered); + ZSTD_CCtxParams_init_internal(&cctxParams, ¶ms, cdict->compressionLevel); + } + /* Increase window log to fit the entire dictionary and source if the + * source size is known. Limit the increase to 19, which is the + * window log for compression level 1 with the largest source size. + */ + if (pledgedSrcSize != ZSTD_CONTENTSIZE_UNKNOWN) { + U32 const limitedSrcSize = (U32)MIN(pledgedSrcSize, 1U << 19); + U32 const limitedSrcLog = limitedSrcSize > 1 ? ZSTD_highbit32(limitedSrcSize - 1) + 1 : 1; + cctxParams.cParams.windowLog = MAX(cctxParams.cParams.windowLog, limitedSrcLog); } + return ZSTD_compressBegin_internal(cctx, + NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, + cdict, + &cctxParams, pledgedSrcSize, + ZSTDb_not_buffered); +} + + +/* ZSTD_compressBegin_usingCDict_advanced() : + * This function is DEPRECATED. + * cdict must be != NULL */ +size_t ZSTD_compressBegin_usingCDict_advanced( + ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, + ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize) +{ + return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, pledgedSrcSize); } /* ZSTD_compressBegin_usingCDict() : - * pledgedSrcSize=0 means "unknown" - * if pledgedSrcSize>0, it will enable contentSizeFlag */ + * cdict must be != NULL */ size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) { ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; - DEBUGLOG(4, "ZSTD_compressBegin_usingCDict : dictIDFlag == %u", !fParams.noDictIDFlag); - return ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN); + return ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, ZSTD_CONTENTSIZE_UNKNOWN); } -size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx, +/*! ZSTD_compress_usingCDict_internal(): + * Implementation of various ZSTD_compress_usingCDict* functions. + */ +static size_t ZSTD_compress_usingCDict_internal(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams) { - FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */ + FORWARD_IF_ERROR(ZSTD_compressBegin_usingCDict_internal(cctx, cdict, fParams, srcSize), ""); /* will check if cdict != NULL */ return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize); } +/*! ZSTD_compress_usingCDict_advanced(): + * This function is DEPRECATED. + */ +size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx, + void* dst, size_t dstCapacity, + const void* src, size_t srcSize, + const ZSTD_CDict* cdict, ZSTD_frameParameters fParams) +{ + return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams); +} + /*! ZSTD_compress_usingCDict() : * Compression using a digested Dictionary. * Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times. @@ -3599,7 +5570,7 @@ size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict) { ZSTD_frameParameters const fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ }; - return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, fParams); + return ZSTD_compress_usingCDict_internal(cctx, dst, dstCapacity, src, srcSize, cdict, fParams); } @@ -3640,32 +5611,12 @@ size_t ZSTD_CStreamOutSize(void) return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ; } -static size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx, - const void* const dict, size_t const dictSize, ZSTD_dictContentType_e const dictContentType, - const ZSTD_CDict* const cdict, - ZSTD_CCtx_params params, unsigned long long const pledgedSrcSize) +static ZSTD_cParamMode_e ZSTD_getCParamMode(ZSTD_CDict const* cdict, ZSTD_CCtx_params const* params, U64 pledgedSrcSize) { - DEBUGLOG(4, "ZSTD_resetCStream_internal"); - /* Finalize the compression parameters */ - params.cParams = ZSTD_getCParamsFromCCtxParams(¶ms, pledgedSrcSize, dictSize); - /* params are supposed to be fully validated at this point */ - assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams))); - assert(!((dict) && (cdict))); /* either dict or cdict, not both */ - - FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx, - dict, dictSize, dictContentType, ZSTD_dtlm_fast, - cdict, - ¶ms, pledgedSrcSize, - ZSTDb_buffered) , ""); - - cctx->inToCompress = 0; - cctx->inBuffPos = 0; - cctx->inBuffTarget = cctx->blockSize - + (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 */ - cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0; - cctx->streamStage = zcss_load; - cctx->frameEnded = 0; - return 0; /* ready to go */ + if (cdict != NULL && ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) + return ZSTD_cpm_attachDict; + else + return ZSTD_cpm_noAttachDict; } /* ZSTD_resetCStream(): @@ -3749,7 +5700,7 @@ size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) , ""); FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) , ""); FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) , ""); - zcs->requestedParams = ZSTD_assignParamsToCCtxParams(&zcs->requestedParams, ¶ms); + ZSTD_CCtxParams_setZstdParams(&zcs->requestedParams, ¶ms); FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) , ""); return 0; } @@ -3791,36 +5742,54 @@ size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel) static size_t ZSTD_nextInputSizeHint(const ZSTD_CCtx* cctx) { - size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos; - if (hintInSize==0) hintInSize = cctx->blockSize; - return hintInSize; + if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { + return cctx->blockSize - cctx->stableIn_notConsumed; + } + assert(cctx->appliedParams.inBufferMode == ZSTD_bm_buffered); + { size_t hintInSize = cctx->inBuffTarget - cctx->inBuffPos; + if (hintInSize==0) hintInSize = cctx->blockSize; + return hintInSize; + } } /** ZSTD_compressStream_generic(): * internal function for all *compressStream*() variants - * non-static, because can be called from zstdmt_compress.c - * @return : hint size for next input */ + * @return : hint size for next input to complete ongoing block */ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input, ZSTD_EndDirective const flushMode) { - const char* const istart = (const char*)input->src; - const char* const iend = input->size != 0 ? istart + input->size : istart; - const char* ip = input->pos != 0 ? istart + input->pos : istart; - char* const ostart = (char*)output->dst; - char* const oend = output->size != 0 ? ostart + output->size : ostart; - char* op = output->pos != 0 ? ostart + output->pos : ostart; + const char* const istart = (assert(input != NULL), (const char*)input->src); + const char* const iend = (istart != NULL) ? istart + input->size : istart; + const char* ip = (istart != NULL) ? istart + input->pos : istart; + char* const ostart = (assert(output != NULL), (char*)output->dst); + char* const oend = (ostart != NULL) ? ostart + output->size : ostart; + char* op = (ostart != NULL) ? ostart + output->pos : ostart; U32 someMoreWork = 1; /* check expectations */ - DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (unsigned)flushMode); - assert(zcs->inBuff != NULL); - assert(zcs->inBuffSize > 0); - assert(zcs->outBuff != NULL); - assert(zcs->outBuffSize > 0); - assert(output->pos <= output->size); + DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%i, srcSize = %zu", (int)flushMode, input->size - input->pos); + assert(zcs != NULL); + if (zcs->appliedParams.inBufferMode == ZSTD_bm_stable) { + assert(input->pos >= zcs->stableIn_notConsumed); + input->pos -= zcs->stableIn_notConsumed; + ip -= zcs->stableIn_notConsumed; + zcs->stableIn_notConsumed = 0; + } + if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) { + assert(zcs->inBuff != NULL); + assert(zcs->inBuffSize > 0); + } + if (zcs->appliedParams.outBufferMode == ZSTD_bm_buffered) { + assert(zcs->outBuff != NULL); + assert(zcs->outBuffSize > 0); + } + if (input->src == NULL) assert(input->size == 0); assert(input->pos <= input->size); + if (output->dst == NULL) assert(output->size == 0); + assert(output->pos <= output->size); + assert((U32)flushMode <= (U32)ZSTD_e_end); while (someMoreWork) { switch(zcs->streamStage) @@ -3830,7 +5799,8 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, case zcss_load: if ( (flushMode == ZSTD_e_end) - && ((size_t)(oend-op) >= ZSTD_compressBound(iend-ip)) /* enough dstCapacity */ + && ( (size_t)(oend-op) >= ZSTD_compressBound(iend-ip) /* Enough output space */ + || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) /* OR we are allowed to return dstSizeTooSmall */ && (zcs->inBuffPos == 0) ) { /* shortcut to compression pass directly into output buffer */ size_t const cSize = ZSTD_compressEnd(zcs, @@ -3843,14 +5813,14 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only); someMoreWork = 0; break; } - /* complete loading into inBuffer */ - { size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos; + /* complete loading into inBuffer in buffered mode */ + if (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered) { + size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos; size_t const loaded = ZSTD_limitCopy( zcs->inBuff + zcs->inBuffPos, toLoad, ip, iend-ip); zcs->inBuffPos += loaded; - if (loaded != 0) - ip += loaded; + if (ip) ip += loaded; if ( (flushMode == ZSTD_e_continue) && (zcs->inBuffPos < zcs->inBuffTarget) ) { /* not enough input to fill full block : stop here */ @@ -3861,34 +5831,62 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, /* empty */ someMoreWork = 0; break; } + } else { + assert(zcs->appliedParams.inBufferMode == ZSTD_bm_stable); + if ( (flushMode == ZSTD_e_continue) + && ( (size_t)(iend - ip) < zcs->blockSize) ) { + /* can't compress a full block : stop here */ + zcs->stableIn_notConsumed = (size_t)(iend - ip); + ip = iend; /* pretend to have consumed input */ + someMoreWork = 0; break; + } + if ( (flushMode == ZSTD_e_flush) + && (ip == iend) ) { + /* empty */ + someMoreWork = 0; break; + } } /* compress current block (note : this stage cannot be stopped in the middle) */ DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode); - { void* cDst; + { int const inputBuffered = (zcs->appliedParams.inBufferMode == ZSTD_bm_buffered); + void* cDst; size_t cSize; - size_t const iSize = zcs->inBuffPos - zcs->inToCompress; size_t oSize = oend-op; - unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend); - if (oSize >= ZSTD_compressBound(iSize)) + size_t const iSize = inputBuffered ? zcs->inBuffPos - zcs->inToCompress + : MIN((size_t)(iend - ip), zcs->blockSize); + if (oSize >= ZSTD_compressBound(iSize) || zcs->appliedParams.outBufferMode == ZSTD_bm_stable) cDst = op; /* compress into output buffer, to skip flush stage */ else cDst = zcs->outBuff, oSize = zcs->outBuffSize; - cSize = lastBlock ? - ZSTD_compressEnd(zcs, cDst, oSize, - zcs->inBuff + zcs->inToCompress, iSize) : - ZSTD_compressContinue(zcs, cDst, oSize, - zcs->inBuff + zcs->inToCompress, iSize); - FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed"); - zcs->frameEnded = lastBlock; - /* prepare next block */ - zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize; - if (zcs->inBuffTarget > zcs->inBuffSize) - zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize; - DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u", - (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize); - if (!lastBlock) - assert(zcs->inBuffTarget <= zcs->inBuffSize); - zcs->inToCompress = zcs->inBuffPos; + if (inputBuffered) { + unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend); + cSize = lastBlock ? + ZSTD_compressEnd(zcs, cDst, oSize, + zcs->inBuff + zcs->inToCompress, iSize) : + ZSTD_compressContinue(zcs, cDst, oSize, + zcs->inBuff + zcs->inToCompress, iSize); + FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed"); + zcs->frameEnded = lastBlock; + /* prepare next block */ + zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize; + if (zcs->inBuffTarget > zcs->inBuffSize) + zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize; + DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u", + (unsigned)zcs->inBuffTarget, (unsigned)zcs->inBuffSize); + if (!lastBlock) + assert(zcs->inBuffTarget <= zcs->inBuffSize); + zcs->inToCompress = zcs->inBuffPos; + } else { /* !inputBuffered, hence ZSTD_bm_stable */ + unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip + iSize == iend); + cSize = lastBlock ? + ZSTD_compressEnd(zcs, cDst, oSize, ip, iSize) : + ZSTD_compressContinue(zcs, cDst, oSize, ip, iSize); + /* Consume the input prior to error checking to mirror buffered mode. */ + if (ip) ip += iSize; + FORWARD_IF_ERROR(cSize, "%s", lastBlock ? "ZSTD_compressEnd failed" : "ZSTD_compressContinue failed"); + zcs->frameEnded = lastBlock; + if (lastBlock) assert(ip == iend); + } if (cDst == op) { /* no need to flush */ op += cSize; if (zcs->frameEnded) { @@ -3902,9 +5900,10 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs, zcs->outBuffFlushedSize = 0; zcs->streamStage = zcss_flush; /* pass-through to flush stage */ } - /* fall-through */ + ZSTD_FALLTHROUGH; case zcss_flush: DEBUGLOG(5, "flush stage"); + assert(zcs->appliedParams.outBufferMode == ZSTD_bm_buffered); { size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize; size_t const flushed = ZSTD_limitCopy(op, (size_t)(oend-op), zcs->outBuff + zcs->outBuffFlushedSize, toFlush); @@ -3959,7 +5958,141 @@ size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuf return ZSTD_nextInputSizeHint_MTorST(zcs); } +/* After a compression call set the expected input/output buffer. + * This is validated at the start of the next compression call. + */ +static void +ZSTD_setBufferExpectations(ZSTD_CCtx* cctx, const ZSTD_outBuffer* output, const ZSTD_inBuffer* input) +{ + DEBUGLOG(5, "ZSTD_setBufferExpectations (for advanced stable in/out modes)"); + if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { + cctx->expectedInBuffer = *input; + } + if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) { + cctx->expectedOutBufferSize = output->size - output->pos; + } +} + +/* Validate that the input/output buffers match the expectations set by + * ZSTD_setBufferExpectations. + */ +static size_t ZSTD_checkBufferStability(ZSTD_CCtx const* cctx, + ZSTD_outBuffer const* output, + ZSTD_inBuffer const* input, + ZSTD_EndDirective endOp) +{ + if (cctx->appliedParams.inBufferMode == ZSTD_bm_stable) { + ZSTD_inBuffer const expect = cctx->expectedInBuffer; + if (expect.src != input->src || expect.pos != input->pos) + RETURN_ERROR(stabilityCondition_notRespected, "ZSTD_c_stableInBuffer enabled but input differs!"); + } + (void)endOp; + if (cctx->appliedParams.outBufferMode == ZSTD_bm_stable) { + size_t const outBufferSize = output->size - output->pos; + if (cctx->expectedOutBufferSize != outBufferSize) + RETURN_ERROR(stabilityCondition_notRespected, "ZSTD_c_stableOutBuffer enabled but output size differs!"); + } + return 0; +} + +static size_t ZSTD_CCtx_init_compressStream2(ZSTD_CCtx* cctx, + ZSTD_EndDirective endOp, + size_t inSize) +{ + ZSTD_CCtx_params params = cctx->requestedParams; + ZSTD_prefixDict const prefixDict = cctx->prefixDict; + FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */ + ZSTD_memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */ + assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */ + if (cctx->cdict && !cctx->localDict.cdict) { + /* Let the cdict's compression level take priority over the requested params. + * But do not take the cdict's compression level if the "cdict" is actually a localDict + * generated from ZSTD_initLocalDict(). + */ + params.compressionLevel = cctx->cdict->compressionLevel; + } + DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage"); + if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = inSize + 1; /* auto-determine pledgedSrcSize */ + + { size_t const dictSize = prefixDict.dict + ? prefixDict.dictSize + : (cctx->cdict ? cctx->cdict->dictContentSize : 0); + ZSTD_cParamMode_e const mode = ZSTD_getCParamMode(cctx->cdict, ¶ms, cctx->pledgedSrcSizePlusOne - 1); + params.cParams = ZSTD_getCParamsFromCCtxParams( + ¶ms, cctx->pledgedSrcSizePlusOne-1, + dictSize, mode); + } + + params.useBlockSplitter = ZSTD_resolveBlockSplitterMode(params.useBlockSplitter, ¶ms.cParams); + params.ldmParams.enableLdm = ZSTD_resolveEnableLdm(params.ldmParams.enableLdm, ¶ms.cParams); + params.useRowMatchFinder = ZSTD_resolveRowMatchFinderMode(params.useRowMatchFinder, ¶ms.cParams); + params.validateSequences = ZSTD_resolveExternalSequenceValidation(params.validateSequences); + params.maxBlockSize = ZSTD_resolveMaxBlockSize(params.maxBlockSize); + params.searchForExternalRepcodes = ZSTD_resolveExternalRepcodeSearch(params.searchForExternalRepcodes, params.compressionLevel); + +#ifdef ZSTD_MULTITHREAD + /* If external matchfinder is enabled, make sure to fail before checking job size (for consistency) */ + RETURN_ERROR_IF( + params.useSequenceProducer == 1 && params.nbWorkers >= 1, + parameter_combination_unsupported, + "External sequence producer isn't supported with nbWorkers >= 1" + ); + + if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) { + params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */ + } + if (params.nbWorkers > 0) { +#if ZSTD_TRACE + cctx->traceCtx = (ZSTD_trace_compress_begin != NULL) ? ZSTD_trace_compress_begin(cctx) : 0; +#endif + /* mt context creation */ + if (cctx->mtctx == NULL) { + DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u", + params.nbWorkers); + cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem, cctx->pool); + RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!"); + } + /* mt compression */ + DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers); + FORWARD_IF_ERROR( ZSTDMT_initCStream_internal( + cctx->mtctx, + prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, + cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , ""); + cctx->dictID = cctx->cdict ? cctx->cdict->dictID : 0; + cctx->dictContentSize = cctx->cdict ? cctx->cdict->dictContentSize : prefixDict.dictSize; + cctx->consumedSrcSize = 0; + cctx->producedCSize = 0; + cctx->streamStage = zcss_load; + cctx->appliedParams = params; + } else +#endif /* ZSTD_MULTITHREAD */ + { U64 const pledgedSrcSize = cctx->pledgedSrcSizePlusOne - 1; + assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams))); + FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx, + prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, ZSTD_dtlm_fast, + cctx->cdict, + ¶ms, pledgedSrcSize, + ZSTDb_buffered) , ""); + assert(cctx->appliedParams.nbWorkers == 0); + cctx->inToCompress = 0; + cctx->inBuffPos = 0; + if (cctx->appliedParams.inBufferMode == ZSTD_bm_buffered) { + /* 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 + */ + cctx->inBuffTarget = cctx->blockSize + (cctx->blockSize == pledgedSrcSize); + } else { + cctx->inBuffTarget = 0; + } + cctx->outBuffContentSize = cctx->outBuffFlushedSize = 0; + cctx->streamStage = zcss_load; + cctx->frameEnded = 0; + } + return 0; +} +/* @return provides a minimum amount of data remaining to be flushed from internal buffers + */ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input, @@ -3967,82 +6100,95 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx, { DEBUGLOG(5, "ZSTD_compressStream2, endOp=%u ", (unsigned)endOp); /* check conditions */ - RETURN_ERROR_IF(output->pos > output->size, GENERIC, "invalid buffer"); - RETURN_ERROR_IF(input->pos > input->size, GENERIC, "invalid buffer"); - assert(cctx!=NULL); + RETURN_ERROR_IF(output->pos > output->size, dstSize_tooSmall, "invalid output buffer"); + RETURN_ERROR_IF(input->pos > input->size, srcSize_wrong, "invalid input buffer"); + RETURN_ERROR_IF((U32)endOp > (U32)ZSTD_e_end, parameter_outOfBound, "invalid endDirective"); + assert(cctx != NULL); /* transparent initialization stage */ if (cctx->streamStage == zcss_init) { - ZSTD_CCtx_params params = cctx->requestedParams; - ZSTD_prefixDict const prefixDict = cctx->prefixDict; - FORWARD_IF_ERROR( ZSTD_initLocalDict(cctx) , ""); /* Init the local dict if present. */ - memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict)); /* single usage */ - assert(prefixDict.dict==NULL || cctx->cdict==NULL); /* only one can be set */ - DEBUGLOG(4, "ZSTD_compressStream2 : transparent init stage"); - if (endOp == ZSTD_e_end) cctx->pledgedSrcSizePlusOne = input->size + 1; /* auto-fix pledgedSrcSize */ - params.cParams = ZSTD_getCParamsFromCCtxParams( - &cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1, 0 /*dictSize*/); - - -#ifdef ZSTD_MULTITHREAD - if ((cctx->pledgedSrcSizePlusOne-1) <= ZSTDMT_JOBSIZE_MIN) { - params.nbWorkers = 0; /* do not invoke multi-threading when src size is too small */ - } - if (params.nbWorkers > 0) { - /* mt context creation */ - if (cctx->mtctx == NULL) { - DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u", - params.nbWorkers); - cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem); - RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation, "NULL pointer!"); + size_t const inputSize = input->size - input->pos; /* no obligation to start from pos==0 */ + size_t const totalInputSize = inputSize + cctx->stableIn_notConsumed; + if ( (cctx->requestedParams.inBufferMode == ZSTD_bm_stable) /* input is presumed stable, across invocations */ + && (endOp == ZSTD_e_continue) /* no flush requested, more input to come */ + && (totalInputSize < ZSTD_BLOCKSIZE_MAX) ) { /* not even reached one block yet */ + if (cctx->stableIn_notConsumed) { /* not the first time */ + /* check stable source guarantees */ + RETURN_ERROR_IF(input->src != cctx->expectedInBuffer.src, stabilityCondition_notRespected, "stableInBuffer condition not respected: wrong src pointer"); + RETURN_ERROR_IF(input->pos != cctx->expectedInBuffer.size, stabilityCondition_notRespected, "stableInBuffer condition not respected: externally modified pos"); } - /* mt compression */ - DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbWorkers=%u", params.nbWorkers); - FORWARD_IF_ERROR( ZSTDMT_initCStream_internal( - cctx->mtctx, - prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, - cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) , ""); - cctx->streamStage = zcss_load; - cctx->appliedParams.nbWorkers = params.nbWorkers; - } else -#endif - { FORWARD_IF_ERROR( ZSTD_resetCStream_internal(cctx, - prefixDict.dict, prefixDict.dictSize, prefixDict.dictContentType, - cctx->cdict, - params, cctx->pledgedSrcSizePlusOne-1) , ""); - assert(cctx->streamStage == zcss_load); - assert(cctx->appliedParams.nbWorkers == 0); - } } + /* pretend input was consumed, to give a sense forward progress */ + input->pos = input->size; + /* save stable inBuffer, for later control, and flush/end */ + cctx->expectedInBuffer = *input; + /* but actually input wasn't consumed, so keep track of position from where compression shall resume */ + cctx->stableIn_notConsumed += inputSize; + /* don't initialize yet, wait for the first block of flush() order, for better parameters adaptation */ + return ZSTD_FRAMEHEADERSIZE_MIN(cctx->requestedParams.format); /* at least some header to produce */ + } + FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, endOp, totalInputSize), "compressStream2 initialization failed"); + ZSTD_setBufferExpectations(cctx, output, input); /* Set initial buffer expectations now that we've initialized */ + } /* end of transparent initialization stage */ + FORWARD_IF_ERROR(ZSTD_checkBufferStability(cctx, output, input, endOp), "invalid buffers"); /* compression stage */ #ifdef ZSTD_MULTITHREAD if (cctx->appliedParams.nbWorkers > 0) { - int const forceMaxProgress = (endOp == ZSTD_e_flush || endOp == ZSTD_e_end); size_t flushMin; - assert(forceMaxProgress || endOp == ZSTD_e_continue /* Protection for a new flush type */); if (cctx->cParamsChanged) { ZSTDMT_updateCParams_whileCompressing(cctx->mtctx, &cctx->requestedParams); cctx->cParamsChanged = 0; } - do { + if (cctx->stableIn_notConsumed) { + assert(cctx->appliedParams.inBufferMode == ZSTD_bm_stable); + /* some early data was skipped - make it available for consumption */ + assert(input->pos >= cctx->stableIn_notConsumed); + input->pos -= cctx->stableIn_notConsumed; + cctx->stableIn_notConsumed = 0; + } + for (;;) { + size_t const ipos = input->pos; + size_t const opos = output->pos; flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp); + cctx->consumedSrcSize += (U64)(input->pos - ipos); + cctx->producedCSize += (U64)(output->pos - opos); if ( ZSTD_isError(flushMin) || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */ + if (flushMin == 0) + ZSTD_CCtx_trace(cctx, 0); ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only); } FORWARD_IF_ERROR(flushMin, "ZSTDMT_compressStream_generic failed"); - } while (forceMaxProgress && flushMin != 0 && output->pos < output->size); + + if (endOp == ZSTD_e_continue) { + /* We only require some progress with ZSTD_e_continue, not maximal progress. + * We're done if we've consumed or produced any bytes, or either buffer is + * full. + */ + if (input->pos != ipos || output->pos != opos || input->pos == input->size || output->pos == output->size) + break; + } else { + assert(endOp == ZSTD_e_flush || endOp == ZSTD_e_end); + /* We require maximal progress. We're done when the flush is complete or the + * output buffer is full. + */ + if (flushMin == 0 || output->pos == output->size) + break; + } + } DEBUGLOG(5, "completed ZSTD_compressStream2 delegating to ZSTDMT_compressStream_generic"); /* Either we don't require maximum forward progress, we've finished the * flush, or we are out of output space. */ - assert(!forceMaxProgress || flushMin == 0 || output->pos == output->size); + assert(endOp == ZSTD_e_continue || flushMin == 0 || output->pos == output->size); + ZSTD_setBufferExpectations(cctx, output, input); return flushMin; } -#endif +#endif /* ZSTD_MULTITHREAD */ FORWARD_IF_ERROR( ZSTD_compressStream_generic(cctx, output, input, endOp) , ""); DEBUGLOG(5, "completed ZSTD_compressStream2"); + ZSTD_setBufferExpectations(cctx, output, input); return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */ } @@ -4052,27 +6198,43 @@ size_t ZSTD_compressStream2_simpleArgs ( const void* src, size_t srcSize, size_t* srcPos, ZSTD_EndDirective endOp) { - ZSTD_outBuffer output = { dst, dstCapacity, *dstPos }; - ZSTD_inBuffer input = { src, srcSize, *srcPos }; + ZSTD_outBuffer output; + ZSTD_inBuffer input; + output.dst = dst; + output.size = dstCapacity; + output.pos = *dstPos; + input.src = src; + input.size = srcSize; + input.pos = *srcPos; /* ZSTD_compressStream2() will check validity of dstPos and srcPos */ - size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp); - *dstPos = output.pos; - *srcPos = input.pos; - return cErr; + { size_t const cErr = ZSTD_compressStream2(cctx, &output, &input, endOp); + *dstPos = output.pos; + *srcPos = input.pos; + return cErr; + } } size_t ZSTD_compress2(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize) { + ZSTD_bufferMode_e const originalInBufferMode = cctx->requestedParams.inBufferMode; + ZSTD_bufferMode_e const originalOutBufferMode = cctx->requestedParams.outBufferMode; DEBUGLOG(4, "ZSTD_compress2 (srcSize=%u)", (unsigned)srcSize); ZSTD_CCtx_reset(cctx, ZSTD_reset_session_only); + /* Enable stable input/output buffers. */ + cctx->requestedParams.inBufferMode = ZSTD_bm_stable; + cctx->requestedParams.outBufferMode = ZSTD_bm_stable; { size_t oPos = 0; size_t iPos = 0; size_t const result = ZSTD_compressStream2_simpleArgs(cctx, dst, dstCapacity, &oPos, src, srcSize, &iPos, ZSTD_e_end); + /* Reset to the original values. */ + cctx->requestedParams.inBufferMode = originalInBufferMode; + cctx->requestedParams.outBufferMode = originalOutBufferMode; + FORWARD_IF_ERROR(result, "ZSTD_compressStream2_simpleArgs failed"); if (result != 0) { /* compression not completed, due to lack of output space */ assert(oPos == dstCapacity); @@ -4083,22 +6245,512 @@ size_t ZSTD_compress2(ZSTD_CCtx* cctx, } } +/* ZSTD_validateSequence() : + * @offCode : is presumed to follow format required by ZSTD_storeSeq() + * @returns a ZSTD error code if sequence is not valid + */ +static size_t +ZSTD_validateSequence(U32 offCode, U32 matchLength, U32 minMatch, + size_t posInSrc, U32 windowLog, size_t dictSize, int useSequenceProducer) +{ + U32 const windowSize = 1u << windowLog; + /* posInSrc represents the amount of data the decoder would decode up to this point. + * As long as the amount of data decoded is less than or equal to window size, offsets may be + * larger than the total length of output decoded in order to reference the dict, even larger than + * window size. After output surpasses windowSize, we're limited to windowSize offsets again. + */ + size_t const offsetBound = posInSrc > windowSize ? (size_t)windowSize : posInSrc + (size_t)dictSize; + size_t const matchLenLowerBound = (minMatch == 3 || useSequenceProducer) ? 3 : 4; + RETURN_ERROR_IF(offCode > OFFSET_TO_OFFBASE(offsetBound), externalSequences_invalid, "Offset too large!"); + /* Validate maxNbSeq is large enough for the given matchLength and minMatch */ + RETURN_ERROR_IF(matchLength < matchLenLowerBound, externalSequences_invalid, "Matchlength too small for the minMatch"); + return 0; +} + +/* Returns an offset code, given a sequence's raw offset, the ongoing repcode array, and whether litLength == 0 */ +static U32 ZSTD_finalizeOffBase(U32 rawOffset, const U32 rep[ZSTD_REP_NUM], U32 ll0) +{ + U32 offBase = OFFSET_TO_OFFBASE(rawOffset); + + if (!ll0 && rawOffset == rep[0]) { + offBase = REPCODE1_TO_OFFBASE; + } else if (rawOffset == rep[1]) { + offBase = REPCODE_TO_OFFBASE(2 - ll0); + } else if (rawOffset == rep[2]) { + offBase = REPCODE_TO_OFFBASE(3 - ll0); + } else if (ll0 && rawOffset == rep[0] - 1) { + offBase = REPCODE3_TO_OFFBASE; + } + return offBase; +} + +size_t +ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx, + ZSTD_sequencePosition* seqPos, + const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, + const void* src, size_t blockSize, + ZSTD_paramSwitch_e externalRepSearch) +{ + U32 idx = seqPos->idx; + U32 const startIdx = idx; + BYTE const* ip = (BYTE const*)(src); + const BYTE* const iend = ip + blockSize; + repcodes_t updatedRepcodes; + U32 dictSize; + + DEBUGLOG(5, "ZSTD_copySequencesToSeqStoreExplicitBlockDelim (blockSize = %zu)", blockSize); + + if (cctx->cdict) { + dictSize = (U32)cctx->cdict->dictContentSize; + } else if (cctx->prefixDict.dict) { + dictSize = (U32)cctx->prefixDict.dictSize; + } else { + dictSize = 0; + } + ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); + for (; idx < inSeqsSize && (inSeqs[idx].matchLength != 0 || inSeqs[idx].offset != 0); ++idx) { + U32 const litLength = inSeqs[idx].litLength; + U32 const matchLength = inSeqs[idx].matchLength; + U32 offBase; + + if (externalRepSearch == ZSTD_ps_disable) { + offBase = OFFSET_TO_OFFBASE(inSeqs[idx].offset); + } else { + U32 const ll0 = (litLength == 0); + offBase = ZSTD_finalizeOffBase(inSeqs[idx].offset, updatedRepcodes.rep, ll0); + ZSTD_updateRep(updatedRepcodes.rep, offBase, ll0); + } + + DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offBase, matchLength, litLength); + if (cctx->appliedParams.validateSequences) { + seqPos->posInSrc += litLength + matchLength; + FORWARD_IF_ERROR(ZSTD_validateSequence(offBase, matchLength, cctx->appliedParams.cParams.minMatch, seqPos->posInSrc, + cctx->appliedParams.cParams.windowLog, dictSize, cctx->appliedParams.useSequenceProducer), + "Sequence validation failed"); + } + RETURN_ERROR_IF(idx - seqPos->idx >= cctx->seqStore.maxNbSeq, externalSequences_invalid, + "Not enough memory allocated. Try adjusting ZSTD_c_minMatch."); + ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offBase, matchLength); + ip += matchLength + litLength; + } + + /* If we skipped repcode search while parsing, we need to update repcodes now */ + assert(externalRepSearch != ZSTD_ps_auto); + assert(idx >= startIdx); + if (externalRepSearch == ZSTD_ps_disable && idx != startIdx) { + U32* const rep = updatedRepcodes.rep; + U32 lastSeqIdx = idx - 1; /* index of last non-block-delimiter sequence */ + + if (lastSeqIdx >= startIdx + 2) { + rep[2] = inSeqs[lastSeqIdx - 2].offset; + rep[1] = inSeqs[lastSeqIdx - 1].offset; + rep[0] = inSeqs[lastSeqIdx].offset; + } else if (lastSeqIdx == startIdx + 1) { + rep[2] = rep[0]; + rep[1] = inSeqs[lastSeqIdx - 1].offset; + rep[0] = inSeqs[lastSeqIdx].offset; + } else { + assert(lastSeqIdx == startIdx); + rep[2] = rep[1]; + rep[1] = rep[0]; + rep[0] = inSeqs[lastSeqIdx].offset; + } + } + + ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t)); + + if (inSeqs[idx].litLength) { + DEBUGLOG(6, "Storing last literals of size: %u", inSeqs[idx].litLength); + ZSTD_storeLastLiterals(&cctx->seqStore, ip, inSeqs[idx].litLength); + ip += inSeqs[idx].litLength; + seqPos->posInSrc += inSeqs[idx].litLength; + } + RETURN_ERROR_IF(ip != iend, externalSequences_invalid, "Blocksize doesn't agree with block delimiter!"); + seqPos->idx = idx+1; + return 0; +} + +size_t +ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos, + const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, + const void* src, size_t blockSize, ZSTD_paramSwitch_e externalRepSearch) +{ + U32 idx = seqPos->idx; + U32 startPosInSequence = seqPos->posInSequence; + U32 endPosInSequence = seqPos->posInSequence + (U32)blockSize; + size_t dictSize; + BYTE const* ip = (BYTE const*)(src); + BYTE const* iend = ip + blockSize; /* May be adjusted if we decide to process fewer than blockSize bytes */ + repcodes_t updatedRepcodes; + U32 bytesAdjustment = 0; + U32 finalMatchSplit = 0; + + /* TODO(embg) support fast parsing mode in noBlockDelim mode */ + (void)externalRepSearch; + + if (cctx->cdict) { + dictSize = cctx->cdict->dictContentSize; + } else if (cctx->prefixDict.dict) { + dictSize = cctx->prefixDict.dictSize; + } else { + dictSize = 0; + } + DEBUGLOG(5, "ZSTD_copySequencesToSeqStoreNoBlockDelim: idx: %u PIS: %u blockSize: %zu", idx, startPosInSequence, blockSize); + DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); + ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t)); + while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) { + const ZSTD_Sequence currSeq = inSeqs[idx]; + U32 litLength = currSeq.litLength; + U32 matchLength = currSeq.matchLength; + U32 const rawOffset = currSeq.offset; + U32 offBase; + + /* Modify the sequence depending on where endPosInSequence lies */ + if (endPosInSequence >= currSeq.litLength + currSeq.matchLength) { + if (startPosInSequence >= litLength) { + startPosInSequence -= litLength; + litLength = 0; + matchLength -= startPosInSequence; + } else { + litLength -= startPosInSequence; + } + /* Move to the next sequence */ + endPosInSequence -= currSeq.litLength + currSeq.matchLength; + startPosInSequence = 0; + } else { + /* This is the final (partial) sequence we're adding from inSeqs, and endPosInSequence + does not reach the end of the match. So, we have to split the sequence */ + DEBUGLOG(6, "Require a split: diff: %u, idx: %u PIS: %u", + currSeq.litLength + currSeq.matchLength - endPosInSequence, idx, endPosInSequence); + if (endPosInSequence > litLength) { + U32 firstHalfMatchLength; + litLength = startPosInSequence >= litLength ? 0 : litLength - startPosInSequence; + firstHalfMatchLength = endPosInSequence - startPosInSequence - litLength; + if (matchLength > blockSize && firstHalfMatchLength >= cctx->appliedParams.cParams.minMatch) { + /* Only ever split the match if it is larger than the block size */ + U32 secondHalfMatchLength = currSeq.matchLength + currSeq.litLength - endPosInSequence; + if (secondHalfMatchLength < cctx->appliedParams.cParams.minMatch) { + /* Move the endPosInSequence backward so that it creates match of minMatch length */ + endPosInSequence -= cctx->appliedParams.cParams.minMatch - secondHalfMatchLength; + bytesAdjustment = cctx->appliedParams.cParams.minMatch - secondHalfMatchLength; + firstHalfMatchLength -= bytesAdjustment; + } + matchLength = firstHalfMatchLength; + /* Flag that we split the last match - after storing the sequence, exit the loop, + but keep the value of endPosInSequence */ + finalMatchSplit = 1; + } else { + /* Move the position in sequence backwards so that we don't split match, and break to store + * the last literals. We use the original currSeq.litLength as a marker for where endPosInSequence + * should go. We prefer to do this whenever it is not necessary to split the match, or if doing so + * would cause the first half of the match to be too small + */ + bytesAdjustment = endPosInSequence - currSeq.litLength; + endPosInSequence = currSeq.litLength; + break; + } + } else { + /* This sequence ends inside the literals, break to store the last literals */ + break; + } + } + /* Check if this offset can be represented with a repcode */ + { U32 const ll0 = (litLength == 0); + offBase = ZSTD_finalizeOffBase(rawOffset, updatedRepcodes.rep, ll0); + ZSTD_updateRep(updatedRepcodes.rep, offBase, ll0); + } + + if (cctx->appliedParams.validateSequences) { + seqPos->posInSrc += litLength + matchLength; + FORWARD_IF_ERROR(ZSTD_validateSequence(offBase, matchLength, cctx->appliedParams.cParams.minMatch, seqPos->posInSrc, + cctx->appliedParams.cParams.windowLog, dictSize, cctx->appliedParams.useSequenceProducer), + "Sequence validation failed"); + } + DEBUGLOG(6, "Storing sequence: (of: %u, ml: %u, ll: %u)", offBase, matchLength, litLength); + RETURN_ERROR_IF(idx - seqPos->idx >= cctx->seqStore.maxNbSeq, externalSequences_invalid, + "Not enough memory allocated. Try adjusting ZSTD_c_minMatch."); + ZSTD_storeSeq(&cctx->seqStore, litLength, ip, iend, offBase, matchLength); + ip += matchLength + litLength; + if (!finalMatchSplit) + idx++; /* Next Sequence */ + } + DEBUGLOG(5, "Ending seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength); + assert(idx == inSeqsSize || endPosInSequence <= inSeqs[idx].litLength + inSeqs[idx].matchLength); + seqPos->idx = idx; + seqPos->posInSequence = endPosInSequence; + ZSTD_memcpy(cctx->blockState.nextCBlock->rep, updatedRepcodes.rep, sizeof(repcodes_t)); + + iend -= bytesAdjustment; + if (ip != iend) { + /* Store any last literals */ + U32 lastLLSize = (U32)(iend - ip); + assert(ip <= iend); + DEBUGLOG(6, "Storing last literals of size: %u", lastLLSize); + ZSTD_storeLastLiterals(&cctx->seqStore, ip, lastLLSize); + seqPos->posInSrc += lastLLSize; + } + + return bytesAdjustment; +} + +typedef size_t (*ZSTD_sequenceCopier) (ZSTD_CCtx* cctx, ZSTD_sequencePosition* seqPos, + const ZSTD_Sequence* const inSeqs, size_t inSeqsSize, + const void* src, size_t blockSize, ZSTD_paramSwitch_e externalRepSearch); +static ZSTD_sequenceCopier ZSTD_selectSequenceCopier(ZSTD_sequenceFormat_e mode) +{ + ZSTD_sequenceCopier sequenceCopier = NULL; + assert(ZSTD_cParam_withinBounds(ZSTD_c_blockDelimiters, mode)); + if (mode == ZSTD_sf_explicitBlockDelimiters) { + return ZSTD_copySequencesToSeqStoreExplicitBlockDelim; + } else if (mode == ZSTD_sf_noBlockDelimiters) { + return ZSTD_copySequencesToSeqStoreNoBlockDelim; + } + assert(sequenceCopier != NULL); + return sequenceCopier; +} + +/* Discover the size of next block by searching for the delimiter. + * Note that a block delimiter **must** exist in this mode, + * otherwise it's an input error. + * The block size retrieved will be later compared to ensure it remains within bounds */ +static size_t +blockSize_explicitDelimiter(const ZSTD_Sequence* inSeqs, size_t inSeqsSize, ZSTD_sequencePosition seqPos) +{ + int end = 0; + size_t blockSize = 0; + size_t spos = seqPos.idx; + DEBUGLOG(6, "blockSize_explicitDelimiter : seq %zu / %zu", spos, inSeqsSize); + assert(spos <= inSeqsSize); + while (spos < inSeqsSize) { + end = (inSeqs[spos].offset == 0); + blockSize += inSeqs[spos].litLength + inSeqs[spos].matchLength; + if (end) { + if (inSeqs[spos].matchLength != 0) + RETURN_ERROR(externalSequences_invalid, "delimiter format error : both matchlength and offset must be == 0"); + break; + } + spos++; + } + if (!end) + RETURN_ERROR(externalSequences_invalid, "Reached end of sequences without finding a block delimiter"); + return blockSize; +} + +/* More a "target" block size */ +static size_t blockSize_noDelimiter(size_t blockSize, size_t remaining) +{ + int const lastBlock = (remaining <= blockSize); + return lastBlock ? remaining : blockSize; +} + +static size_t determine_blockSize(ZSTD_sequenceFormat_e mode, + size_t blockSize, size_t remaining, + const ZSTD_Sequence* inSeqs, size_t inSeqsSize, ZSTD_sequencePosition seqPos) +{ + DEBUGLOG(6, "determine_blockSize : remainingSize = %zu", remaining); + if (mode == ZSTD_sf_noBlockDelimiters) + return blockSize_noDelimiter(blockSize, remaining); + { size_t const explicitBlockSize = blockSize_explicitDelimiter(inSeqs, inSeqsSize, seqPos); + FORWARD_IF_ERROR(explicitBlockSize, "Error while determining block size with explicit delimiters"); + if (explicitBlockSize > blockSize) + RETURN_ERROR(externalSequences_invalid, "sequences incorrectly define a too large block"); + if (explicitBlockSize > remaining) + RETURN_ERROR(externalSequences_invalid, "sequences define a frame longer than source"); + return explicitBlockSize; + } +} + +/* Compress, block-by-block, all of the sequences given. + * + * Returns the cumulative size of all compressed blocks (including their headers), + * otherwise a ZSTD error. + */ +static size_t +ZSTD_compressSequences_internal(ZSTD_CCtx* cctx, + void* dst, size_t dstCapacity, + const ZSTD_Sequence* inSeqs, size_t inSeqsSize, + const void* src, size_t srcSize) +{ + size_t cSize = 0; + size_t remaining = srcSize; + ZSTD_sequencePosition seqPos = {0, 0, 0}; + + BYTE const* ip = (BYTE const*)src; + BYTE* op = (BYTE*)dst; + ZSTD_sequenceCopier const sequenceCopier = ZSTD_selectSequenceCopier(cctx->appliedParams.blockDelimiters); + + DEBUGLOG(4, "ZSTD_compressSequences_internal srcSize: %zu, inSeqsSize: %zu", srcSize, inSeqsSize); + /* Special case: empty frame */ + if (remaining == 0) { + U32 const cBlockHeader24 = 1 /* last block */ + (((U32)bt_raw)<<1); + RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "No room for empty frame block header"); + MEM_writeLE32(op, cBlockHeader24); + op += ZSTD_blockHeaderSize; + dstCapacity -= ZSTD_blockHeaderSize; + cSize += ZSTD_blockHeaderSize; + } + + while (remaining) { + size_t compressedSeqsSize; + size_t cBlockSize; + size_t additionalByteAdjustment; + size_t blockSize = determine_blockSize(cctx->appliedParams.blockDelimiters, + cctx->blockSize, remaining, + inSeqs, inSeqsSize, seqPos); + U32 const lastBlock = (blockSize == remaining); + FORWARD_IF_ERROR(blockSize, "Error while trying to determine block size"); + assert(blockSize <= remaining); + ZSTD_resetSeqStore(&cctx->seqStore); + DEBUGLOG(5, "Working on new block. Blocksize: %zu (total:%zu)", blockSize, (ip - (const BYTE*)src) + blockSize); + + additionalByteAdjustment = sequenceCopier(cctx, &seqPos, inSeqs, inSeqsSize, ip, blockSize, cctx->appliedParams.searchForExternalRepcodes); + FORWARD_IF_ERROR(additionalByteAdjustment, "Bad sequence copy"); + blockSize -= additionalByteAdjustment; + + /* If blocks are too small, emit as a nocompress block */ + /* TODO: See 3090. We reduced MIN_CBLOCK_SIZE from 3 to 2 so to compensate we are adding + * additional 1. We need to revisit and change this logic to be more consistent */ + if (blockSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1+1) { + cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); + FORWARD_IF_ERROR(cBlockSize, "Nocompress block failed"); + DEBUGLOG(5, "Block too small, writing out nocompress block: cSize: %zu", cBlockSize); + cSize += cBlockSize; + ip += blockSize; + op += cBlockSize; + remaining -= blockSize; + dstCapacity -= cBlockSize; + continue; + } + + RETURN_ERROR_IF(dstCapacity < ZSTD_blockHeaderSize, dstSize_tooSmall, "not enough dstCapacity to write a new compressed block"); + compressedSeqsSize = ZSTD_entropyCompressSeqStore(&cctx->seqStore, + &cctx->blockState.prevCBlock->entropy, &cctx->blockState.nextCBlock->entropy, + &cctx->appliedParams, + op + ZSTD_blockHeaderSize /* Leave space for block header */, dstCapacity - ZSTD_blockHeaderSize, + blockSize, + cctx->entropyWorkspace, ENTROPY_WORKSPACE_SIZE /* statically allocated in resetCCtx */, + cctx->bmi2); + FORWARD_IF_ERROR(compressedSeqsSize, "Compressing sequences of block failed"); + DEBUGLOG(5, "Compressed sequences size: %zu", compressedSeqsSize); + + if (!cctx->isFirstBlock && + ZSTD_maybeRLE(&cctx->seqStore) && + ZSTD_isRLE(ip, blockSize)) { + /* We don't want to emit our first block as a RLE even if it qualifies because + * doing so will cause the decoder (cli only) to throw a "should consume all input error." + * This is only an issue for zstd <= v1.4.3 + */ + compressedSeqsSize = 1; + } + + if (compressedSeqsSize == 0) { + /* ZSTD_noCompressBlock writes the block header as well */ + cBlockSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock); + FORWARD_IF_ERROR(cBlockSize, "ZSTD_noCompressBlock failed"); + DEBUGLOG(5, "Writing out nocompress block, size: %zu", cBlockSize); + } else if (compressedSeqsSize == 1) { + cBlockSize = ZSTD_rleCompressBlock(op, dstCapacity, *ip, blockSize, lastBlock); + FORWARD_IF_ERROR(cBlockSize, "ZSTD_rleCompressBlock failed"); + DEBUGLOG(5, "Writing out RLE block, size: %zu", cBlockSize); + } else { + U32 cBlockHeader; + /* Error checking and repcodes update */ + ZSTD_blockState_confirmRepcodesAndEntropyTables(&cctx->blockState); + if (cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid) + cctx->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check; + + /* Write block header into beginning of block*/ + cBlockHeader = lastBlock + (((U32)bt_compressed)<<1) + (U32)(compressedSeqsSize << 3); + MEM_writeLE24(op, cBlockHeader); + cBlockSize = ZSTD_blockHeaderSize + compressedSeqsSize; + DEBUGLOG(5, "Writing out compressed block, size: %zu", cBlockSize); + } + + cSize += cBlockSize; + + if (lastBlock) { + break; + } else { + ip += blockSize; + op += cBlockSize; + remaining -= blockSize; + dstCapacity -= cBlockSize; + cctx->isFirstBlock = 0; + } + DEBUGLOG(5, "cSize running total: %zu (remaining dstCapacity=%zu)", cSize, dstCapacity); + } + + DEBUGLOG(4, "cSize final total: %zu", cSize); + return cSize; +} + +size_t ZSTD_compressSequences(ZSTD_CCtx* cctx, + void* dst, size_t dstCapacity, + const ZSTD_Sequence* inSeqs, size_t inSeqsSize, + const void* src, size_t srcSize) +{ + BYTE* op = (BYTE*)dst; + size_t cSize = 0; + size_t compressedBlocksSize = 0; + size_t frameHeaderSize = 0; + + /* Transparent initialization stage, same as compressStream2() */ + DEBUGLOG(4, "ZSTD_compressSequences (dstCapacity=%zu)", dstCapacity); + assert(cctx != NULL); + FORWARD_IF_ERROR(ZSTD_CCtx_init_compressStream2(cctx, ZSTD_e_end, srcSize), "CCtx initialization failed"); + /* Begin writing output, starting with frame header */ + frameHeaderSize = ZSTD_writeFrameHeader(op, dstCapacity, &cctx->appliedParams, srcSize, cctx->dictID); + op += frameHeaderSize; + dstCapacity -= frameHeaderSize; + cSize += frameHeaderSize; + if (cctx->appliedParams.fParams.checksumFlag && srcSize) { + XXH64_update(&cctx->xxhState, src, srcSize); + } + /* cSize includes block header size and compressed sequences size */ + compressedBlocksSize = ZSTD_compressSequences_internal(cctx, + op, dstCapacity, + inSeqs, inSeqsSize, + src, srcSize); + FORWARD_IF_ERROR(compressedBlocksSize, "Compressing blocks failed!"); + cSize += compressedBlocksSize; + dstCapacity -= compressedBlocksSize; + + if (cctx->appliedParams.fParams.checksumFlag) { + U32 const checksum = (U32) XXH64_digest(&cctx->xxhState); + RETURN_ERROR_IF(dstCapacity<4, dstSize_tooSmall, "no room for checksum"); + DEBUGLOG(4, "Write checksum : %08X", (unsigned)checksum); + MEM_writeLE32((char*)dst + cSize, checksum); + cSize += 4; + } + + DEBUGLOG(4, "Final compressed size: %zu", cSize); + return cSize; +} + /*====== Finalize ======*/ +static ZSTD_inBuffer inBuffer_forEndFlush(const ZSTD_CStream* zcs) +{ + const ZSTD_inBuffer nullInput = { NULL, 0, 0 }; + const int stableInput = (zcs->appliedParams.inBufferMode == ZSTD_bm_stable); + return stableInput ? zcs->expectedInBuffer : nullInput; +} + /*! ZSTD_flushStream() : * @return : amount of data remaining to flush */ size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output) { - ZSTD_inBuffer input = { NULL, 0, 0 }; + ZSTD_inBuffer input = inBuffer_forEndFlush(zcs); + input.size = input.pos; /* do not ingest more input during flush */ return ZSTD_compressStream2(zcs, output, &input, ZSTD_e_flush); } size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output) { - ZSTD_inBuffer input = { NULL, 0, 0 }; + ZSTD_inBuffer input = inBuffer_forEndFlush(zcs); size_t const remainingToFlush = ZSTD_compressStream2(zcs, output, &input, ZSTD_e_end); - FORWARD_IF_ERROR( remainingToFlush , "ZSTD_compressStream2 failed"); + FORWARD_IF_ERROR(remainingToFlush , "ZSTD_compressStream2(,,ZSTD_e_end) failed"); if (zcs->appliedParams.nbWorkers > 0) return remainingToFlush; /* minimal estimation */ /* single thread mode : attempt to calculate remaining to flush more precisely */ { size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE; @@ -4111,137 +6763,116 @@ size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output) /*-===== Pre-defined compression levels =====-*/ +#include "clevels.h" -#define ZSTD_MAX_CLEVEL 22 int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; } int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; } +int ZSTD_defaultCLevel(void) { return ZSTD_CLEVEL_DEFAULT; } -static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEVEL+1] = { -{ /* "default" - for any srcSize > 256 KB */ - /* W, C, H, S, L, TL, strat */ - { 19, 12, 13, 1, 6, 1, ZSTD_fast }, /* base for negative levels */ - { 19, 13, 14, 1, 7, 0, ZSTD_fast }, /* level 1 */ - { 20, 15, 16, 1, 6, 0, ZSTD_fast }, /* level 2 */ - { 21, 16, 17, 1, 5, 0, ZSTD_dfast }, /* level 3 */ - { 21, 18, 18, 1, 5, 0, ZSTD_dfast }, /* level 4 */ - { 21, 18, 19, 2, 5, 2, ZSTD_greedy }, /* level 5 */ - { 21, 19, 19, 3, 5, 4, ZSTD_greedy }, /* level 6 */ - { 21, 19, 19, 3, 5, 8, ZSTD_lazy }, /* level 7 */ - { 21, 19, 19, 3, 5, 16, ZSTD_lazy2 }, /* level 8 */ - { 21, 19, 20, 4, 5, 16, ZSTD_lazy2 }, /* level 9 */ - { 22, 20, 21, 4, 5, 16, ZSTD_lazy2 }, /* level 10 */ - { 22, 21, 22, 4, 5, 16, ZSTD_lazy2 }, /* level 11 */ - { 22, 21, 22, 5, 5, 16, ZSTD_lazy2 }, /* level 12 */ - { 22, 21, 22, 5, 5, 32, ZSTD_btlazy2 }, /* level 13 */ - { 22, 22, 23, 5, 5, 32, ZSTD_btlazy2 }, /* level 14 */ - { 22, 23, 23, 6, 5, 32, ZSTD_btlazy2 }, /* level 15 */ - { 22, 22, 22, 5, 5, 48, ZSTD_btopt }, /* level 16 */ - { 23, 23, 22, 5, 4, 64, ZSTD_btopt }, /* level 17 */ - { 23, 23, 22, 6, 3, 64, ZSTD_btultra }, /* level 18 */ - { 23, 24, 22, 7, 3,256, ZSTD_btultra2}, /* level 19 */ - { 25, 25, 23, 7, 3,256, ZSTD_btultra2}, /* level 20 */ - { 26, 26, 24, 7, 3,512, ZSTD_btultra2}, /* level 21 */ - { 27, 27, 25, 9, 3,999, ZSTD_btultra2}, /* level 22 */ -}, -{ /* for srcSize <= 256 KB */ - /* W, C, H, S, L, T, strat */ - { 18, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */ - { 18, 13, 14, 1, 6, 0, ZSTD_fast }, /* level 1 */ - { 18, 14, 14, 1, 5, 0, ZSTD_dfast }, /* level 2 */ - { 18, 16, 16, 1, 4, 0, ZSTD_dfast }, /* level 3 */ - { 18, 16, 17, 2, 5, 2, ZSTD_greedy }, /* level 4.*/ - { 18, 18, 18, 3, 5, 2, ZSTD_greedy }, /* level 5.*/ - { 18, 18, 19, 3, 5, 4, ZSTD_lazy }, /* level 6.*/ - { 18, 18, 19, 4, 4, 4, ZSTD_lazy }, /* level 7 */ - { 18, 18, 19, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */ - { 18, 18, 19, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */ - { 18, 18, 19, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */ - { 18, 18, 19, 5, 4, 12, ZSTD_btlazy2 }, /* level 11.*/ - { 18, 19, 19, 7, 4, 12, ZSTD_btlazy2 }, /* level 12.*/ - { 18, 18, 19, 4, 4, 16, ZSTD_btopt }, /* level 13 */ - { 18, 18, 19, 4, 3, 32, ZSTD_btopt }, /* level 14.*/ - { 18, 18, 19, 6, 3,128, ZSTD_btopt }, /* level 15.*/ - { 18, 19, 19, 6, 3,128, ZSTD_btultra }, /* level 16.*/ - { 18, 19, 19, 8, 3,256, ZSTD_btultra }, /* level 17.*/ - { 18, 19, 19, 6, 3,128, ZSTD_btultra2}, /* level 18.*/ - { 18, 19, 19, 8, 3,256, ZSTD_btultra2}, /* level 19.*/ - { 18, 19, 19, 10, 3,512, ZSTD_btultra2}, /* level 20.*/ - { 18, 19, 19, 12, 3,512, ZSTD_btultra2}, /* level 21.*/ - { 18, 19, 19, 13, 3,999, ZSTD_btultra2}, /* level 22.*/ -}, -{ /* for srcSize <= 128 KB */ - /* W, C, H, S, L, T, strat */ - { 17, 12, 12, 1, 5, 1, ZSTD_fast }, /* base for negative levels */ - { 17, 12, 13, 1, 6, 0, ZSTD_fast }, /* level 1 */ - { 17, 13, 15, 1, 5, 0, ZSTD_fast }, /* level 2 */ - { 17, 15, 16, 2, 5, 0, ZSTD_dfast }, /* level 3 */ - { 17, 17, 17, 2, 4, 0, ZSTD_dfast }, /* level 4 */ - { 17, 16, 17, 3, 4, 2, ZSTD_greedy }, /* level 5 */ - { 17, 17, 17, 3, 4, 4, ZSTD_lazy }, /* level 6 */ - { 17, 17, 17, 3, 4, 8, ZSTD_lazy2 }, /* level 7 */ - { 17, 17, 17, 4, 4, 8, ZSTD_lazy2 }, /* level 8 */ - { 17, 17, 17, 5, 4, 8, ZSTD_lazy2 }, /* level 9 */ - { 17, 17, 17, 6, 4, 8, ZSTD_lazy2 }, /* level 10 */ - { 17, 17, 17, 5, 4, 8, ZSTD_btlazy2 }, /* level 11 */ - { 17, 18, 17, 7, 4, 12, ZSTD_btlazy2 }, /* level 12 */ - { 17, 18, 17, 3, 4, 12, ZSTD_btopt }, /* level 13.*/ - { 17, 18, 17, 4, 3, 32, ZSTD_btopt }, /* level 14.*/ - { 17, 18, 17, 6, 3,256, ZSTD_btopt }, /* level 15.*/ - { 17, 18, 17, 6, 3,128, ZSTD_btultra }, /* level 16.*/ - { 17, 18, 17, 8, 3,256, ZSTD_btultra }, /* level 17.*/ - { 17, 18, 17, 10, 3,512, ZSTD_btultra }, /* level 18.*/ - { 17, 18, 17, 5, 3,256, ZSTD_btultra2}, /* level 19.*/ - { 17, 18, 17, 7, 3,512, ZSTD_btultra2}, /* level 20.*/ - { 17, 18, 17, 9, 3,512, ZSTD_btultra2}, /* level 21.*/ - { 17, 18, 17, 11, 3,999, ZSTD_btultra2}, /* level 22.*/ -}, -{ /* for srcSize <= 16 KB */ - /* W, C, H, S, L, T, strat */ - { 14, 12, 13, 1, 5, 1, ZSTD_fast }, /* base for negative levels */ - { 14, 14, 15, 1, 5, 0, ZSTD_fast }, /* level 1 */ - { 14, 14, 15, 1, 4, 0, ZSTD_fast }, /* level 2 */ - { 14, 14, 15, 2, 4, 0, ZSTD_dfast }, /* level 3 */ - { 14, 14, 14, 4, 4, 2, ZSTD_greedy }, /* level 4 */ - { 14, 14, 14, 3, 4, 4, ZSTD_lazy }, /* level 5.*/ - { 14, 14, 14, 4, 4, 8, ZSTD_lazy2 }, /* level 6 */ - { 14, 14, 14, 6, 4, 8, ZSTD_lazy2 }, /* level 7 */ - { 14, 14, 14, 8, 4, 8, ZSTD_lazy2 }, /* level 8.*/ - { 14, 15, 14, 5, 4, 8, ZSTD_btlazy2 }, /* level 9.*/ - { 14, 15, 14, 9, 4, 8, ZSTD_btlazy2 }, /* level 10.*/ - { 14, 15, 14, 3, 4, 12, ZSTD_btopt }, /* level 11.*/ - { 14, 15, 14, 4, 3, 24, ZSTD_btopt }, /* level 12.*/ - { 14, 15, 14, 5, 3, 32, ZSTD_btultra }, /* level 13.*/ - { 14, 15, 15, 6, 3, 64, ZSTD_btultra }, /* level 14.*/ - { 14, 15, 15, 7, 3,256, ZSTD_btultra }, /* level 15.*/ - { 14, 15, 15, 5, 3, 48, ZSTD_btultra2}, /* level 16.*/ - { 14, 15, 15, 6, 3,128, ZSTD_btultra2}, /* level 17.*/ - { 14, 15, 15, 7, 3,256, ZSTD_btultra2}, /* level 18.*/ - { 14, 15, 15, 8, 3,256, ZSTD_btultra2}, /* level 19.*/ - { 14, 15, 15, 8, 3,512, ZSTD_btultra2}, /* level 20.*/ - { 14, 15, 15, 9, 3,512, ZSTD_btultra2}, /* level 21.*/ - { 14, 15, 15, 10, 3,999, ZSTD_btultra2}, /* level 22.*/ -}, -}; +static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel, size_t const dictSize) +{ + ZSTD_compressionParameters cParams = ZSTD_getCParams_internal(compressionLevel, 0, dictSize, ZSTD_cpm_createCDict); + switch (cParams.strategy) { + case ZSTD_fast: + case ZSTD_dfast: + break; + case ZSTD_greedy: + case ZSTD_lazy: + case ZSTD_lazy2: + cParams.hashLog += ZSTD_LAZY_DDSS_BUCKET_LOG; + break; + case ZSTD_btlazy2: + case ZSTD_btopt: + case ZSTD_btultra: + case ZSTD_btultra2: + break; + } + return cParams; +} + +static int ZSTD_dedicatedDictSearch_isSupported( + ZSTD_compressionParameters const* cParams) +{ + return (cParams->strategy >= ZSTD_greedy) + && (cParams->strategy <= ZSTD_lazy2) + && (cParams->hashLog > cParams->chainLog) + && (cParams->chainLog <= 24); +} + +/** + * Reverses the adjustment applied to cparams when enabling dedicated dict + * search. This is used to recover the params set to be used in the working + * context. (Otherwise, those tables would also grow.) + */ +static void ZSTD_dedicatedDictSearch_revertCParams( + ZSTD_compressionParameters* cParams) { + switch (cParams->strategy) { + case ZSTD_fast: + case ZSTD_dfast: + break; + case ZSTD_greedy: + case ZSTD_lazy: + case ZSTD_lazy2: + cParams->hashLog -= ZSTD_LAZY_DDSS_BUCKET_LOG; + if (cParams->hashLog < ZSTD_HASHLOG_MIN) { + cParams->hashLog = ZSTD_HASHLOG_MIN; + } + break; + case ZSTD_btlazy2: + case ZSTD_btopt: + case ZSTD_btultra: + case ZSTD_btultra2: + break; + } +} + +static U64 ZSTD_getCParamRowSize(U64 srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) +{ + switch (mode) { + case ZSTD_cpm_unknown: + case ZSTD_cpm_noAttachDict: + case ZSTD_cpm_createCDict: + break; + case ZSTD_cpm_attachDict: + dictSize = 0; + break; + default: + assert(0); + break; + } + { int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN; + size_t const addedSize = unknown && dictSize > 0 ? 500 : 0; + return unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize; + } +} /*! ZSTD_getCParams_internal() : * @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize. * Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown. - * Use dictSize == 0 for unknown or unused. */ -static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) + * Use dictSize == 0 for unknown or unused. + * Note: `mode` controls how we treat the `dictSize`. See docs for `ZSTD_cParamMode_e`. */ +static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) { - int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN; - size_t const addedSize = unknown && dictSize > 0 ? 500 : 0; - U64 const rSize = unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize; + U64 const rSize = ZSTD_getCParamRowSize(srcSizeHint, dictSize, mode); U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB); - int row = compressionLevel; + int row; DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel); + + /* row */ if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */ - if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */ - if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL; + else if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */ + else if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL; + else row = compressionLevel; + { ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row]; - if (compressionLevel < 0) cp.targetLength = (unsigned)(-compressionLevel); /* acceleration factor */ + DEBUGLOG(5, "ZSTD_getCParams_internal selected tableID: %u row: %u strat: %u", tableID, row, (U32)cp.strategy); + /* acceleration factor */ + if (compressionLevel < 0) { + int const clampedCompressionLevel = MAX(ZSTD_minCLevel(), compressionLevel); + cp.targetLength = (unsigned)(-clampedCompressionLevel); + } /* refine parameters based on srcSize & dictSize */ - return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize); + return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode, ZSTD_ps_auto); } } @@ -4251,18 +6882,18 @@ static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) { if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN; - return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize); + return ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown); } /*! ZSTD_getParams() : * same idea as ZSTD_getCParams() * @return a `ZSTD_parameters` structure (instead of `ZSTD_compressionParameters`). * Fields of `ZSTD_frameParameters` are set to default values */ -static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) { +static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_cParamMode_e mode) { ZSTD_parameters params; - ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize); + ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, mode); DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel); - memset(¶ms, 0, sizeof(params)); + ZSTD_memset(¶ms, 0, sizeof(params)); params.cParams = cParams; params.fParams.contentSizeFlag = 1; return params; @@ -4274,5 +6905,23 @@ static ZSTD_parameters ZSTD_getParams_internal(int compressionLevel, unsigned lo * Fields of `ZSTD_frameParameters` are set to default values */ ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) { if (srcSizeHint == 0) srcSizeHint = ZSTD_CONTENTSIZE_UNKNOWN; - return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize); + return ZSTD_getParams_internal(compressionLevel, srcSizeHint, dictSize, ZSTD_cpm_unknown); +} + +void ZSTD_registerSequenceProducer( + ZSTD_CCtx* zc, void* mState, + ZSTD_sequenceProducer_F* mFinder +) { + if (mFinder != NULL) { + ZSTD_externalMatchCtx emctx; + emctx.mState = mState; + emctx.mFinder = mFinder; + emctx.seqBuffer = NULL; + emctx.seqBufferCapacity = 0; + zc->externalMatchCtx = emctx; + zc->requestedParams.useSequenceProducer = 1; + } else { + ZSTD_memset(&zc->externalMatchCtx, 0, sizeof(zc->externalMatchCtx)); + zc->requestedParams.useSequenceProducer = 0; + } } |