ddidderr/zstd-rs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge branch 'dev' into adapt

Browse Source
This commit is contained in:
Yann Collet
2018-09-19 12:43:42 -07:00
parent c71c4f23d7 c8ff5200f9
commit 2f78228f65
78 changed files with 4962 additions and 1660 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/BUCK
+1
View File
@@ -69,6 +69,7 @@ cxx_library(
]),
headers=subdir_glob([
('dictBuilder', 'divsufsort.h'),
('dictBuilder', 'cover.h'),
]),
srcs=glob(['dictBuilder/*.c']),
deps=[':common'],
lib/Makefile
+1 -1
View File
@@ -23,7 +23,7 @@ ifeq ($(OS),Windows_NT) # MinGW assumed
CPPFLAGS += -D__USE_MINGW_ANSI_STDIO # compatibility with %zu formatting
endif
CFLAGS ?= -O3
DEBUGFLAGS = -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow \
DEBUGFLAGS= -Wall -Wextra -Wcast-qual -Wcast-align -Wshadow \
-Wstrict-aliasing=1 -Wswitch-enum -Wdeclaration-after-statement \
-Wstrict-prototypes -Wundef -Wpointer-arith -Wformat-security \
-Wvla -Wformat=2 -Winit-self -Wfloat-equal -Wwrite-strings \
lib/common/compiler.h
+22 -5
View File
@@ -89,20 +89,37 @@
#endif
/* prefetch
* can be disabled, by declaring NO_PREFETCH macro */
* can be disabled, by declaring NO_PREFETCH macro
* All prefetch invocations use a single default locality 2,
* generating instruction prefetcht1,
* which, according to Intel, means "load data into L2 cache".
* This is a good enough "middle ground" for the time being,
* though in theory, it would be better to specialize locality depending on data being prefetched.
* Tests could not determine any sensible difference based on locality value. */
#if defined(NO_PREFETCH)
# define PREFETCH(ptr) /* disabled */
# define PREFETCH(ptr) (void)(ptr) /* disabled */
#else
# if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86)) /* _mm_prefetch() is not defined outside of x86/x64 */
# include <mmintrin.h> /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
# define PREFETCH(ptr) _mm_prefetch((const char*)ptr, _MM_HINT_T0)
# define PREFETCH(ptr) _mm_prefetch((const char*)(ptr), _MM_HINT_T1)
# elif defined(__GNUC__) && ( (__GNUC__ >= 4) || ( (__GNUC__ == 3) && (__GNUC_MINOR__ >= 1) ) )
# define PREFETCH(ptr) __builtin_prefetch(ptr, 0, 0)
# define PREFETCH(ptr) __builtin_prefetch((ptr), 0 /* rw==read */, 2 /* locality */)
# else
# define PREFETCH(ptr) /* disabled */
# define PREFETCH(ptr) (void)(ptr) /* disabled */
# endif
#endif /* NO_PREFETCH */
#define CACHELINE_SIZE 64
#define PREFETCH_AREA(p, s) { \
const char* const _ptr = (const char*)(p); \
size_t const _size = (size_t)(s); \
size_t _pos; \
for (_pos=0; _pos<_size; _pos+=CACHELINE_SIZE) { \
PREFETCH(_ptr + _pos); \
} \
}
/* disable warnings */
#ifdef _MSC_VER /* Visual Studio */
# include <intrin.h> /* For Visual 2005 */
lib/common/cpu.h
+1 -1
View File
@@ -36,7 +36,7 @@ MEM_STATIC ZSTD_cpuid_t ZSTD_cpuid(void) {
U32 f1d = 0;
U32 f7b = 0;
U32 f7c = 0;
#ifdef _MSC_VER
#if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_IX86))
int reg[4];
__cpuid((int*)reg, 0);
{
lib/common/zstd_internal.h
+3 -2
View File
@@ -79,8 +79,7 @@ static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
static const size_t ZSTD_fcs_fieldSize[4] = { 0, 2, 4, 8 };
static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 };
#define ZSTD_FRAMEIDSIZE 4
static const size_t ZSTD_frameIdSize = ZSTD_FRAMEIDSIZE; /* magic number size */
#define ZSTD_FRAMEIDSIZE 4 /* magic number size */
#define ZSTD_BLOCKHEADERSIZE 3 /* C standard doesn't allow `static const` variable to be init using another `static const` variable */
static const size_t ZSTD_blockHeaderSize = ZSTD_BLOCKHEADERSIZE;
@@ -193,6 +192,8 @@ typedef struct {
BYTE* llCode;
BYTE* mlCode;
BYTE* ofCode;
size_t maxNbSeq;
size_t maxNbLit;
U32 longLengthID; /* 0 == no longLength; 1 == Lit.longLength; 2 == Match.longLength; */
U32 longLengthPos;
} seqStore_t;
lib/compress/fse_compress.c
+51 -34
View File
@@ -83,7 +83,9 @@
* wkspSize should be sized to handle worst case situation, which is `1<<max_tableLog * sizeof(FSE_FUNCTION_TYPE)`
* workSpace must also be properly aligned with FSE_FUNCTION_TYPE requirements
*/
size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize)
size_t FSE_buildCTable_wksp(FSE_CTable* ct,
const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
void* workSpace, size_t wkspSize)
{
U32 const tableSize = 1 << tableLog;
U32 const tableMask = tableSize - 1;
@@ -101,10 +103,14 @@ size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsi
if (((size_t)1 << tableLog) * sizeof(FSE_FUNCTION_TYPE) > wkspSize) return ERROR(tableLog_tooLarge);
tableU16[-2] = (U16) tableLog;
tableU16[-1] = (U16) maxSymbolValue;
assert(tableLog < 16); /* required for the threshold strategy to work */
assert(tableLog < 16); /* required for threshold strategy to work */
/* For explanations on how to distribute symbol values over the table :
* http://fastcompression.blogspot.fr/2014/02/fse-distributing-symbol-values.html */
* http://fastcompression.blogspot.fr/2014/02/fse-distributing-symbol-values.html */
#ifdef __clang_analyzer__
memset(tableSymbol, 0, sizeof(*tableSymbol) * tableSize); /* useless initialization, just to keep scan-build happy */
#endif
/* symbol start positions */
{ U32 u;
@@ -124,13 +130,15 @@ size_t FSE_buildCTable_wksp(FSE_CTable* ct, const short* normalizedCounter, unsi
U32 symbol;
for (symbol=0; symbol<=maxSymbolValue; symbol++) {
int nbOccurences;
for (nbOccurences=0; nbOccurences<normalizedCounter[symbol]; nbOccurences++) {
int const freq = normalizedCounter[symbol];
for (nbOccurences=0; nbOccurences<freq; nbOccurences++) {
tableSymbol[position] = (FSE_FUNCTION_TYPE)symbol;
position = (position + step) & tableMask;
while (position > highThreshold) position = (position + step) & tableMask; /* Low proba area */
while (position > highThreshold)
position = (position + step) & tableMask; /* Low proba area */
} }
if (position!=0) return ERROR(GENERIC); /* Must have gone through all positions */
assert(position==0); /* Must have initialized all positions */
}
/* Build table */
@@ -201,9 +209,10 @@ size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
return maxSymbolValue ? maxHeaderSize : FSE_NCOUNTBOUND; /* maxSymbolValue==0 ? use default */
}
static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
unsigned writeIsSafe)
static size_t
FSE_writeNCount_generic (void* header, size_t headerBufferSize,
const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog,
unsigned writeIsSafe)
{
BYTE* const ostart = (BYTE*) header;
BYTE* out = ostart;
@@ -212,13 +221,12 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
const int tableSize = 1 << tableLog;
int remaining;
int threshold;
U32 bitStream;
int bitCount;
unsigned charnum = 0;
int previous0 = 0;
U32 bitStream = 0;
int bitCount = 0;
unsigned symbol = 0;
unsigned const alphabetSize = maxSymbolValue + 1;
int previousIs0 = 0;
bitStream = 0;
bitCount = 0;
/* Table Size */
bitStream += (tableLog-FSE_MIN_TABLELOG) << bitCount;
bitCount += 4;
@@ -228,48 +236,53 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
threshold = tableSize;
nbBits = tableLog+1;
while (remaining>1) { /* stops at 1 */
if (previous0) {
unsigned start = charnum;
while (!normalizedCounter[charnum]) charnum++;
while (charnum >= start+24) {
while ((symbol < alphabetSize) && (remaining>1)) { /* stops at 1 */
if (previousIs0) {
unsigned start = symbol;
while ((symbol < alphabetSize) && !normalizedCounter[symbol]) symbol++;
if (symbol == alphabetSize) break; /* incorrect distribution */
while (symbol >= start+24) {
start+=24;
bitStream += 0xFFFFU << bitCount;
if ((!writeIsSafe) && (out > oend-2)) return ERROR(dstSize_tooSmall); /* Buffer overflow */
if ((!writeIsSafe) && (out > oend-2))
return ERROR(dstSize_tooSmall); /* Buffer overflow */
out[0] = (BYTE) bitStream;
out[1] = (BYTE)(bitStream>>8);
out+=2;
bitStream>>=16;
}
while (charnum >= start+3) {
while (symbol >= start+3) {
start+=3;
bitStream += 3 << bitCount;
bitCount += 2;
}
bitStream += (charnum-start) << bitCount;
bitStream += (symbol-start) << bitCount;
bitCount += 2;
if (bitCount>16) {
if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall); /* Buffer overflow */
if ((!writeIsSafe) && (out > oend - 2))
return ERROR(dstSize_tooSmall); /* Buffer overflow */
out[0] = (BYTE)bitStream;
out[1] = (BYTE)(bitStream>>8);
out += 2;
bitStream >>= 16;
bitCount -= 16;
} }
{ int count = normalizedCounter[charnum++];
int const max = (2*threshold-1)-remaining;
{ int count = normalizedCounter[symbol++];
int const max = (2*threshold-1) - remaining;
remaining -= count < 0 ? -count : count;
count++; /* +1 for extra accuracy */
if (count>=threshold) count += max; /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
if (count>=threshold)
count += max; /* [0..max[ [max..threshold[ (...) [threshold+max 2*threshold[ */
bitStream += count << bitCount;
bitCount += nbBits;
bitCount -= (count<max);
previous0 = (count==1);
previousIs0 = (count==1);
if (remaining<1) return ERROR(GENERIC);
while (remaining<threshold) { nbBits--; threshold>>=1; }
}
if (bitCount>16) {
if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall); /* Buffer overflow */
if ((!writeIsSafe) && (out > oend - 2))
return ERROR(dstSize_tooSmall); /* Buffer overflow */
out[0] = (BYTE)bitStream;
out[1] = (BYTE)(bitStream>>8);
out += 2;
@@ -277,19 +290,23 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
bitCount -= 16;
} }
if (remaining != 1)
return ERROR(GENERIC); /* incorrect normalized distribution */
assert(symbol <= alphabetSize);
/* flush remaining bitStream */
if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall); /* Buffer overflow */
if ((!writeIsSafe) && (out > oend - 2))
return ERROR(dstSize_tooSmall); /* Buffer overflow */
out[0] = (BYTE)bitStream;
out[1] = (BYTE)(bitStream>>8);
out+= (bitCount+7) /8;
if (charnum > maxSymbolValue + 1) return ERROR(GENERIC);
return (out-ostart);
}
size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
size_t FSE_writeNCount (void* buffer, size_t bufferSize,
const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
{
if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported */
if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported */
@@ -297,7 +314,7 @@ size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalized
if (bufferSize < FSE_NCountWriteBound(maxSymbolValue, tableLog))
return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 0);
return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 1);
return FSE_writeNCount_generic(buffer, bufferSize, normalizedCounter, maxSymbolValue, tableLog, 1 /* write in buffer is safe */);
}
lib/compress/hist.h
+1 -1
View File
@@ -50,7 +50,7 @@
size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize);
unsigned HIST_isError(size_t code); /*< tells if a return value is an error code */
unsigned HIST_isError(size_t code); /**< tells if a return value is an error code */
/* --- advanced histogram functions --- */
lib/compress/zstd_compress.c
+52 -28
View File
@@ -805,7 +805,7 @@ size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
U32 const divider = (cParams.searchLength==3) ? 3 : 4;
size_t const maxNbSeq = blockSize / divider;
size_t const tokenSpace = blockSize + 11*maxNbSeq;
size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
size_t const entropySpace = HUF_WORKSPACE_SIZE;
size_t const blockStateSpace = 2 * sizeof(ZSTD_compressedBlockState_t);
size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1);
@@ -949,33 +949,51 @@ typedef enum { ZSTDb_not_buffered, ZSTDb_buffered } ZSTD_buffered_policy_e;
/* ZSTD_sufficientBuff() :
* check internal buffers exist for streaming if buffPol == ZSTDb_buffered .
* Note : they are assumed to be correctly sized if ZSTD_equivalentCParams()==1 */
static U32 ZSTD_sufficientBuff(size_t bufferSize1, size_t blockSize1,
static U32 ZSTD_sufficientBuff(size_t bufferSize1, size_t maxNbSeq1,
size_t maxNbLit1,
ZSTD_buffered_policy_e buffPol2,
ZSTD_compressionParameters cParams2,
U64 pledgedSrcSize)
{
size_t const windowSize2 = MAX(1, (size_t)MIN(((U64)1 << cParams2.windowLog), pledgedSrcSize));
size_t const blockSize2 = MIN(ZSTD_BLOCKSIZE_MAX, windowSize2);
size_t const maxNbSeq2 = blockSize2 / ((cParams2.searchLength == 3) ? 3 : 4);
size_t const maxNbLit2 = blockSize2;
size_t const neededBufferSize2 = (buffPol2==ZSTDb_buffered) ? windowSize2 + blockSize2 : 0;
DEBUGLOG(4, "ZSTD_sufficientBuff: is windowSize2=%u <= wlog1=%u",
(U32)windowSize2, cParams2.windowLog);
DEBUGLOG(4, "ZSTD_sufficientBuff: is blockSize2=%u <= blockSize1=%u",
(U32)blockSize2, (U32)blockSize1);
return (blockSize2 <= blockSize1) /* seqStore space depends on blockSize */
DEBUGLOG(4, "ZSTD_sufficientBuff: is neededBufferSize2=%u <= bufferSize1=%u",
(U32)neededBufferSize2, (U32)bufferSize1);
DEBUGLOG(4, "ZSTD_sufficientBuff: is maxNbSeq2=%u <= maxNbSeq1=%u",
(U32)maxNbSeq2, (U32)maxNbSeq1);
DEBUGLOG(4, "ZSTD_sufficientBuff: is maxNbLit2=%u <= maxNbLit1=%u",
(U32)maxNbLit2, (U32)maxNbLit1);
return (maxNbLit2 <= maxNbLit1)
& (maxNbSeq2 <= maxNbSeq1)
& (neededBufferSize2 <= bufferSize1);
}
/** Equivalence for resetCCtx purposes */
static U32 ZSTD_equivalentParams(ZSTD_CCtx_params params1,
ZSTD_CCtx_params params2,
size_t buffSize1, size_t blockSize1,
size_t buffSize1,
size_t maxNbSeq1, size_t maxNbLit1,
ZSTD_buffered_policy_e buffPol2,
U64 pledgedSrcSize)
{
DEBUGLOG(4, "ZSTD_equivalentParams: pledgedSrcSize=%u", (U32)pledgedSrcSize);
return ZSTD_equivalentCParams(params1.cParams, params2.cParams) &&
ZSTD_equivalentLdmParams(params1.ldmParams, params2.ldmParams) &&
ZSTD_sufficientBuff(buffSize1, blockSize1, buffPol2, params2.cParams, pledgedSrcSize);
if (!ZSTD_equivalentCParams(params1.cParams, params2.cParams)) {
DEBUGLOG(4, "ZSTD_equivalentCParams() == 0");
return 0;
}
if (!ZSTD_equivalentLdmParams(params1.ldmParams, params2.ldmParams)) {
DEBUGLOG(4, "ZSTD_equivalentLdmParams() == 0");
return 0;
}
if (!ZSTD_sufficientBuff(buffSize1, maxNbSeq1, maxNbLit1, buffPol2,
params2.cParams, pledgedSrcSize)) {
DEBUGLOG(4, "ZSTD_sufficientBuff() == 0");
return 0;
}
return 1;
}
static void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs)
@@ -1103,8 +1121,9 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
if (crp == ZSTDcrp_continue) {
if (ZSTD_equivalentParams(zc->appliedParams, params,
zc->inBuffSize, zc->blockSize,
zbuff, pledgedSrcSize)) {
zc->inBuffSize,
zc->seqStore.maxNbSeq, zc->seqStore.maxNbLit,
zbuff, pledgedSrcSize)) {
DEBUGLOG(4, "ZSTD_equivalentParams()==1 -> continue mode (wLog1=%u, blockSize1=%zu)",
zc->appliedParams.cParams.windowLog, zc->blockSize);
zc->workSpaceOversizedDuration += (zc->workSpaceOversizedDuration > 0); /* if it was too large, it still is */
@@ -1125,7 +1144,7 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
U32 const divider = (params.cParams.searchLength==3) ? 3 : 4;
size_t const maxNbSeq = blockSize / divider;
size_t const tokenSpace = blockSize + 11*maxNbSeq;
size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
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(&params.cParams, /* forCCtx */ 1);
@@ -1165,7 +1184,6 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
if (zc->workSpace == NULL) return ERROR(memory_allocation);
zc->workSpaceSize = neededSpace;
zc->workSpaceOversizedDuration = 0;
ptr = zc->workSpace;
/* Statically sized space.
* entropyWorkspace never moves,
@@ -1216,13 +1234,18 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
ptr = ZSTD_reset_matchState(&zc->blockState.matchState, ptr, &params.cParams, crp, /* forCCtx */ 1);
/* sequences storage */
zc->seqStore.maxNbSeq = maxNbSeq;
zc->seqStore.sequencesStart = (seqDef*)ptr;
ptr = zc->seqStore.sequencesStart + maxNbSeq;
zc->seqStore.llCode = (BYTE*) ptr;
zc->seqStore.mlCode = zc->seqStore.llCode + maxNbSeq;
zc->seqStore.ofCode = zc->seqStore.mlCode + maxNbSeq;
zc->seqStore.litStart = zc->seqStore.ofCode + maxNbSeq;
ptr = zc->seqStore.litStart + blockSize;
/* ZSTD_wildcopy() is used to copy into the literals buffer,
* so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
*/
zc->seqStore.maxNbLit = blockSize;
ptr = zc->seqStore.litStart + blockSize + WILDCOPY_OVERLENGTH;
/* ldm bucketOffsets table */
if (params.ldmParams.enableLdm) {
@@ -1341,8 +1364,7 @@ static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
}
/* copy dictionary offsets */
{
ZSTD_matchState_t const* srcMatchState = &cdict->matchState;
{ ZSTD_matchState_t const* srcMatchState = &cdict->matchState;
ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState;
dstMatchState->window = srcMatchState->window;
dstMatchState->nextToUpdate = srcMatchState->nextToUpdate;
@@ -1666,6 +1688,7 @@ void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
BYTE* const mlCodeTable = seqStorePtr->mlCode;
U32 const nbSeq = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
U32 u;
assert(nbSeq <= seqStorePtr->maxNbSeq);
for (u=0; u<nbSeq; u++) {
U32 const llv = sequences[u].litLength;
U32 const mlv = sequences[u].matchLength;
@@ -2254,13 +2277,6 @@ MEM_STATIC size_t ZSTD_compressSequences(seqStore_t* seqStorePtr,
if (cSize >= maxCSize) return 0; /* block not compressed */
}
/* We check that dictionaries have offset codes available for the first
* block. After the first block, the offcode table might not have large
* enough codes to represent the offsets in the data.
*/
if (nextEntropy->fse.offcode_repeatMode == FSE_repeat_valid)
nextEntropy->fse.offcode_repeatMode = FSE_repeat_check;
return cSize;
}
@@ -2399,12 +2415,20 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
&zc->appliedParams,
dst, dstCapacity,
srcSize, zc->entropyWorkspace, zc->bmi2);
if (ZSTD_isError(cSize) || cSize == 0) return cSize;
/* confirm repcodes and entropy tables */
{ ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;
if (!ZSTD_isError(cSize) && cSize != 0) {
/* confirm repcodes and entropy tables */
ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;
zc->blockState.prevCBlock = zc->blockState.nextCBlock;
zc->blockState.nextCBlock = tmp;
}
/* We check that dictionaries have offset codes available for the first
* block. After the first block, the offcode table might not have large
* enough codes to represent the offsets in the data.
*/
if (zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode == FSE_repeat_valid)
zc->blockState.prevCBlock->entropy.fse.offcode_repeatMode = FSE_repeat_check;
return cSize;
}
}
lib/compress/zstd_compress_internal.h
+3 -1
View File
@@ -314,8 +314,10 @@ MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const v
pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)offsetCode);
}
#endif
assert((size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart) < seqStorePtr->maxNbSeq);
/* copy Literals */
assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + 128 KB);
assert(seqStorePtr->maxNbLit <= 128 KB);
assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + seqStorePtr->maxNbLit);
ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
seqStorePtr->lit += litLength;
lib/compress/zstd_opt.c
+1 -1
View File
@@ -970,7 +970,7 @@ _shortestPath: /* cur, last_pos, best_mlen, best_off have to be set */
U32 seqPos = cur;
DEBUGLOG(6, "start reverse traversal (last_pos:%u, cur:%u)",
last_pos, cur);
last_pos, cur); (void)last_pos;
assert(storeEnd < ZSTD_OPT_NUM);
DEBUGLOG(6, "last sequence copied into pos=%u (llen=%u,mlen=%u,ofc=%u)",
storeEnd, lastSequence.litlen, lastSequence.mlen, lastSequence.off);
lib/compress/zstdmt_compress.c
+2 -1
View File
@@ -320,7 +320,8 @@ static void ZSTDMT_setNbSeq(ZSTDMT_seqPool* const seqPool, size_t const nbSeq)
static ZSTDMT_seqPool* ZSTDMT_createSeqPool(unsigned nbWorkers, ZSTD_customMem cMem)
{
ZSTDMT_seqPool* seqPool = ZSTDMT_createBufferPool(nbWorkers, cMem);
ZSTDMT_seqPool* const seqPool = ZSTDMT_createBufferPool(nbWorkers, cMem);
if (seqPool == NULL) return NULL;
ZSTDMT_setNbSeq(seqPool, 0);
return seqPool;
}
lib/decompress/huf_decompress.c
+3 -3
View File
@@ -533,9 +533,9 @@ static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
}
}
size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src,
size_t srcSize, void* workSpace,
size_t wkspSize)
size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
const void* src, size_t srcSize,
void* workSpace, size_t wkspSize)
{
U32 tableLog, maxW, sizeOfSort, nbSymbols;
DTableDesc dtd = HUF_getDTableDesc(DTable);
lib/decompress/zstd_decompress.c
+161 -99
View File
@@ -40,7 +40,6 @@
# define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_DEFAULTMAX) + 1)
#endif
/*!
* NO_FORWARD_PROGRESS_MAX :
* maximum allowed nb of calls to ZSTD_decompressStream() and ZSTD_decompress_generic()
@@ -52,11 +51,13 @@
# define ZSTD_NO_FORWARD_PROGRESS_MAX 16
#endif
/*-*******************************************************
* Dependencies
*********************************************************/
#include <string.h> /* memcpy, memmove, memset */
#include "cpu.h"
#include "compiler.h" /* prefetch */
#include "cpu.h" /* bmi2 */
#include "mem.h" /* low level memory routines */
#define FSE_STATIC_LINKING_ONLY
#include "fse.h"
@@ -68,6 +69,9 @@
# include "zstd_legacy.h"
#endif
static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict);
static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict);
/*-*************************************
* Errors
@@ -110,11 +114,10 @@ typedef struct {
#define SEQSYMBOL_TABLE_SIZE(log) (1 + (1 << (log)))
typedef struct {
ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)];
ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)];
ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)];
ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)]; /* Note : Space reserved for FSE Tables */
ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)]; /* is also used as temporary workspace while building hufTable during DDict creation */
ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)]; /* and therefore must be at least HUF_DECOMPRESS_WORKSPACE_SIZE large */
HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)]; /* can accommodate HUF_decompress4X */
U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
U32 rep[ZSTD_REP_NUM];
} ZSTD_entropyDTables_t;
@@ -125,6 +128,7 @@ struct ZSTD_DCtx_s
const ZSTD_seqSymbol* OFTptr;
const HUF_DTable* HUFptr;
ZSTD_entropyDTables_t entropy;
U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32]; /* space needed when building huffman tables */
const void* previousDstEnd; /* detect continuity */
const void* prefixStart; /* start of current segment */
const void* virtualStart; /* virtual start of previous segment if it was just before current one */
@@ -138,7 +142,6 @@ struct ZSTD_DCtx_s
U32 fseEntropy;
XXH64_state_t xxhState;
size_t headerSize;
U32 dictID;
ZSTD_format_e format;
const BYTE* litPtr;
ZSTD_customMem customMem;
@@ -147,9 +150,13 @@ struct ZSTD_DCtx_s
size_t staticSize;
int bmi2; /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
/* streaming */
/* dictionary */
ZSTD_DDict* ddictLocal;
const ZSTD_DDict* ddict;
const ZSTD_DDict* ddict; /* set by ZSTD_initDStream_usingDDict(), or ZSTD_DCtx_refDDict() */
U32 dictID;
int ddictIsCold; /* if == 1 : dictionary is "new" for working context, and presumed "cold" (not in cpu cache) */
/* streaming */
ZSTD_dStreamStage streamStage;
char* inBuff;
size_t inBuffSize;
@@ -185,7 +192,7 @@ size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
static size_t ZSTD_startingInputLength(ZSTD_format_e format)
{
size_t const startingInputLength = (format==ZSTD_f_zstd1_magicless) ?
ZSTD_frameHeaderSize_prefix - ZSTD_frameIdSize :
ZSTD_frameHeaderSize_prefix - ZSTD_FRAMEIDSIZE :
ZSTD_frameHeaderSize_prefix;
ZSTD_STATIC_ASSERT(ZSTD_FRAMEHEADERSIZE_PREFIX >= ZSTD_FRAMEIDSIZE);
/* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
@@ -200,6 +207,8 @@ static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
dctx->ddict = NULL;
dctx->ddictLocal = NULL;
dctx->dictEnd = NULL;
dctx->ddictIsCold = 0;
dctx->inBuff = NULL;
dctx->inBuffSize = 0;
dctx->outBuffSize = 0;
@@ -278,7 +287,7 @@ void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
* Note 3 : Skippable Frame Identifiers are considered valid. */
unsigned ZSTD_isFrame(const void* buffer, size_t size)
{
if (size < ZSTD_frameIdSize) return 0;
if (size < ZSTD_FRAMEIDSIZE) return 0;
{ U32 const magic = MEM_readLE32(buffer);
if (magic == ZSTD_MAGICNUMBER) return 1;
if ((magic & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) return 1;
@@ -330,7 +339,9 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
const BYTE* ip = (const BYTE*)src;
size_t const minInputSize = ZSTD_startingInputLength(format);
memset(zfhPtr, 0, sizeof(*zfhPtr)); /* not strictly necessary, but static analyzer do not understand that zfhPtr is only going to be read only if return value is zero, since they are 2 different signals */
if (srcSize < minInputSize) return minInputSize;
if (src==NULL) return ERROR(GENERIC); /* invalid parameter */
if ( (format != ZSTD_f_zstd1_magicless)
&& (MEM_readLE32(src) != ZSTD_MAGICNUMBER) ) {
@@ -339,7 +350,7 @@ size_t ZSTD_getFrameHeader_advanced(ZSTD_frameHeader* zfhPtr, const void* src, s
if (srcSize < ZSTD_skippableHeaderSize)
return ZSTD_skippableHeaderSize; /* magic number + frame length */
memset(zfhPtr, 0, sizeof(*zfhPtr));
zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_frameIdSize);
zfhPtr->frameContentSize = MEM_readLE32((const char *)src + ZSTD_FRAMEIDSIZE);
zfhPtr->frameType = ZSTD_skippableFrame;
return 0;
}
@@ -451,7 +462,7 @@ unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
size_t skippableSize;
if (srcSize < ZSTD_skippableHeaderSize)
return ERROR(srcSize_wrong);
skippableSize = MEM_readLE32((const BYTE *)src + ZSTD_frameIdSize)
skippableSize = MEM_readLE32((const BYTE *)src + ZSTD_FRAMEIDSIZE)
+ ZSTD_skippableHeaderSize;
if (srcSize < skippableSize) {
return ZSTD_CONTENTSIZE_ERROR;
@@ -540,6 +551,7 @@ size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
static size_t ZSTD_copyRawBlock(void* dst, size_t dstCapacity,
const void* src, size_t srcSize)
{
if (dst==NULL) return ERROR(dstSize_tooSmall);
if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall);
memcpy(dst, src, srcSize);
return srcSize;
@@ -572,6 +584,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
case set_repeat:
if (dctx->litEntropy==0) return ERROR(dictionary_corrupted);
/* fall-through */
case set_compressed:
if (srcSize < 5) return ERROR(corruption_detected); /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3 */
{ size_t lhSize, litSize, litCSize;
@@ -603,15 +616,20 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
/* prefetch huffman table if cold */
if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable));
}
if (HUF_isError((litEncType==set_repeat) ?
( singleStream ?
HUF_decompress1X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) :
HUF_decompress4X_usingDTable_bmi2(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr, dctx->bmi2) ) :
( singleStream ?
HUF_decompress1X1_DCtx_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
dctx->entropy.workspace, sizeof(dctx->entropy.workspace), dctx->bmi2) :
dctx->workspace, sizeof(dctx->workspace), dctx->bmi2) :
HUF_decompress4X_hufOnly_wksp_bmi2(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
dctx->entropy.workspace, sizeof(dctx->entropy.workspace), dctx->bmi2))))
dctx->workspace, sizeof(dctx->workspace), dctx->bmi2))))
return ERROR(corruption_detected);
dctx->litPtr = dctx->litBuffer;
@@ -883,7 +901,8 @@ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymb
symbolEncodingType_e type, U32 max, U32 maxLog,
const void* src, size_t srcSize,
const U32* baseValue, const U32* nbAdditionalBits,
const ZSTD_seqSymbol* defaultTable, U32 flagRepeatTable)
const ZSTD_seqSymbol* defaultTable, U32 flagRepeatTable,
int ddictIsCold, int nbSeq)
{
switch(type)
{
@@ -902,6 +921,12 @@ static size_t ZSTD_buildSeqTable(ZSTD_seqSymbol* DTableSpace, const ZSTD_seqSymb
return 0;
case set_repeat:
if (!flagRepeatTable) return ERROR(corruption_detected);
/* prefetch FSE table if used */
if (ddictIsCold && (nbSeq > 24 /* heuristic */)) {
const void* const pStart = *DTablePtr;
size_t const pSize = sizeof(ZSTD_seqSymbol) * (SEQSYMBOL_TABLE_SIZE(maxLog));
PREFETCH_AREA(pStart, pSize);
}
return 0;
case set_compressed :
{ U32 tableLog;
@@ -954,25 +979,25 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
const BYTE* const istart = (const BYTE* const)src;
const BYTE* const iend = istart + srcSize;
const BYTE* ip = istart;
int nbSeq;
DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
/* check */
if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong);
/* SeqHead */
{ int nbSeq = *ip++;
if (!nbSeq) { *nbSeqPtr=0; return 1; }
if (nbSeq > 0x7F) {
if (nbSeq == 0xFF) {
if (ip+2 > iend) return ERROR(srcSize_wrong);
nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
} else {
if (ip >= iend) return ERROR(srcSize_wrong);
nbSeq = ((nbSeq-0x80)<<8) + *ip++;
}
nbSeq = *ip++;
if (!nbSeq) { *nbSeqPtr=0; return 1; }
if (nbSeq > 0x7F) {
if (nbSeq == 0xFF) {
if (ip+2 > iend) return ERROR(srcSize_wrong);
nbSeq = MEM_readLE16(ip) + LONGNBSEQ, ip+=2;
} else {
if (ip >= iend) return ERROR(srcSize_wrong);
nbSeq = ((nbSeq-0x80)<<8) + *ip++;
}
*nbSeqPtr = nbSeq;
}
*nbSeqPtr = nbSeq;
/* FSE table descriptors */
if (ip+4 > iend) return ERROR(srcSize_wrong); /* minimum possible size */
@@ -986,7 +1011,8 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
LLtype, MaxLL, LLFSELog,
ip, iend-ip,
LL_base, LL_bits,
LL_defaultDTable, dctx->fseEntropy);
LL_defaultDTable, dctx->fseEntropy,
dctx->ddictIsCold, nbSeq);
if (ZSTD_isError(llhSize)) return ERROR(corruption_detected);
ip += llhSize;
}
@@ -995,7 +1021,8 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
OFtype, MaxOff, OffFSELog,
ip, iend-ip,
OF_base, OF_bits,
OF_defaultDTable, dctx->fseEntropy);
OF_defaultDTable, dctx->fseEntropy,
dctx->ddictIsCold, nbSeq);
if (ZSTD_isError(ofhSize)) return ERROR(corruption_detected);
ip += ofhSize;
}
@@ -1004,12 +1031,23 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
MLtype, MaxML, MLFSELog,
ip, iend-ip,
ML_base, ML_bits,
ML_defaultDTable, dctx->fseEntropy);
ML_defaultDTable, dctx->fseEntropy,
dctx->ddictIsCold, nbSeq);
if (ZSTD_isError(mlhSize)) return ERROR(corruption_detected);
ip += mlhSize;
}
}
/* prefetch dictionary content */
if (dctx->ddictIsCold) {
size_t const dictSize = (const char*)dctx->prefixStart - (const char*)dctx->virtualStart;
size_t const psmin = MIN(dictSize, (size_t)(64*nbSeq) /* heuristic */ );
size_t const pSize = MIN(psmin, 128 KB /* protection */ );
const void* const pStart = (const char*)dctx->dictEnd - pSize;
PREFETCH_AREA(pStart, pSize);
dctx->ddictIsCold = 0;
}
return ip-istart;
}
@@ -1676,7 +1714,8 @@ static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
/* isLongOffset must be true if there are long offsets.
* Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN.
* We don't expect that to be the case in 64-bit mode.
* In block mode, window size is not known, so we have to be conservative. (note: but it could be evaluated from current-lowLimit)
* In block mode, window size is not known, so we have to be conservative.
* (note: but it could be evaluated from current-lowLimit)
*/
ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN)));
DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize);
@@ -1763,7 +1802,7 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
#endif
if ( (srcSize >= ZSTD_skippableHeaderSize)
&& (MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START ) {
return ZSTD_skippableHeaderSize + MEM_readLE32((const BYTE*)src + ZSTD_frameIdSize);
return ZSTD_skippableHeaderSize + MEM_readLE32((const BYTE*)src + ZSTD_FRAMEIDSIZE);
} else {
const BYTE* ip = (const BYTE*)src;
const BYTE* const ipstart = ip;
@@ -1797,7 +1836,6 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
if (zfh.checksumFlag) { /* Final frame content checksum */
if (remainingSize < 4) return ERROR(srcSize_wrong);
ip += 4;
remainingSize -= 4;
}
return ip - ipstart;
@@ -1885,9 +1923,6 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
return op-ostart;
}
static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict);
static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict);
static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
@@ -1896,6 +1931,8 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
{
void* const dststart = dst;
int moreThan1Frame = 0;
DEBUGLOG(5, "ZSTD_decompressMultiFrame");
assert(dict==NULL || ddict==NULL); /* either dict or ddict set, not both */
if (ddict) {
@@ -1932,7 +1969,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
size_t skippableSize;
if (srcSize < ZSTD_skippableHeaderSize)
return ERROR(srcSize_wrong);
skippableSize = MEM_readLE32((const BYTE*)src + ZSTD_frameIdSize)
skippableSize = MEM_readLE32((const BYTE*)src + ZSTD_FRAMEIDSIZE)
+ ZSTD_skippableHeaderSize;
if (srcSize < skippableSize) return ERROR(srcSize_wrong);
@@ -2057,7 +2094,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
case ZSTDds_getFrameHeaderSize :
assert(src != NULL);
if (dctx->format == ZSTD_f_zstd1) { /* allows header */
assert(srcSize >= ZSTD_frameIdSize); /* to read skippable magic number */
assert(srcSize >= ZSTD_FRAMEIDSIZE); /* to read skippable magic number */
if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
memcpy(dctx->headerBuffer, src, srcSize);
dctx->expected = ZSTD_skippableHeaderSize - srcSize; /* remaining to load to get full skippable frame header */
@@ -2167,7 +2204,7 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
assert(src != NULL);
assert(srcSize <= ZSTD_skippableHeaderSize);
memcpy(dctx->headerBuffer + (ZSTD_skippableHeaderSize - srcSize), src, srcSize); /* complete skippable header */
dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_frameIdSize); /* note : dctx->expected can grow seriously large, beyond local buffer size */
dctx->expected = MEM_readLE32(dctx->headerBuffer + ZSTD_FRAMEIDSIZE); /* note : dctx->expected can grow seriously large, beyond local buffer size */
dctx->stage = ZSTDds_skipFrame;
return 0;
@@ -2191,21 +2228,27 @@ static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dict
return 0;
}
/* ZSTD_loadEntropy() :
* dict : must point at beginning of a valid zstd dictionary
/*! ZSTD_loadEntropy() :
* dict : must point at beginning of a valid zstd dictionary.
* @return : size of entropy tables read */
static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const dict, size_t const dictSize)
static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy,
const void* const dict, size_t const dictSize)
{
const BYTE* dictPtr = (const BYTE*)dict;
const BYTE* const dictEnd = dictPtr + dictSize;
if (dictSize <= 8) return ERROR(dictionary_corrupted);
assert(MEM_readLE32(dict) == ZSTD_MAGIC_DICTIONARY); /* dict must be valid */
dictPtr += 8; /* skip header = magic + dictID */
{ size_t const hSize = HUF_readDTableX2_wksp(
entropy->hufTable, dictPtr, dictEnd - dictPtr,
entropy->workspace, sizeof(entropy->workspace));
ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, OFTable) == offsetof(ZSTD_entropyDTables_t, LLTable) + sizeof(entropy->LLTable));
ZSTD_STATIC_ASSERT(offsetof(ZSTD_entropyDTables_t, MLTable) == offsetof(ZSTD_entropyDTables_t, OFTable) + sizeof(entropy->OFTable));
ZSTD_STATIC_ASSERT(sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable) >= HUF_DECOMPRESS_WORKSPACE_SIZE);
{ void* const workspace = &entropy->LLTable; /* use fse tables as temporary workspace; implies fse tables are grouped together */
size_t const workspaceSize = sizeof(entropy->LLTable) + sizeof(entropy->OFTable) + sizeof(entropy->MLTable);
size_t const hSize = HUF_readDTableX2_wksp(entropy->hufTable,
dictPtr, dictEnd - dictPtr,
workspace, workspaceSize);
if (HUF_isError(hSize)) return ERROR(dictionary_corrupted);
dictPtr += hSize;
}
@@ -2216,7 +2259,7 @@ static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const
if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
if (offcodeMaxValue > MaxOff) return ERROR(dictionary_corrupted);
if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
ZSTD_buildFSETable(entropy->OFTable,
ZSTD_buildFSETable( entropy->OFTable,
offcodeNCount, offcodeMaxValue,
OF_base, OF_bits,
offcodeLog);
@@ -2229,7 +2272,7 @@ static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const
if (FSE_isError(matchlengthHeaderSize)) return ERROR(dictionary_corrupted);
if (matchlengthMaxValue > MaxML) return ERROR(dictionary_corrupted);
if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
ZSTD_buildFSETable(entropy->MLTable,
ZSTD_buildFSETable( entropy->MLTable,
matchlengthNCount, matchlengthMaxValue,
ML_base, ML_bits,
matchlengthLog);
@@ -2242,7 +2285,7 @@ static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const
if (FSE_isError(litlengthHeaderSize)) return ERROR(dictionary_corrupted);
if (litlengthMaxValue > MaxLL) return ERROR(dictionary_corrupted);
if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
ZSTD_buildFSETable(entropy->LLTable,
ZSTD_buildFSETable( entropy->LLTable,
litlengthNCount, litlengthMaxValue,
LL_base, LL_bits,
litlengthLog);
@@ -2268,7 +2311,7 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
if (magic != ZSTD_MAGIC_DICTIONARY) {
return ZSTD_refDictContent(dctx, dict, dictSize); /* pure content mode */
} }
dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_frameIdSize);
dctx->dictID = MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
/* load entropy tables */
{ size_t const eSize = ZSTD_loadEntropy(&dctx->entropy, dict, dictSize);
@@ -2282,7 +2325,6 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
return ZSTD_refDictContent(dctx, dict, dictSize);
}
/* Note : this function cannot fail */
size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
{
assert(dctx != NULL);
@@ -2328,42 +2370,53 @@ struct ZSTD_DDict_s {
static const void* ZSTD_DDictDictContent(const ZSTD_DDict* ddict)
{
assert(ddict != NULL);
return ddict->dictContent;
}
static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict)
{
assert(ddict != NULL);
return ddict->dictSize;
}
size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict)
size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
{
CHECK_F( ZSTD_decompressBegin(dstDCtx) );
if (ddict) { /* support begin on NULL */
dstDCtx->dictID = ddict->dictID;
dstDCtx->prefixStart = ddict->dictContent;
dstDCtx->virtualStart = ddict->dictContent;
dstDCtx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
dstDCtx->previousDstEnd = dstDCtx->dictEnd;
DEBUGLOG(4, "ZSTD_decompressBegin_usingDDict");
assert(dctx != NULL);
if (ddict) {
dctx->ddictIsCold = (dctx->dictEnd != (const char*)ddict->dictContent + ddict->dictSize);
DEBUGLOG(4, "DDict is %s",
dctx->ddictIsCold ? "~cold~" : "hot!");
}
CHECK_F( ZSTD_decompressBegin(dctx) );
if (ddict) { /* NULL ddict is equivalent to no dictionary */
dctx->dictID = ddict->dictID;
dctx->prefixStart = ddict->dictContent;
dctx->virtualStart = ddict->dictContent;
dctx->dictEnd = (const BYTE*)ddict->dictContent + ddict->dictSize;
dctx->previousDstEnd = dctx->dictEnd;
if (ddict->entropyPresent) {
dstDCtx->litEntropy = 1;
dstDCtx->fseEntropy = 1;
dstDCtx->LLTptr = ddict->entropy.LLTable;
dstDCtx->MLTptr = ddict->entropy.MLTable;
dstDCtx->OFTptr = ddict->entropy.OFTable;
dstDCtx->HUFptr = ddict->entropy.hufTable;
dstDCtx->entropy.rep[0] = ddict->entropy.rep[0];
dstDCtx->entropy.rep[1] = ddict->entropy.rep[1];
dstDCtx->entropy.rep[2] = ddict->entropy.rep[2];
dctx->litEntropy = 1;
dctx->fseEntropy = 1;
dctx->LLTptr = ddict->entropy.LLTable;
dctx->MLTptr = ddict->entropy.MLTable;
dctx->OFTptr = ddict->entropy.OFTable;
dctx->HUFptr = ddict->entropy.hufTable;
dctx->entropy.rep[0] = ddict->entropy.rep[0];
dctx->entropy.rep[1] = ddict->entropy.rep[1];
dctx->entropy.rep[2] = ddict->entropy.rep[2];
} else {
dstDCtx->litEntropy = 0;
dstDCtx->fseEntropy = 0;
dctx->litEntropy = 0;
dctx->fseEntropy = 0;
}
}
return 0;
}
static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict, ZSTD_dictContentType_e dictContentType)
static size_t
ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict,
ZSTD_dictContentType_e dictContentType)
{
ddict->dictID = 0;
ddict->entropyPresent = 0;
@@ -2381,10 +2434,12 @@ static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict, ZSTD_dictContentType_e
return 0; /* pure content mode */
}
}
ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_frameIdSize);
ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + ZSTD_FRAMEIDSIZE);
/* load entropy tables */
CHECK_E( ZSTD_loadEntropy(&ddict->entropy, ddict->dictContent, ddict->dictSize), dictionary_corrupted );
CHECK_E( ZSTD_loadEntropy(&ddict->entropy,
ddict->dictContent, ddict->dictSize),
dictionary_corrupted );
ddict->entropyPresent = 1;
return 0;
}
@@ -2398,6 +2453,7 @@ static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) {
ddict->dictBuffer = NULL;
ddict->dictContent = dict;
if (!dict) dictSize = 0;
} else {
void* const internalBuffer = ZSTD_malloc(dictSize, ddict->cMem);
ddict->dictBuffer = internalBuffer;
@@ -2422,14 +2478,15 @@ ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
{ ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem);
if (!ddict) return NULL;
if (ddict == NULL) return NULL;
ddict->cMem = customMem;
if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, dictLoadMethod, dictContentType) )) {
ZSTD_freeDDict(ddict);
return NULL;
}
{ size_t const initResult = ZSTD_initDDict_internal(ddict,
dict, dictSize,
dictLoadMethod, dictContentType);
if (ZSTD_isError(initResult)) {
ZSTD_freeDDict(ddict);
return NULL;
} }
return ddict;
}
}
@@ -2456,23 +2513,25 @@ ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize
const ZSTD_DDict* ZSTD_initStaticDDict(
void* workspace, size_t workspaceSize,
void* sBuffer, size_t sBufferSize,
const void* dict, size_t dictSize,
ZSTD_dictLoadMethod_e dictLoadMethod,
ZSTD_dictContentType_e dictContentType)
{
size_t const neededSpace =
sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
ZSTD_DDict* const ddict = (ZSTD_DDict*)workspace;
assert(workspace != NULL);
size_t const neededSpace = sizeof(ZSTD_DDict)
+ (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
ZSTD_DDict* const ddict = (ZSTD_DDict*)sBuffer;
assert(sBuffer != NULL);
assert(dict != NULL);
if ((size_t)workspace & 7) return NULL; /* 8-aligned */
if (workspaceSize < neededSpace) return NULL;
if ((size_t)sBuffer & 7) return NULL; /* 8-aligned */
if (sBufferSize < neededSpace) return NULL;
if (dictLoadMethod == ZSTD_dlm_byCopy) {
memcpy(ddict+1, dict, dictSize); /* local copy */
dict = ddict+1;
}
if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, ZSTD_dlm_byRef, dictContentType) ))
if (ZSTD_isError( ZSTD_initDDict_internal(ddict,
dict, dictSize,
ZSTD_dlm_byRef, dictContentType) ))
return NULL;
return ddict;
}
@@ -2510,7 +2569,7 @@ unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
{
if (dictSize < 8) return 0;
if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0;
return MEM_readLE32((const char*)dict + ZSTD_frameIdSize);
return MEM_readLE32((const char*)dict + ZSTD_FRAMEIDSIZE);
}
/*! ZSTD_getDictID_fromDDict() :
@@ -2586,12 +2645,15 @@ size_t ZSTD_freeDStream(ZSTD_DStream* zds)
}
/* *** Initialization *** */
/* *** Initialization *** */
size_t ZSTD_DStreamInSize(void) { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; }
size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; }
size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType)
size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
const void* dict, size_t dictSize,
ZSTD_dictLoadMethod_e dictLoadMethod,
ZSTD_dictContentType_e dictContentType)
{
if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
ZSTD_freeDDict(dctx->ddictLocal);
@@ -2645,13 +2707,6 @@ size_t ZSTD_initDStream(ZSTD_DStream* zds)
return ZSTD_initDStream_usingDict(zds, NULL, 0);
}
size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
{
if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
dctx->ddict = ddict;
return 0;
}
/* ZSTD_initDStream_usingDDict() :
* ddict will just be referenced, and must outlive decompression session
* this function cannot fail */
@@ -2690,6 +2745,13 @@ size_t ZSTD_setDStreamParameter(ZSTD_DStream* dctx,
return 0;
}
size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict)
{
if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
dctx->ddict = ddict;
return 0;
}
size_t ZSTD_DCtx_setMaxWindowSize(ZSTD_DCtx* dctx, size_t maxWindowSize)
{
if (dctx->streamStage != zdss_init) return ERROR(stage_wrong);
@@ -2855,7 +2917,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
CHECK_F(ZSTD_decompressBegin_usingDDict(zds, zds->ddict));
if ((MEM_readLE32(zds->headerBuffer) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) { /* skippable frame */
zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_frameIdSize);
zds->expected = MEM_readLE32(zds->headerBuffer + ZSTD_FRAMEIDSIZE);
zds->stage = ZSTDds_skipFrame;
} else {
CHECK_F(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize));
lib/dictBuilder/cover.c
+67 -78
View File
@@ -29,6 +29,7 @@
#include "mem.h" /* read */
#include "pool.h"
#include "threading.h"
#include "cover.h"
#include "zstd_internal.h" /* includes zstd.h */
#ifndef ZDICT_STATIC_LINKING_ONLY
#define ZDICT_STATIC_LINKING_ONLY
@@ -185,7 +186,7 @@ static void COVER_map_remove(COVER_map_t *map, U32 key) {
}
/**
* Destroyes a map that is inited with COVER_map_init().
* Destroys a map that is inited with COVER_map_init().
*/
static void COVER_map_destroy(COVER_map_t *map) {
if (map->data) {
@@ -223,7 +224,7 @@ static COVER_ctx_t *g_ctx = NULL;
/**
* Returns the sum of the sample sizes.
*/
static size_t COVER_sum(const size_t *samplesSizes, unsigned nbSamples) {
size_t COVER_sum(const size_t *samplesSizes, unsigned nbSamples) {
size_t sum = 0;
unsigned i;
for (i = 0; i < nbSamples; ++i) {
@@ -380,14 +381,6 @@ static void COVER_group(COVER_ctx_t *ctx, const void *group,
ctx->suffix[dmerId] = freq;
}
/**
* A segment is a range in the source as well as the score of the segment.
*/
typedef struct {
U32 begin;
U32 end;
U32 score;
} COVER_segment_t;
/**
* Selects the best segment in an epoch.
@@ -737,28 +730,65 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
}
}
/**
* COVER_best_t is used for two purposes:
* 1. Synchronizing threads.
* 2. Saving the best parameters and dictionary.
*
* All of the methods except COVER_best_init() are thread safe if zstd is
* compiled with multithreaded support.
*/
typedef struct COVER_best_s {
ZSTD_pthread_mutex_t mutex;
ZSTD_pthread_cond_t cond;
size_t liveJobs;
void *dict;
size_t dictSize;
ZDICT_cover_params_t parameters;
size_t compressedSize;
} COVER_best_t;
size_t COVER_checkTotalCompressedSize(const ZDICT_cover_params_t parameters,
const size_t *samplesSizes, const BYTE *samples,
size_t *offsets,
size_t nbTrainSamples, size_t nbSamples,
BYTE *const dict, size_t dictBufferCapacity) {
size_t totalCompressedSize = ERROR(GENERIC);
/* Pointers */
ZSTD_CCtx *cctx;
ZSTD_CDict *cdict;
void *dst;
/* Local variables */
size_t dstCapacity;
size_t i;
/* Allocate dst with enough space to compress the maximum sized sample */
{
size_t maxSampleSize = 0;
i = parameters.splitPoint < 1.0 ? nbTrainSamples : 0;
for (; i < nbSamples; ++i) {
maxSampleSize = MAX(samplesSizes[i], maxSampleSize);
}
dstCapacity = ZSTD_compressBound(maxSampleSize);
dst = malloc(dstCapacity);
}
/* Create the cctx and cdict */
cctx = ZSTD_createCCtx();
cdict = ZSTD_createCDict(dict, dictBufferCapacity,
parameters.zParams.compressionLevel);
if (!dst || !cctx || !cdict) {
goto _compressCleanup;
}
/* Compress each sample and sum their sizes (or error) */
totalCompressedSize = dictBufferCapacity;
i = parameters.splitPoint < 1.0 ? nbTrainSamples : 0;
for (; i < nbSamples; ++i) {
const size_t size = ZSTD_compress_usingCDict(
cctx, dst, dstCapacity, samples + offsets[i],
samplesSizes[i], cdict);
if (ZSTD_isError(size)) {
totalCompressedSize = ERROR(GENERIC);
goto _compressCleanup;
}
totalCompressedSize += size;
}
_compressCleanup:
ZSTD_freeCCtx(cctx);
ZSTD_freeCDict(cdict);
if (dst) {
free(dst);
}
return totalCompressedSize;
}
/**
* Initialize the `COVER_best_t`.
*/
static void COVER_best_init(COVER_best_t *best) {
void COVER_best_init(COVER_best_t *best) {
if (best==NULL) return; /* compatible with init on NULL */
(void)ZSTD_pthread_mutex_init(&best->mutex, NULL);
(void)ZSTD_pthread_cond_init(&best->cond, NULL);
@@ -772,7 +802,7 @@ static void COVER_best_init(COVER_best_t *best) {
/**
* Wait until liveJobs == 0.
*/
static void COVER_best_wait(COVER_best_t *best) {
void COVER_best_wait(COVER_best_t *best) {
if (!best) {
return;
}
@@ -786,7 +816,7 @@ static void COVER_best_wait(COVER_best_t *best) {
/**
* Call COVER_best_wait() and then destroy the COVER_best_t.
*/
static void COVER_best_destroy(COVER_best_t *best) {
void COVER_best_destroy(COVER_best_t *best) {
if (!best) {
return;
}
@@ -802,7 +832,7 @@ static void COVER_best_destroy(COVER_best_t *best) {
* Called when a thread is about to be launched.
* Increments liveJobs.
*/
static void COVER_best_start(COVER_best_t *best) {
void COVER_best_start(COVER_best_t *best) {
if (!best) {
return;
}
@@ -816,7 +846,7 @@ static void COVER_best_start(COVER_best_t *best) {
* Decrements liveJobs and signals any waiting threads if liveJobs == 0.
* If this dictionary is the best so far save it and its parameters.
*/
static void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
ZDICT_cover_params_t parameters, void *dict,
size_t dictSize) {
if (!best) {
@@ -847,10 +877,10 @@ static void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
best->parameters = parameters;
best->compressedSize = compressedSize;
}
ZSTD_pthread_mutex_unlock(&best->mutex);
if (liveJobs == 0) {
ZSTD_pthread_cond_broadcast(&best->cond);
}
ZSTD_pthread_mutex_unlock(&best->mutex);
}
}
@@ -904,51 +934,10 @@ static void COVER_tryParameters(void *opaque) {
}
}
/* Check total compressed size */
{
/* Pointers */
ZSTD_CCtx *cctx;
ZSTD_CDict *cdict;
void *dst;
/* Local variables */
size_t dstCapacity;
size_t i;
/* Allocate dst with enough space to compress the maximum sized sample */
{
size_t maxSampleSize = 0;
i = parameters.splitPoint < 1.0 ? ctx->nbTrainSamples : 0;
for (; i < ctx->nbSamples; ++i) {
maxSampleSize = MAX(ctx->samplesSizes[i], maxSampleSize);
}
dstCapacity = ZSTD_compressBound(maxSampleSize);
dst = malloc(dstCapacity);
}
/* Create the cctx and cdict */
cctx = ZSTD_createCCtx();
cdict = ZSTD_createCDict(dict, dictBufferCapacity,
parameters.zParams.compressionLevel);
if (!dst || !cctx || !cdict) {
goto _compressCleanup;
}
/* Compress each sample and sum their sizes (or error) */
totalCompressedSize = dictBufferCapacity;
i = parameters.splitPoint < 1.0 ? ctx->nbTrainSamples : 0;
for (; i < ctx->nbSamples; ++i) {
const size_t size = ZSTD_compress_usingCDict(
cctx, dst, dstCapacity, ctx->samples + ctx->offsets[i],
ctx->samplesSizes[i], cdict);
if (ZSTD_isError(size)) {
totalCompressedSize = ERROR(GENERIC);
goto _compressCleanup;
}
totalCompressedSize += size;
}
_compressCleanup:
ZSTD_freeCCtx(cctx);
ZSTD_freeCDict(cdict);
if (dst) {
free(dst);
}
}
totalCompressedSize = COVER_checkTotalCompressedSize(parameters, ctx->samplesSizes,
ctx->samples, ctx->offsets,
ctx->nbTrainSamples, ctx->nbSamples,
dict, dictBufferCapacity);
_cleanup:
COVER_best_finish(data->best, totalCompressedSize, parameters, dict,
lib/dictBuilder/cover.h
+83
View File
@@ -0,0 +1,83 @@
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* memset */
#include <time.h> /* clock */
#include "mem.h" /* read */
#include "pool.h"
#include "threading.h"
#include "zstd_internal.h" /* includes zstd.h */
#ifndef ZDICT_STATIC_LINKING_ONLY
#define ZDICT_STATIC_LINKING_ONLY
#endif
#include "zdict.h"
/**
* COVER_best_t is used for two purposes:
* 1. Synchronizing threads.
* 2. Saving the best parameters and dictionary.
*
* All of the methods except COVER_best_init() are thread safe if zstd is
* compiled with multithreaded support.
*/
typedef struct COVER_best_s {
ZSTD_pthread_mutex_t mutex;
ZSTD_pthread_cond_t cond;
size_t liveJobs;
void *dict;
size_t dictSize;
ZDICT_cover_params_t parameters;
size_t compressedSize;
} COVER_best_t;
/**
* A segment is a range in the source as well as the score of the segment.
*/
typedef struct {
U32 begin;
U32 end;
U32 score;
} COVER_segment_t;
/**
* Checks total compressed size of a dictionary
*/
size_t COVER_checkTotalCompressedSize(const ZDICT_cover_params_t parameters,
const size_t *samplesSizes, const BYTE *samples,
size_t *offsets,
size_t nbTrainSamples, size_t nbSamples,
BYTE *const dict, size_t dictBufferCapacity);
/**
* Returns the sum of the sample sizes.
*/
size_t COVER_sum(const size_t *samplesSizes, unsigned nbSamples) ;
/**
* Initialize the `COVER_best_t`.
*/
void COVER_best_init(COVER_best_t *best);
/**
* Wait until liveJobs == 0.
*/
void COVER_best_wait(COVER_best_t *best);
/**
* Call COVER_best_wait() and then destroy the COVER_best_t.
*/
void COVER_best_destroy(COVER_best_t *best);
/**
* Called when a thread is about to be launched.
* Increments liveJobs.
*/
void COVER_best_start(COVER_best_t *best);
/**
* Called when a thread finishes executing, both on error or success.
* Decrements liveJobs and signals any waiting threads if liveJobs == 0.
* If this dictionary is the best so far save it and its parameters.
*/
void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
ZDICT_cover_params_t parameters, void *dict,
size_t dictSize);
lib/dictBuilder/divsufsort.c
+3 -3
View File
@@ -1637,7 +1637,7 @@ construct_SA(const unsigned char *T, int *SA,
if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
k = SA + BUCKET_B(c2 = c0, c1);
}
assert(k < j);
assert(k < j); assert(k != NULL);
*k-- = s;
} else {
assert(((s == 0) && (T[s] == c1)) || (s < 0));
@@ -1701,7 +1701,7 @@ construct_BWT(const unsigned char *T, int *SA,
if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
k = SA + BUCKET_B(c2 = c0, c1);
}
assert(k < j);
assert(k < j); assert(k != NULL);
*k-- = s;
} else if(s != 0) {
*j = ~s;
@@ -1785,7 +1785,7 @@ construct_BWT_indexes(const unsigned char *T, int *SA,
if(0 <= c2) { BUCKET_B(c2, c1) = k - SA; }
k = SA + BUCKET_B(c2 = c0, c1);
}
assert(k < j);
assert(k < j); assert(k != NULL);
*k-- = s;
} else if(s != 0) {
*j = ~s;
lib/dictBuilder/fastcover.c
+701
View File
@@ -0,0 +1,701 @@
/*-*************************************
* Dependencies
***************************************/
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* memset */
#include <time.h> /* clock */
#include "mem.h" /* read */
#include "pool.h"
#include "threading.h"
#include "cover.h"
#include "zstd_internal.h" /* includes zstd.h */
#ifndef ZDICT_STATIC_LINKING_ONLY
#define ZDICT_STATIC_LINKING_ONLY
#endif
#include "zdict.h"
/*-*************************************
* Constants
***************************************/
#define FASTCOVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((U32)-1) : ((U32)1 GB))
#define FASTCOVER_MAX_F 31
#define FASTCOVER_MAX_ACCEL 10
#define DEFAULT_SPLITPOINT 0.75
#define DEFAULT_F 20
#define DEFAULT_ACCEL 1
/*-*************************************
* Console display
***************************************/
static int g_displayLevel = 2;
#define DISPLAY(...) \
{ \
fprintf(stderr, __VA_ARGS__); \
fflush(stderr); \
}
#define LOCALDISPLAYLEVEL(displayLevel, l, ...) \
if (displayLevel >= l) { \
DISPLAY(__VA_ARGS__); \
} /* 0 : no display; 1: errors; 2: default; 3: details; 4: debug */
#define DISPLAYLEVEL(l, ...) LOCALDISPLAYLEVEL(g_displayLevel, l, __VA_ARGS__)
#define LOCALDISPLAYUPDATE(displayLevel, l, ...) \
if (displayLevel >= l) { \
if ((clock() - g_time > refreshRate) || (displayLevel >= 4)) { \
g_time = clock(); \
DISPLAY(__VA_ARGS__); \
} \
}
#define DISPLAYUPDATE(l, ...) LOCALDISPLAYUPDATE(g_displayLevel, l, __VA_ARGS__)
static const clock_t refreshRate = CLOCKS_PER_SEC * 15 / 100;
static clock_t g_time = 0;
/*-*************************************
* Hash Functions
***************************************/
static const U64 prime6bytes = 227718039650203ULL;
static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u << (64-48)) * prime6bytes) >> (64-h)) ; }
static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
/**
* Hash the d-byte value pointed to by p and mod 2^f
*/
static size_t FASTCOVER_hashPtrToIndex(const void* p, U32 h, unsigned d) {
if (d == 6) {
return ZSTD_hash6Ptr(p, h) & ((1 << h) - 1);
}
return ZSTD_hash8Ptr(p, h) & ((1 << h) - 1);
}
/*-*************************************
* Acceleration
***************************************/
typedef struct {
unsigned finalize; /* Percentage of training samples used for ZDICT_finalizeDictionary */
unsigned skip; /* Number of dmer skipped between each dmer counted in computeFrequency */
} FASTCOVER_accel_t;
static const FASTCOVER_accel_t FASTCOVER_defaultAccelParameters[FASTCOVER_MAX_ACCEL+1] = {
{ 100, 0 }, /* accel = 0, should not happen because accel = 0 defaults to accel = 1 */
{ 100, 0 }, /* accel = 1 */
{ 50, 1 }, /* accel = 2 */
{ 34, 2 }, /* accel = 3 */
{ 25, 3 }, /* accel = 4 */
{ 20, 4 }, /* accel = 5 */
{ 17, 5 }, /* accel = 6 */
{ 14, 6 }, /* accel = 7 */
{ 13, 7 }, /* accel = 8 */
{ 11, 8 }, /* accel = 9 */
{ 10, 9 }, /* accel = 10 */
};
/*-*************************************
* Context
***************************************/
typedef struct {
const BYTE *samples;
size_t *offsets;
const size_t *samplesSizes;
size_t nbSamples;
size_t nbTrainSamples;
size_t nbTestSamples;
size_t nbDmers;
U32 *freqs;
unsigned d;
unsigned f;
FASTCOVER_accel_t accelParams;
} FASTCOVER_ctx_t;
/*-*************************************
* Helper functions
***************************************/
/**
* Selects the best segment in an epoch.
* Segments of are scored according to the function:
*
* Let F(d) be the frequency of all dmers with hash value d.
* Let S_i be hash value of the dmer at position i of segment S which has length k.
*
* Score(S) = F(S_1) + F(S_2) + ... + F(S_{k-d+1})
*
* Once the dmer with hash value d is in the dictionay we set F(d) = 0.
*/
static COVER_segment_t FASTCOVER_selectSegment(const FASTCOVER_ctx_t *ctx,
U32 *freqs, U32 begin, U32 end,
ZDICT_cover_params_t parameters,
U16* segmentFreqs) {
/* Constants */
const U32 k = parameters.k;
const U32 d = parameters.d;
const U32 f = ctx->f;
const U32 dmersInK = k - d + 1;
/* Try each segment (activeSegment) and save the best (bestSegment) */
COVER_segment_t bestSegment = {0, 0, 0};
COVER_segment_t activeSegment;
/* Reset the activeDmers in the segment */
/* The activeSegment starts at the beginning of the epoch. */
activeSegment.begin = begin;
activeSegment.end = begin;
activeSegment.score = 0;
/* Slide the activeSegment through the whole epoch.
* Save the best segment in bestSegment.
*/
while (activeSegment.end < end) {
/* Get hash value of current dmer */
const size_t index = FASTCOVER_hashPtrToIndex(ctx->samples + activeSegment.end, f, d);
/* Add frequency of this index to score if this is the first occurence of index in active segment */
if (segmentFreqs[index] == 0) {
activeSegment.score += freqs[index];
}
/* Increment end of segment and segmentFreqs*/
activeSegment.end += 1;
segmentFreqs[index] += 1;
/* If the window is now too large, drop the first position */
if (activeSegment.end - activeSegment.begin == dmersInK + 1) {
/* Get hash value of the dmer to be eliminated from active segment */
const size_t delIndex = FASTCOVER_hashPtrToIndex(ctx->samples + activeSegment.begin, f, d);
segmentFreqs[delIndex] -= 1;
/* Subtract frequency of this index from score if this is the last occurrence of this index in active segment */
if (segmentFreqs[delIndex] == 0) {
activeSegment.score -= freqs[delIndex];
}
/* Increment start of segment */
activeSegment.begin += 1;
}
/* If this segment is the best so far save it */
if (activeSegment.score > bestSegment.score) {
bestSegment = activeSegment;
}
}
/* Zero out rest of segmentFreqs array */
while (activeSegment.begin < end) {
const size_t delIndex = FASTCOVER_hashPtrToIndex(ctx->samples + activeSegment.begin, f, d);
segmentFreqs[delIndex] -= 1;
activeSegment.begin += 1;
}
{
/* Zero the frequency of hash value of each dmer covered by the chosen segment. */
U32 pos;
for (pos = bestSegment.begin; pos != bestSegment.end; ++pos) {
const size_t i = FASTCOVER_hashPtrToIndex(ctx->samples + pos, f, d);
freqs[i] = 0;
}
}
return bestSegment;
}
static int FASTCOVER_checkParameters(ZDICT_cover_params_t parameters,
size_t maxDictSize, unsigned f,
unsigned accel) {
/* k, d, and f are required parameters */
if (parameters.d == 0 || parameters.k == 0) {
return 0;
}
/* d has to be 6 or 8 */
if (parameters.d != 6 && parameters.d != 8) {
return 0;
}
/* k <= maxDictSize */
if (parameters.k > maxDictSize) {
return 0;
}
/* d <= k */
if (parameters.d > parameters.k) {
return 0;
}
/* 0 < f <= FASTCOVER_MAX_F*/
if (f > FASTCOVER_MAX_F || f == 0) {
return 0;
}
/* 0 < splitPoint <= 1 */
if (parameters.splitPoint <= 0 || parameters.splitPoint > 1) {
return 0;
}
/* 0 < accel <= 10 */
if (accel > 10 || accel == 0) {
return 0;
}
return 1;
}
/**
* Clean up a context initialized with `FASTCOVER_ctx_init()`.
*/
static void FASTCOVER_ctx_destroy(FASTCOVER_ctx_t *ctx) {
if (!ctx) {
return;
}
free(ctx->freqs);
ctx->freqs = NULL;
free(ctx->offsets);
ctx->offsets = NULL;
}
/**
* Calculate for frequency of hash value of each dmer in ctx->samples
*/
static void FASTCOVER_computeFrequency(U32 *freqs, FASTCOVER_ctx_t *ctx){
const unsigned f = ctx->f;
const unsigned d = ctx->d;
const unsigned skip = ctx->accelParams.skip;
const unsigned readLength = MAX(d, 8);
size_t start; /* start of current dmer */
size_t i;
for (i = 0; i < ctx->nbTrainSamples; i++) {
size_t currSampleStart = ctx->offsets[i];
size_t currSampleEnd = ctx->offsets[i+1];
start = currSampleStart;
while (start + readLength <= currSampleEnd) {
const size_t dmerIndex = FASTCOVER_hashPtrToIndex(ctx->samples + start, f, d);
freqs[dmerIndex]++;
start = start + skip + 1;
}
}
}
/**
* Prepare a context for dictionary building.
* The context is only dependent on the parameter `d` and can used multiple
* times.
* Returns 1 on success or zero on error.
* The context must be destroyed with `FASTCOVER_ctx_destroy()`.
*/
static int FASTCOVER_ctx_init(FASTCOVER_ctx_t *ctx, const void *samplesBuffer,
const size_t *samplesSizes, unsigned nbSamples,
unsigned d, double splitPoint, unsigned f,
FASTCOVER_accel_t accelParams) {
const BYTE *const samples = (const BYTE *)samplesBuffer;
const size_t totalSamplesSize = COVER_sum(samplesSizes, nbSamples);
/* Split samples into testing and training sets */
const unsigned nbTrainSamples = splitPoint < 1.0 ? (unsigned)((double)nbSamples * splitPoint) : nbSamples;
const unsigned nbTestSamples = splitPoint < 1.0 ? nbSamples - nbTrainSamples : nbSamples;
const size_t trainingSamplesSize = splitPoint < 1.0 ? COVER_sum(samplesSizes, nbTrainSamples) : totalSamplesSize;
const size_t testSamplesSize = splitPoint < 1.0 ? COVER_sum(samplesSizes + nbTrainSamples, nbTestSamples) : totalSamplesSize;
/* Checks */
if (totalSamplesSize < MAX(d, sizeof(U64)) ||
totalSamplesSize >= (size_t)FASTCOVER_MAX_SAMPLES_SIZE) {
DISPLAYLEVEL(1, "Total samples size is too large (%u MB), maximum size is %u MB\n",
(U32)(totalSamplesSize >> 20), (FASTCOVER_MAX_SAMPLES_SIZE >> 20));
return 0;
}
/* Check if there are at least 5 training samples */
if (nbTrainSamples < 5) {
DISPLAYLEVEL(1, "Total number of training samples is %u and is invalid\n", nbTrainSamples);
return 0;
}
/* Check if there's testing sample */
if (nbTestSamples < 1) {
DISPLAYLEVEL(1, "Total number of testing samples is %u and is invalid.\n", nbTestSamples);
return 0;
}
/* Zero the context */
memset(ctx, 0, sizeof(*ctx));
DISPLAYLEVEL(2, "Training on %u samples of total size %u\n", nbTrainSamples,
(U32)trainingSamplesSize);
DISPLAYLEVEL(2, "Testing on %u samples of total size %u\n", nbTestSamples,
(U32)testSamplesSize);
ctx->samples = samples;
ctx->samplesSizes = samplesSizes;
ctx->nbSamples = nbSamples;
ctx->nbTrainSamples = nbTrainSamples;
ctx->nbTestSamples = nbTestSamples;
ctx->nbDmers = trainingSamplesSize - MAX(d, sizeof(U64)) + 1;
ctx->d = d;
ctx->f = f;
ctx->accelParams = accelParams;
/* The offsets of each file */
ctx->offsets = (size_t *)malloc((nbSamples + 1) * sizeof(size_t));
if (!ctx->offsets) {
DISPLAYLEVEL(1, "Failed to allocate scratch buffers\n");
FASTCOVER_ctx_destroy(ctx);
return 0;
}
/* Fill offsets from the samplesSizes */
{
U32 i;
ctx->offsets[0] = 0;
for (i = 1; i <= nbSamples; ++i) {
ctx->offsets[i] = ctx->offsets[i - 1] + samplesSizes[i - 1];
}
}
/* Initialize frequency array of size 2^f */
ctx->freqs = (U32 *)calloc(((U64)1 << f), sizeof(U32));
DISPLAYLEVEL(2, "Computing frequencies\n");
FASTCOVER_computeFrequency(ctx->freqs, ctx);
return 1;
}
/**
* Given the prepared context build the dictionary.
*/
static size_t FASTCOVER_buildDictionary(const FASTCOVER_ctx_t *ctx, U32 *freqs,
void *dictBuffer, size_t dictBufferCapacity,
ZDICT_cover_params_t parameters, U16* segmentFreqs){
BYTE *const dict = (BYTE *)dictBuffer;
size_t tail = dictBufferCapacity;
/* Divide the data up into epochs of equal size.
* We will select at least one segment from each epoch.
*/
const U32 epochs = MAX(1, (U32)(dictBufferCapacity / parameters.k));
const U32 epochSize = (U32)(ctx->nbDmers / epochs);
size_t epoch;
DISPLAYLEVEL(2, "Breaking content into %u epochs of size %u\n", epochs,
epochSize);
/* Loop through the epochs until there are no more segments or the dictionary
* is full.
*/
for (epoch = 0; tail > 0; epoch = (epoch + 1) % epochs) {
const U32 epochBegin = (U32)(epoch * epochSize);
const U32 epochEnd = epochBegin + epochSize;
size_t segmentSize;
/* Select a segment */
COVER_segment_t segment = FASTCOVER_selectSegment(
ctx, freqs, epochBegin, epochEnd, parameters, segmentFreqs);
/* If the segment covers no dmers, then we are out of content */
if (segment.score == 0) {
break;
}
/* Trim the segment if necessary and if it is too small then we are done */
segmentSize = MIN(segment.end - segment.begin + parameters.d - 1, tail);
if (segmentSize < parameters.d) {
break;
}
/* We fill the dictionary from the back to allow the best segments to be
* referenced with the smallest offsets.
*/
tail -= segmentSize;
memcpy(dict + tail, ctx->samples + segment.begin, segmentSize);
DISPLAYUPDATE(
2, "\r%u%% ",
(U32)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
}
DISPLAYLEVEL(2, "\r%79s\r", "");
return tail;
}
/**
* Parameters for FASTCOVER_tryParameters().
*/
typedef struct FASTCOVER_tryParameters_data_s {
const FASTCOVER_ctx_t *ctx;
COVER_best_t *best;
size_t dictBufferCapacity;
ZDICT_cover_params_t parameters;
} FASTCOVER_tryParameters_data_t;
/**
* Tries a set of parameters and updates the COVER_best_t with the results.
* This function is thread safe if zstd is compiled with multithreaded support.
* It takes its parameters as an *OWNING* opaque pointer to support threading.
*/
static void FASTCOVER_tryParameters(void *opaque) {
/* Save parameters as local variables */
FASTCOVER_tryParameters_data_t *const data = (FASTCOVER_tryParameters_data_t *)opaque;
const FASTCOVER_ctx_t *const ctx = data->ctx;
const ZDICT_cover_params_t parameters = data->parameters;
size_t dictBufferCapacity = data->dictBufferCapacity;
size_t totalCompressedSize = ERROR(GENERIC);
/* Initialize array to keep track of frequency of dmer within activeSegment */
U16* segmentFreqs = (U16 *)calloc(((U64)1 << ctx->f), sizeof(U16));
/* Allocate space for hash table, dict, and freqs */
BYTE *const dict = (BYTE * const)malloc(dictBufferCapacity);
U32 *freqs = (U32*) malloc(((U64)1 << ctx->f) * sizeof(U32));
if (!segmentFreqs || !dict || !freqs) {
DISPLAYLEVEL(1, "Failed to allocate buffers: out of memory\n");
goto _cleanup;
}
/* Copy the frequencies because we need to modify them */
memcpy(freqs, ctx->freqs, ((U64)1 << ctx->f) * sizeof(U32));
/* Build the dictionary */
{
const size_t tail = FASTCOVER_buildDictionary(ctx, freqs, dict, dictBufferCapacity,
parameters, segmentFreqs);
const unsigned nbFinalizeSamples = (unsigned)(ctx->nbTrainSamples * ctx->accelParams.finalize / 100);
dictBufferCapacity = ZDICT_finalizeDictionary(
dict, dictBufferCapacity, dict + tail, dictBufferCapacity - tail,
ctx->samples, ctx->samplesSizes, nbFinalizeSamples, parameters.zParams);
if (ZDICT_isError(dictBufferCapacity)) {
DISPLAYLEVEL(1, "Failed to finalize dictionary\n");
goto _cleanup;
}
}
/* Check total compressed size */
totalCompressedSize = COVER_checkTotalCompressedSize(parameters, ctx->samplesSizes,
ctx->samples, ctx->offsets,
ctx->nbTrainSamples, ctx->nbSamples,
dict, dictBufferCapacity);
_cleanup:
COVER_best_finish(data->best, totalCompressedSize, parameters, dict,
dictBufferCapacity);
free(data);
free(segmentFreqs);
free(dict);
free(freqs);
}
static void FASTCOVER_convertToCoverParams(ZDICT_fastCover_params_t fastCoverParams,
ZDICT_cover_params_t *coverParams) {
coverParams->k = fastCoverParams.k;
coverParams->d = fastCoverParams.d;
coverParams->steps = fastCoverParams.steps;
coverParams->nbThreads = fastCoverParams.nbThreads;
coverParams->splitPoint = fastCoverParams.splitPoint;
coverParams->zParams = fastCoverParams.zParams;
}
static void FASTCOVER_convertToFastCoverParams(ZDICT_cover_params_t coverParams,
ZDICT_fastCover_params_t *fastCoverParams,
unsigned f, unsigned accel) {
fastCoverParams->k = coverParams.k;
fastCoverParams->d = coverParams.d;
fastCoverParams->steps = coverParams.steps;
fastCoverParams->nbThreads = coverParams.nbThreads;
fastCoverParams->splitPoint = coverParams.splitPoint;
fastCoverParams->f = f;
fastCoverParams->accel = accel;
fastCoverParams->zParams = coverParams.zParams;
}
ZDICTLIB_API size_t ZDICT_trainFromBuffer_fastCover(
void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
const size_t *samplesSizes, unsigned nbSamples, ZDICT_fastCover_params_t parameters) {
BYTE* const dict = (BYTE*)dictBuffer;
FASTCOVER_ctx_t ctx;
ZDICT_cover_params_t coverParams;
FASTCOVER_accel_t accelParams;
/* Initialize global data */
g_displayLevel = parameters.zParams.notificationLevel;
/* Assign splitPoint and f if not provided */
parameters.splitPoint = 1.0;
parameters.f = parameters.f == 0 ? DEFAULT_F : parameters.f;
parameters.accel = parameters.accel == 0 ? DEFAULT_ACCEL : parameters.accel;
/* Convert to cover parameter */
memset(&coverParams, 0 , sizeof(coverParams));
FASTCOVER_convertToCoverParams(parameters, &coverParams);
/* Checks */
if (!FASTCOVER_checkParameters(coverParams, dictBufferCapacity, parameters.f,
parameters.accel)) {
DISPLAYLEVEL(1, "FASTCOVER parameters incorrect\n");
return ERROR(GENERIC);
}
if (nbSamples == 0) {
DISPLAYLEVEL(1, "FASTCOVER must have at least one input file\n");
return ERROR(GENERIC);
}
if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) {
DISPLAYLEVEL(1, "dictBufferCapacity must be at least %u\n",
ZDICT_DICTSIZE_MIN);
return ERROR(dstSize_tooSmall);
}
/* Assign corresponding FASTCOVER_accel_t to accelParams*/
accelParams = FASTCOVER_defaultAccelParameters[parameters.accel];
/* Initialize context */
if (!FASTCOVER_ctx_init(&ctx, samplesBuffer, samplesSizes, nbSamples,
coverParams.d, parameters.splitPoint, parameters.f,
accelParams)) {
DISPLAYLEVEL(1, "Failed to initialize context\n");
return ERROR(GENERIC);
}
/* Build the dictionary */
DISPLAYLEVEL(2, "Building dictionary\n");
{
/* Initialize array to keep track of frequency of dmer within activeSegment */
U16* segmentFreqs = (U16 *)calloc(((U64)1 << parameters.f), sizeof(U16));
const size_t tail = FASTCOVER_buildDictionary(&ctx, ctx.freqs, dictBuffer,
dictBufferCapacity, coverParams, segmentFreqs);
const unsigned nbFinalizeSamples = (unsigned)(ctx.nbTrainSamples * ctx.accelParams.finalize / 100);
const size_t dictionarySize = ZDICT_finalizeDictionary(
dict, dictBufferCapacity, dict + tail, dictBufferCapacity - tail,
samplesBuffer, samplesSizes, nbFinalizeSamples, coverParams.zParams);
if (!ZSTD_isError(dictionarySize)) {
DISPLAYLEVEL(2, "Constructed dictionary of size %u\n",
(U32)dictionarySize);
}
FASTCOVER_ctx_destroy(&ctx);
free(segmentFreqs);
return dictionarySize;
}
}
ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_fastCover(
void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
const size_t *samplesSizes, unsigned nbSamples,
ZDICT_fastCover_params_t *parameters) {
ZDICT_cover_params_t coverParams;
FASTCOVER_accel_t accelParams;
/* constants */
const unsigned nbThreads = parameters->nbThreads;
const double splitPoint =
parameters->splitPoint <= 0.0 ? DEFAULT_SPLITPOINT : parameters->splitPoint;
const unsigned kMinD = parameters->d == 0 ? 6 : parameters->d;
const unsigned kMaxD = parameters->d == 0 ? 8 : parameters->d;
const unsigned kMinK = parameters->k == 0 ? 50 : parameters->k;
const unsigned kMaxK = parameters->k == 0 ? 2000 : parameters->k;
const unsigned kSteps = parameters->steps == 0 ? 40 : parameters->steps;
const unsigned kStepSize = MAX((kMaxK - kMinK) / kSteps, 1);
const unsigned kIterations =
(1 + (kMaxD - kMinD) / 2) * (1 + (kMaxK - kMinK) / kStepSize);
const unsigned f = parameters->f == 0 ? DEFAULT_F : parameters->f;
const unsigned accel = parameters->accel == 0 ? DEFAULT_ACCEL : parameters->accel;
/* Local variables */
const int displayLevel = parameters->zParams.notificationLevel;
unsigned iteration = 1;
unsigned d;
unsigned k;
COVER_best_t best;
POOL_ctx *pool = NULL;
/* Checks */
if (splitPoint <= 0 || splitPoint > 1) {
LOCALDISPLAYLEVEL(displayLevel, 1, "Incorrect splitPoint\n");
return ERROR(GENERIC);
}
if (accel == 0 || accel > FASTCOVER_MAX_ACCEL) {
LOCALDISPLAYLEVEL(displayLevel, 1, "Incorrect accel\n");
return ERROR(GENERIC);
}
if (kMinK < kMaxD || kMaxK < kMinK) {
LOCALDISPLAYLEVEL(displayLevel, 1, "Incorrect k\n");
return ERROR(GENERIC);
}
if (nbSamples == 0) {
LOCALDISPLAYLEVEL(displayLevel, 1, "FASTCOVER must have at least one input file\n");
return ERROR(GENERIC);
}
if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) {
LOCALDISPLAYLEVEL(displayLevel, 1, "dictBufferCapacity must be at least %u\n",
ZDICT_DICTSIZE_MIN);
return ERROR(dstSize_tooSmall);
}
if (nbThreads > 1) {
pool = POOL_create(nbThreads, 1);
if (!pool) {
return ERROR(memory_allocation);
}
}
/* Initialization */
COVER_best_init(&best);
memset(&coverParams, 0 , sizeof(coverParams));
FASTCOVER_convertToCoverParams(*parameters, &coverParams);
accelParams = FASTCOVER_defaultAccelParameters[accel];
/* Turn down global display level to clean up display at level 2 and below */
g_displayLevel = displayLevel == 0 ? 0 : displayLevel - 1;
/* Loop through d first because each new value needs a new context */
LOCALDISPLAYLEVEL(displayLevel, 2, "Trying %u different sets of parameters\n",
kIterations);
for (d = kMinD; d <= kMaxD; d += 2) {
/* Initialize the context for this value of d */
FASTCOVER_ctx_t ctx;
LOCALDISPLAYLEVEL(displayLevel, 3, "d=%u\n", d);
if (!FASTCOVER_ctx_init(&ctx, samplesBuffer, samplesSizes, nbSamples, d, splitPoint, f, accelParams)) {
LOCALDISPLAYLEVEL(displayLevel, 1, "Failed to initialize context\n");
COVER_best_destroy(&best);
POOL_free(pool);
return ERROR(GENERIC);
}
/* Loop through k reusing the same context */
for (k = kMinK; k <= kMaxK; k += kStepSize) {
/* Prepare the arguments */
FASTCOVER_tryParameters_data_t *data = (FASTCOVER_tryParameters_data_t *)malloc(
sizeof(FASTCOVER_tryParameters_data_t));
LOCALDISPLAYLEVEL(displayLevel, 3, "k=%u\n", k);
if (!data) {
LOCALDISPLAYLEVEL(displayLevel, 1, "Failed to allocate parameters\n");
COVER_best_destroy(&best);
FASTCOVER_ctx_destroy(&ctx);
POOL_free(pool);
return ERROR(GENERIC);
}
data->ctx = &ctx;
data->best = &best;
data->dictBufferCapacity = dictBufferCapacity;
data->parameters = coverParams;
data->parameters.k = k;
data->parameters.d = d;
data->parameters.splitPoint = splitPoint;
data->parameters.steps = kSteps;
data->parameters.zParams.notificationLevel = g_displayLevel;
/* Check the parameters */
if (!FASTCOVER_checkParameters(data->parameters, dictBufferCapacity,
data->ctx->f, accel)) {
DISPLAYLEVEL(1, "FASTCOVER parameters incorrect\n");
free(data);
continue;
}
/* Call the function and pass ownership of data to it */
COVER_best_start(&best);
if (pool) {
POOL_add(pool, &FASTCOVER_tryParameters, data);
} else {
FASTCOVER_tryParameters(data);
}
/* Print status */
LOCALDISPLAYUPDATE(displayLevel, 2, "\r%u%% ",
(U32)((iteration * 100) / kIterations));
++iteration;
}
COVER_best_wait(&best);
FASTCOVER_ctx_destroy(&ctx);
}
LOCALDISPLAYLEVEL(displayLevel, 2, "\r%79s\r", "");
/* Fill the output buffer and parameters with output of the best parameters */
{
const size_t dictSize = best.dictSize;
if (ZSTD_isError(best.compressedSize)) {
const size_t compressedSize = best.compressedSize;
COVER_best_destroy(&best);
POOL_free(pool);
return compressedSize;
}
FASTCOVER_convertToFastCoverParams(best.parameters, parameters, f, accel);
memcpy(dictBuffer, best.dict, dictSize);
COVER_best_destroy(&best);
POOL_free(pool);
return dictSize;
}
}
lib/dictBuilder/zdict.c
+9 -7
View File
@@ -698,7 +698,7 @@ static size_t ZDICT_analyzeEntropy(void* dstBuffer, size_t maxDstSize,
short litLengthNCount[MaxLL+1];
U32 repOffset[MAXREPOFFSET];
offsetCount_t bestRepOffset[ZSTD_REP_NUM+1];
EStats_ress_t esr;
EStats_ress_t esr = { NULL, NULL, NULL };
ZSTD_parameters params;
U32 u, huffLog = 11, Offlog = OffFSELog, mlLog = MLFSELog, llLog = LLFSELog, total;
size_t pos = 0, errorCode;
@@ -863,8 +863,8 @@ _cleanup:
size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
const void* customDictContent, size_t dictContentSize,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
ZDICT_params_t params)
const void* samplesBuffer, const size_t* samplesSizes,
unsigned nbSamples, ZDICT_params_t params)
{
size_t hSize;
#define HBUFFSIZE 256 /* should prove large enough for all entropy headers */
@@ -987,8 +987,10 @@ size_t ZDICT_trainFromBuffer_unsafe_legacy(
U32 const pos = dictList[u].pos;
U32 const length = dictList[u].length;
U32 const printedLength = MIN(40, length);
if ((pos > samplesBuffSize) || ((pos + length) > samplesBuffSize))
if ((pos > samplesBuffSize) || ((pos + length) > samplesBuffSize)) {
free(dictList);
return ERROR(GENERIC); /* should never happen */
}
DISPLAYLEVEL(3, "%3u:%3u bytes at pos %8u, savings %7u bytes |",
u, length, pos, dictList[u].savings);
ZDICT_printHex((const char*)samplesBuffer+pos, printedLength);
@@ -1078,17 +1080,17 @@ size_t ZDICT_trainFromBuffer_legacy(void* dictBuffer, size_t dictBufferCapacity,
size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
{
ZDICT_cover_params_t params;
ZDICT_fastCover_params_t params;
DEBUGLOG(3, "ZDICT_trainFromBuffer");
memset(&params, 0, sizeof(params));
params.d = 8;
params.steps = 4;
/* Default to level 6 since no compression level information is available */
params.zParams.compressionLevel = 6;
params.zParams.compressionLevel = 3;
#if defined(DEBUGLEVEL) && (DEBUGLEVEL>=1)
params.zParams.notificationLevel = DEBUGLEVEL;
#endif
return ZDICT_optimizeTrainFromBuffer_cover(dictBuffer, dictBufferCapacity,
return ZDICT_optimizeTrainFromBuffer_fastCover(dictBuffer, dictBufferCapacity,
samplesBuffer, samplesSizes, nbSamples,
&params);
}
lib/dictBuilder/zdict.h
+60 -6
View File
@@ -39,7 +39,8 @@ extern "C" {
/*! ZDICT_trainFromBuffer():
* Train a dictionary from an array of samples.
* Redirect towards ZDICT_optimizeTrainFromBuffer_cover() single-threaded, with d=8 and steps=4.
* Redirect towards ZDICT_optimizeTrainFromBuffer_fastCover() single-threaded, with d=8, steps=4,
* f=20, and accel=1.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* The resulting dictionary will be saved into `dictBuffer`.
@@ -52,7 +53,8 @@ extern "C" {
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
const void* samplesBuffer,
const size_t* samplesSizes, unsigned nbSamples);
/*====== Helper functions ======*/
@@ -84,12 +86,22 @@ typedef struct {
typedef struct {
unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
unsigned steps; /* Number of steps : Only used for optimization : 0 means default (32) : Higher means more parameters checked */
unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
double splitPoint; /* Percentage of samples used for training: the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (1.0), 1.0 when all samples are used for both training and testing */
double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (1.0), 1.0 when all samples are used for both training and testing */
ZDICT_params_t zParams;
} ZDICT_cover_params_t;
typedef struct {
unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
unsigned f; /* log of size of frequency array : constraint: 0 < f <= 31 : 1 means default(20)*/
unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (0.75), 1.0 when all samples are used for both training and testing */
unsigned accel; /* Acceleration level: constraint: 0 < accel <= 10, higher means faster and less accurate, 0 means default(1) */
ZDICT_params_t zParams;
} ZDICT_fastCover_params_t;
/*! ZDICT_trainFromBuffer_cover():
* Train a dictionary from an array of samples using the COVER algorithm.
@@ -116,9 +128,9 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
* dictionary constructed with those parameters is stored in `dictBuffer`.
*
* All of the parameters d, k, steps are optional.
* If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8, 10, 12, 14, 16}.
* If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
* if steps is zero it defaults to its default value.
* If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [16, 2048].
* If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
*
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
@@ -130,6 +142,48 @@ ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
ZDICT_cover_params_t* parameters);
/*! ZDICT_trainFromBuffer_fastCover():
* Train a dictionary from an array of samples using a modified version of COVER algorithm.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* d and k are required.
* All other parameters are optional, will use default values if not provided
* The resulting dictionary will be saved into `dictBuffer`.
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
* Note: ZDICT_trainFromBuffer_fastCover() requires about 1 bytes of memory for each input byte and additionally another 6 * 2^f bytes of memory .
* Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
* It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
* In general, it's recommended to provide a few thousands samples, though this can vary a lot.
* It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer_fastCover(void *dictBuffer,
size_t dictBufferCapacity, const void *samplesBuffer,
const size_t *samplesSizes, unsigned nbSamples,
ZDICT_fastCover_params_t parameters);
/*! ZDICT_optimizeTrainFromBuffer_fastCover():
* The same requirements as above hold for all the parameters except `parameters`.
* This function tries many parameter combinations (specifically, k and d combinations)
* and picks the best parameters. `*parameters` is filled with the best parameters found,
* dictionary constructed with those parameters is stored in `dictBuffer`.
* All of the parameters d, k, steps, f, and accel are optional.
* If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
* if steps is zero it defaults to its default value.
* If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
* If f is zero, default value of 20 is used.
* If accel is zero, default value of 1 is used.
*
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
* On success `*parameters` contains the parameters selected.
* Note: ZDICT_optimizeTrainFromBuffer_fastCover() requires about 1 byte of memory for each input byte and additionally another 6 * 2^f bytes of memory for each thread.
*/
ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_fastCover(void* dictBuffer,
size_t dictBufferCapacity, const void* samplesBuffer,
const size_t* samplesSizes, unsigned nbSamples,
ZDICT_fastCover_params_t* parameters);
/*! ZDICT_finalizeDictionary():
* Given a custom content as a basis for dictionary, and a set of samples,
* finalize dictionary by adding headers and statistics.
lib/legacy/zstd_v04.c
+1
View File
@@ -1093,6 +1093,7 @@ static size_t FSE_buildDTable(FSE_DTable* dt, const short* normalizedCounter, un
if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
/* Init, lay down lowprob symbols */
memset(tableDecode, 0, sizeof(FSE_DECODE_TYPE) * (maxSymbolValue+1) ); /* useless init, but keep static analyzer happy, and we don't need to performance optimize legacy decoders */
DTableH.tableLog = (U16)tableLog;
for (s=0; s<=maxSymbolValue; s++)
{
lib/legacy/zstd_v05.c
+2
View File
@@ -1224,6 +1224,7 @@ size_t FSEv05_buildDTable(FSEv05_DTable* dt, const short* normalizedCounter, uns
if (tableLog > FSEv05_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
/* Init, lay down lowprob symbols */
memset(tableDecode, 0, sizeof(FSEv05_FUNCTION_TYPE) * (maxSymbolValue+1) ); /* useless init, but keep static analyzer happy, and we don't need to performance optimize legacy decoders */
DTableH.tableLog = (U16)tableLog;
for (s=0; s<=maxSymbolValue; s++) {
if (normalizedCounter[s]==-1) {
@@ -2845,6 +2846,7 @@ size_t ZSTDv05_getcBlockSize(const void* src, size_t srcSize, blockProperties_t*
static size_t ZSTDv05_copyRawBlock(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
{
if (dst==NULL) return ERROR(dstSize_tooSmall);
if (srcSize > maxDstSize) return ERROR(dstSize_tooSmall);
memcpy(dst, src, srcSize);
return srcSize;
lib/legacy/zstd_v06.c
+2 -1
View File
@@ -3041,6 +3041,7 @@ size_t ZSTDv06_getcBlockSize(const void* src, size_t srcSize, blockProperties_t*
static size_t ZSTDv06_copyRawBlock(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
{
if (dst==NULL) return ERROR(dstSize_tooSmall);
if (srcSize > dstCapacity) return ERROR(dstSize_tooSmall);
memcpy(dst, src, srcSize);
return srcSize;
@@ -4006,7 +4007,7 @@ size_t ZBUFFv06_decompressContinue(ZBUFFv06_DCtx* zbd,
if (ZSTDv06_isError(hSize)) return hSize;
if (toLoad > (size_t)(iend-ip)) { /* not enough input to load full header */
memcpy(zbd->headerBuffer + zbd->lhSize, ip, iend-ip);
zbd->lhSize += iend-ip; ip = iend; notDone = 0;
zbd->lhSize += iend-ip;
*dstCapacityPtr = 0;
return (hSize - zbd->lhSize) + ZSTDv06_blockHeaderSize; /* remaining header bytes + next block header */
}
lib/legacy/zstd_v07.c
+11 -9
View File
@@ -3150,10 +3150,10 @@ size_t ZSTDv07_getFrameParams(ZSTDv07_frameParams* fparamsPtr, const void* src,
const BYTE* ip = (const BYTE*)src;
if (srcSize < ZSTDv07_frameHeaderSize_min) return ZSTDv07_frameHeaderSize_min;
memset(fparamsPtr, 0, sizeof(*fparamsPtr));
if (MEM_readLE32(src) != ZSTDv07_MAGICNUMBER) {
if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTDv07_MAGIC_SKIPPABLE_START) {
if (srcSize < ZSTDv07_skippableHeaderSize) return ZSTDv07_skippableHeaderSize; /* magic number + skippable frame length */
memset(fparamsPtr, 0, sizeof(*fparamsPtr));
fparamsPtr->frameContentSize = MEM_readLE32((const char *)src + 4);
fparamsPtr->windowSize = 0; /* windowSize==0 means a frame is skippable */
return 0;
@@ -3175,11 +3175,13 @@ size_t ZSTDv07_getFrameParams(ZSTDv07_frameParams* fparamsPtr, const void* src,
U32 windowSize = 0;
U32 dictID = 0;
U64 frameContentSize = 0;
if ((fhdByte & 0x08) != 0) return ERROR(frameParameter_unsupported); /* reserved bits, which must be zero */
if ((fhdByte & 0x08) != 0) /* reserved bits, which must be zero */
return ERROR(frameParameter_unsupported);
if (!directMode) {
BYTE const wlByte = ip[pos++];
U32 const windowLog = (wlByte >> 3) + ZSTDv07_WINDOWLOG_ABSOLUTEMIN;
if (windowLog > ZSTDv07_WINDOWLOG_MAX) return ERROR(frameParameter_unsupported);
if (windowLog > ZSTDv07_WINDOWLOG_MAX)
return ERROR(frameParameter_unsupported);
windowSize = (1U << windowLog);
windowSize += (windowSize >> 3) * (wlByte&7);
}
@@ -3201,7 +3203,8 @@ size_t ZSTDv07_getFrameParams(ZSTDv07_frameParams* fparamsPtr, const void* src,
case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
}
if (!windowSize) windowSize = (U32)frameContentSize;
if (windowSize > windowSizeMax) return ERROR(frameParameter_unsupported);
if (windowSize > windowSizeMax)
return ERROR(frameParameter_unsupported);
fparamsPtr->frameContentSize = frameContentSize;
fparamsPtr->windowSize = windowSize;
fparamsPtr->dictID = dictID;
@@ -3220,11 +3223,10 @@ size_t ZSTDv07_getFrameParams(ZSTDv07_frameParams* fparamsPtr, const void* src,
- frame header not completely provided (`srcSize` too small) */
unsigned long long ZSTDv07_getDecompressedSize(const void* src, size_t srcSize)
{
{ ZSTDv07_frameParams fparams;
size_t const frResult = ZSTDv07_getFrameParams(&fparams, src, srcSize);
if (frResult!=0) return 0;
return fparams.frameContentSize;
}
ZSTDv07_frameParams fparams;
size_t const frResult = ZSTDv07_getFrameParams(&fparams, src, srcSize);
if (frResult!=0) return 0;
return fparams.frameContentSize;
}
lib/zstd.h
+41 -19
View File
@@ -35,26 +35,38 @@ extern "C" {
#endif
/*******************************************************************************************************
/*******************************************************************************
Introduction
zstd, short for Zstandard, is a fast lossless compression algorithm,
targeting real-time compression scenarios at zlib-level and better compression ratios.
The zstd compression library provides in-memory compression and decompression functions.
The library supports compression levels from 1 up to ZSTD_maxCLevel() which is currently 22.
Levels >= 20, labeled `--ultra`, should be used with caution, as they require more memory.
zstd, short for Zstandard, is a fast lossless compression algorithm, targeting
real-time compression scenarios at zlib-level and better compression ratios.
The zstd compression library provides in-memory compression and decompression
functions.
The library supports regular compression levels from 1 up to ZSTD_maxCLevel(),
which is currently 22. Levels >= 20, labeled `--ultra`, should be used with
caution, as they require more memory. The library also offers negative
compression levels, which extend the range of speed vs. ratio preferences.
The lower the level, the faster the speed (at the cost of compression).
Compression can be done in:
- a single step (described as Simple API)
- a single step, reusing a context (described as Explicit context)
- unbounded multiple steps (described as Streaming compression)
The compression ratio achievable on small data can be highly improved using a dictionary in:
- a single step (described as Simple dictionary API)
- a single step, reusing a dictionary (described as Bulk-processing dictionary API)
Advanced experimental functions can be accessed using #define ZSTD_STATIC_LINKING_ONLY before including zstd.h.
Advanced experimental APIs shall never be used with a dynamic library.
They are not "stable", their definition may change in the future. Only static linking is allowed.
*********************************************************************************************************/
The compression ratio achievable on small data can be highly improved using
a dictionary. Dictionary compression can be performed in:
- a single step (described as Simple dictionary API)
- a single step, reusing a dictionary (described as Bulk-processing
dictionary API)
Advanced experimental functions can be accessed using
`#define ZSTD_STATIC_LINKING_ONLY` before including zstd.h.
Advanced experimental APIs should never be used with a dynamically-linked
library. They are not "stable"; their definitions or signatures may change in
the future. Only static linking is allowed.
*******************************************************************************/
/*------ Version ------*/
#define ZSTD_VERSION_MAJOR 1
@@ -211,7 +223,8 @@ typedef struct ZSTD_CDict_s ZSTD_CDict;
* When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
* ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
* ZSTD_CDict can be created once and shared by multiple threads concurrently, since its usage is read-only.
* `dictBuffer` can be released after ZSTD_CDict creation, since its content is copied within CDict */
* `dictBuffer` can be released after ZSTD_CDict creation, since its content is copied within CDict
* Note : A ZSTD_CDict can be created with an empty dictionary, but it is inefficient for small data. */
ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize,
int compressionLevel);
@@ -223,7 +236,9 @@ ZSTDLIB_API size_t ZSTD_freeCDict(ZSTD_CDict* CDict);
* Compression using a digested Dictionary.
* Faster startup than ZSTD_compress_usingDict(), recommended when same dictionary is used multiple times.
* Note that compression level is decided during dictionary creation.
* Frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no) */
* Frame parameters are hardcoded (dictID=yes, contentSize=yes, checksum=no)
* Note : ZSTD_compress_usingCDict() can be used with a ZSTD_CDict created from an empty dictionary.
* But it is inefficient for small data, and it is recommended to use ZSTD_compressCCtx(). */
ZSTDLIB_API size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
@@ -1161,16 +1176,21 @@ ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
/*! ZSTD_CCtx_refPrefix() :
* Reference a prefix (single-usage dictionary) for next compression job.
* Decompression need same prefix to properly regenerate data.
* Prefix is **only used once**. Tables are discarded at end of compression job (ZSTD_e_end).
* Decompression will need same prefix to properly regenerate data.
* Compressing with a prefix is similar in outcome as performing a diff and compressing it,
* but performs much faster, especially during decompression (compression speed is tunable with compression level).
* Note that prefix is **only used once**. Tables are discarded at end of compression job (ZSTD_e_end).
* @result : 0, or an error code (which can be tested with ZSTD_isError()).
* Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
* Note 1 : Prefix buffer is referenced. It **must** outlive compression job.
* Its contain must remain unmodified up to end of compression (ZSTD_e_end).
* Note 2 : Referencing a prefix involves building tables, which are dependent on compression parameters.
* Note 2 : If the intention is to diff some large src data blob with some prior version of itself,
* ensure that the window size is large enough to contain the entire source.
* See ZSTD_p_windowLog.
* Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
* It's a CPU consuming operation, with non-negligible impact on latency.
* If there is a need to use same prefix multiple times, consider loadDictionary instead.
* Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
* Note 4 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
* Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode. */
ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
const void* prefix, size_t prefixSize);
@@ -1353,6 +1373,8 @@ ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
/*! ZSTD_DCtx_refPrefix() :
* Reference a prefix (single-usage dictionary) for next compression job.
* This is the reverse operation of ZSTD_CCtx_refPrefix(),
* and must use the same prefix as the one used during compression.
* Prefix is **only used once**. Reference is discarded at end of frame.
* End of frame is reached when ZSTD_DCtx_decompress_generic() returns 0.
* @result : 0, or an error code (which can be tested with ZSTD_isError()).