c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-22 02:27:36 +03:00
Code Issues Packages Projects Releases Wiki Activity

Explode zstd 1.4.5 into separate upstream files

Browse Source 
It's much nicer to import from upstream this way, and compiles
faster too.

Everything in lib/ is unmodified 1.4.5.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rich Ercolani <rincebrain@gmail.com>
Closes #12978
This commit is contained in:
Rich Ercolani
2022-03-01 08:32:52 -05:00
committed by Brian Behlendorf
parent 669683c4cb
commit 234e9605c1
58 changed files with 23917 additions and 27895 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zstd/lib/compress/fse_compress.c
+698
View File
@@ -0,0 +1,698 @@
/* ******************************************************************
* FSE : Finite State Entropy encoder
* Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
*
* You can contact the author at :
* - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
* - Public forum : https://groups.google.com/forum/#!forum/lz4c
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
****************************************************************** */
/* **************************************************************
* Includes
****************************************************************/
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* memcpy, memset */
#include "../common/compiler.h"
#include "../common/mem.h" /* U32, U16, etc. */
#include "../common/debug.h" /* assert, DEBUGLOG */
#include "hist.h" /* HIST_count_wksp */
#include "../common/bitstream.h"
#define FSE_STATIC_LINKING_ONLY
#include "../common/fse.h"
#include "../common/error_private.h"
/* **************************************************************
* Error Management
****************************************************************/
#define FSE_isError ERR_isError
/* **************************************************************
* Templates
****************************************************************/
/*
designed to be included
for type-specific functions (template emulation in C)
Objective is to write these functions only once, for improved maintenance
*/
/* safety checks */
#ifndef FSE_FUNCTION_EXTENSION
# error "FSE_FUNCTION_EXTENSION must be defined"
#endif
#ifndef FSE_FUNCTION_TYPE
# error "FSE_FUNCTION_TYPE must be defined"
#endif
/* Function names */
#define FSE_CAT(X,Y) X##Y
#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
/* Function templates */
/* FSE_buildCTable_wksp() :
* Same as FSE_buildCTable(), but using an externally allocated scratch buffer (`workSpace`).
* 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)
{
U32 const tableSize = 1 << tableLog;
U32 const tableMask = tableSize - 1;
void* const ptr = ct;
U16* const tableU16 = ( (U16*) ptr) + 2;
void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableLog ? tableSize>>1 : 1) ;
FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
U32 const step = FSE_TABLESTEP(tableSize);
U32 cumul[FSE_MAX_SYMBOL_VALUE+2];
FSE_FUNCTION_TYPE* const tableSymbol = (FSE_FUNCTION_TYPE*)workSpace;
U32 highThreshold = tableSize-1;
/* CTable header */
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 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 */
#ifdef __clang_analyzer__
memset(tableSymbol, 0, sizeof(*tableSymbol) * tableSize); /* useless initialization, just to keep scan-build happy */
#endif
/* symbol start positions */
{ U32 u;
cumul[0] = 0;
for (u=1; u <= maxSymbolValue+1; u++) {
if (normalizedCounter[u-1]==-1) { /* Low proba symbol */
cumul[u] = cumul[u-1] + 1;
tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(u-1);
} else {
cumul[u] = cumul[u-1] + normalizedCounter[u-1];
} }
cumul[maxSymbolValue+1] = tableSize+1;
}
/* Spread symbols */
{ U32 position = 0;
U32 symbol;
for (symbol=0; symbol<=maxSymbolValue; symbol++) {
int nbOccurrences;
int const freq = normalizedCounter[symbol];
for (nbOccurrences=0; nbOccurrences<freq; nbOccurrences++) {
tableSymbol[position] = (FSE_FUNCTION_TYPE)symbol;
position = (position + step) & tableMask;
while (position > highThreshold)
position = (position + step) & tableMask; /* Low proba area */
} }
assert(position==0); /* Must have initialized all positions */
}
/* Build table */
{ U32 u; for (u=0; u<tableSize; u++) {
FSE_FUNCTION_TYPE s = tableSymbol[u]; /* note : static analyzer may not understand tableSymbol is properly initialized */
tableU16[cumul[s]++] = (U16) (tableSize+u); /* TableU16 : sorted by symbol order; gives next state value */
} }
/* Build Symbol Transformation Table */
{ unsigned total = 0;
unsigned s;
for (s=0; s<=maxSymbolValue; s++) {
switch (normalizedCounter[s])
{
case 0:
/* filling nonetheless, for compatibility with FSE_getMaxNbBits() */
symbolTT[s].deltaNbBits = ((tableLog+1) << 16) - (1<<tableLog);
break;
case -1:
case 1:
symbolTT[s].deltaNbBits = (tableLog << 16) - (1<<tableLog);
symbolTT[s].deltaFindState = total - 1;
total ++;
break;
default :
{
U32 const maxBitsOut = tableLog - BIT_highbit32 (normalizedCounter[s]-1);
U32 const minStatePlus = normalizedCounter[s] << maxBitsOut;
symbolTT[s].deltaNbBits = (maxBitsOut << 16) - minStatePlus;
symbolTT[s].deltaFindState = total - normalizedCounter[s];
total += normalizedCounter[s];
} } } }
#if 0 /* debug : symbol costs */
DEBUGLOG(5, "\n --- table statistics : ");
{ U32 symbol;
for (symbol=0; symbol<=maxSymbolValue; symbol++) {
DEBUGLOG(5, "%3u: w=%3i, maxBits=%u, fracBits=%.2f",
symbol, normalizedCounter[symbol],
FSE_getMaxNbBits(symbolTT, symbol),
(double)FSE_bitCost(symbolTT, tableLog, symbol, 8) / 256);
}
}
#endif
return 0;
}
size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
{
FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE]; /* memset() is not necessary, even if static analyzer complain about it */
return FSE_buildCTable_wksp(ct, normalizedCounter, maxSymbolValue, tableLog, tableSymbol, sizeof(tableSymbol));
}
#ifndef FSE_COMMONDEFS_ONLY
/*-**************************************************************
* FSE NCount encoding
****************************************************************/
size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog)
{
size_t const maxHeaderSize = (((maxSymbolValue+1) * tableLog) >> 3) + 3;
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)
{
BYTE* const ostart = (BYTE*) header;
BYTE* out = ostart;
BYTE* const oend = ostart + headerBufferSize;
int nbBits;
const int tableSize = 1 << tableLog;
int remaining;
int threshold;
U32 bitStream = 0;
int bitCount = 0;
unsigned symbol = 0;
unsigned const alphabetSize = maxSymbolValue + 1;
int previousIs0 = 0;
/* Table Size */
bitStream += (tableLog-FSE_MIN_TABLELOG) << bitCount;
bitCount += 4;
/* Init */
remaining = tableSize+1; /* +1 for extra accuracy */
threshold = tableSize;
nbBits = tableLog+1;
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 */
out[0] = (BYTE) bitStream;
out[1] = (BYTE)(bitStream>>8);
out+=2;
bitStream>>=16;
}
while (symbol >= start+3) {
start+=3;
bitStream += 3 << bitCount;
bitCount += 2;
}
bitStream += (symbol-start) << bitCount;
bitCount += 2;
if (bitCount>16) {
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[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[ */
bitStream += count << bitCount;
bitCount += nbBits;
bitCount -= (count<max);
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 */
out[0] = (BYTE)bitStream;
out[1] = (BYTE)(bitStream>>8);
out += 2;
bitStream >>= 16;
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 */
out[0] = (BYTE)bitStream;
out[1] = (BYTE)(bitStream>>8);
out+= (bitCount+7) /8;
return (out-ostart);
}
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 */
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 /* write in buffer is safe */);
}
/*-**************************************************************
* FSE Compression Code
****************************************************************/
FSE_CTable* FSE_createCTable (unsigned maxSymbolValue, unsigned tableLog)
{
size_t size;
if (tableLog > FSE_TABLELOG_ABSOLUTE_MAX) tableLog = FSE_TABLELOG_ABSOLUTE_MAX;
size = FSE_CTABLE_SIZE_U32 (tableLog, maxSymbolValue) * sizeof(U32);
return (FSE_CTable*)malloc(size);
}
void FSE_freeCTable (FSE_CTable* ct) { free(ct); }
/* provides the minimum logSize to safely represent a distribution */
static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
{
U32 minBitsSrc = BIT_highbit32((U32)(srcSize)) + 1;
U32 minBitsSymbols = BIT_highbit32(maxSymbolValue) + 2;
U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
assert(srcSize > 1); /* Not supported, RLE should be used instead */
return minBits;
}
unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue, unsigned minus)
{
U32 maxBitsSrc = BIT_highbit32((U32)(srcSize - 1)) - minus;
U32 tableLog = maxTableLog;
U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
assert(srcSize > 1); /* Not supported, RLE should be used instead */
if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
if (maxBitsSrc < tableLog) tableLog = maxBitsSrc; /* Accuracy can be reduced */
if (minBits > tableLog) tableLog = minBits; /* Need a minimum to safely represent all symbol values */
if (tableLog < FSE_MIN_TABLELOG) tableLog = FSE_MIN_TABLELOG;
if (tableLog > FSE_MAX_TABLELOG) tableLog = FSE_MAX_TABLELOG;
return tableLog;
}
unsigned FSE_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
{
return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 2);
}
/* Secondary normalization method.
To be used when primary method fails. */
static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count, size_t total, U32 maxSymbolValue)
{
short const NOT_YET_ASSIGNED = -2;
U32 s;
U32 distributed = 0;
U32 ToDistribute;
/* Init */
U32 const lowThreshold = (U32)(total >> tableLog);
U32 lowOne = (U32)((total * 3) >> (tableLog + 1));
for (s=0; s<=maxSymbolValue; s++) {
if (count[s] == 0) {
norm[s]=0;
continue;
}
if (count[s] <= lowThreshold) {
norm[s] = -1;
distributed++;
total -= count[s];
continue;
}
if (count[s] <= lowOne) {
norm[s] = 1;
distributed++;
total -= count[s];
continue;
}
norm[s]=NOT_YET_ASSIGNED;
}
ToDistribute = (1 << tableLog) - distributed;
if (ToDistribute == 0)
return 0;
if ((total / ToDistribute) > lowOne) {
/* risk of rounding to zero */
lowOne = (U32)((total * 3) / (ToDistribute * 2));
for (s=0; s<=maxSymbolValue; s++) {
if ((norm[s] == NOT_YET_ASSIGNED) && (count[s] <= lowOne)) {
norm[s] = 1;
distributed++;
total -= count[s];
continue;
} }
ToDistribute = (1 << tableLog) - distributed;
}
if (distributed == maxSymbolValue+1) {
/* all values are pretty poor;
probably incompressible data (should have already been detected);
find max, then give all remaining points to max */
U32 maxV = 0, maxC = 0;
for (s=0; s<=maxSymbolValue; s++)
if (count[s] > maxC) { maxV=s; maxC=count[s]; }
norm[maxV] += (short)ToDistribute;
return 0;
}
if (total == 0) {
/* all of the symbols were low enough for the lowOne or lowThreshold */
for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
if (norm[s] > 0) { ToDistribute--; norm[s]++; }
return 0;
}
{ U64 const vStepLog = 62 - tableLog;
U64 const mid = (1ULL << (vStepLog-1)) - 1;
U64 const rStep = ((((U64)1<<vStepLog) * ToDistribute) + mid) / total; /* scale on remaining */
U64 tmpTotal = mid;
for (s=0; s<=maxSymbolValue; s++) {
if (norm[s]==NOT_YET_ASSIGNED) {
U64 const end = tmpTotal + (count[s] * rStep);
U32 const sStart = (U32)(tmpTotal >> vStepLog);
U32 const sEnd = (U32)(end >> vStepLog);
U32 const weight = sEnd - sStart;
if (weight < 1)
return ERROR(GENERIC);
norm[s] = (short)weight;
tmpTotal = end;
} } }
return 0;
}
size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
const unsigned* count, size_t total,
unsigned maxSymbolValue)
{
/* Sanity checks */
if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
if (tableLog < FSE_MIN_TABLELOG) return ERROR(GENERIC); /* Unsupported size */
if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge); /* Unsupported size */
if (tableLog < FSE_minTableLog(total, maxSymbolValue)) return ERROR(GENERIC); /* Too small tableLog, compression potentially impossible */
{ static U32 const rtbTable[] = { 0, 473195, 504333, 520860, 550000, 700000, 750000, 830000 };
U64 const scale = 62 - tableLog;
U64 const step = ((U64)1<<62) / total; /* <== here, one division ! */
U64 const vStep = 1ULL<<(scale-20);
int stillToDistribute = 1<<tableLog;
unsigned s;
unsigned largest=0;
short largestP=0;
U32 lowThreshold = (U32)(total >> tableLog);
for (s=0; s<=maxSymbolValue; s++) {
if (count[s] == total) return 0; /* rle special case */
if (count[s] == 0) { normalizedCounter[s]=0; continue; }
if (count[s] <= lowThreshold) {
normalizedCounter[s] = -1;
stillToDistribute--;
} else {
short proba = (short)((count[s]*step) >> scale);
if (proba<8) {
U64 restToBeat = vStep * rtbTable[proba];
proba += (count[s]*step) - ((U64)proba<<scale) > restToBeat;
}
if (proba > largestP) { largestP=proba; largest=s; }
normalizedCounter[s] = proba;
stillToDistribute -= proba;
} }
if (-stillToDistribute >= (normalizedCounter[largest] >> 1)) {
/* corner case, need another normalization method */
size_t const errorCode = FSE_normalizeM2(normalizedCounter, tableLog, count, total, maxSymbolValue);
if (FSE_isError(errorCode)) return errorCode;
}
else normalizedCounter[largest] += (short)stillToDistribute;
}
#if 0
{ /* Print Table (debug) */
U32 s;
U32 nTotal = 0;
for (s=0; s<=maxSymbolValue; s++)
RAWLOG(2, "%3i: %4i \n", s, normalizedCounter[s]);
for (s=0; s<=maxSymbolValue; s++)
nTotal += abs(normalizedCounter[s]);
if (nTotal != (1U<<tableLog))
RAWLOG(2, "Warning !!! Total == %u != %u !!!", nTotal, 1U<<tableLog);
getchar();
}
#endif
return tableLog;
}
/* fake FSE_CTable, for raw (uncompressed) input */
size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits)
{
const unsigned tableSize = 1 << nbBits;
const unsigned tableMask = tableSize - 1;
const unsigned maxSymbolValue = tableMask;
void* const ptr = ct;
U16* const tableU16 = ( (U16*) ptr) + 2;
void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableSize>>1); /* assumption : tableLog >= 1 */
FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
unsigned s;
/* Sanity checks */
if (nbBits < 1) return ERROR(GENERIC); /* min size */
/* header */
tableU16[-2] = (U16) nbBits;
tableU16[-1] = (U16) maxSymbolValue;
/* Build table */
for (s=0; s<tableSize; s++)
tableU16[s] = (U16)(tableSize + s);
/* Build Symbol Transformation Table */
{ const U32 deltaNbBits = (nbBits << 16) - (1 << nbBits);
for (s=0; s<=maxSymbolValue; s++) {
symbolTT[s].deltaNbBits = deltaNbBits;
symbolTT[s].deltaFindState = s-1;
} }
return 0;
}
/* fake FSE_CTable, for rle input (always same symbol) */
size_t FSE_buildCTable_rle (FSE_CTable* ct, BYTE symbolValue)
{
void* ptr = ct;
U16* tableU16 = ( (U16*) ptr) + 2;
void* FSCTptr = (U32*)ptr + 2;
FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) FSCTptr;
/* header */
tableU16[-2] = (U16) 0;
tableU16[-1] = (U16) symbolValue;
/* Build table */
tableU16[0] = 0;
tableU16[1] = 0; /* just in case */
/* Build Symbol Transformation Table */
symbolTT[symbolValue].deltaNbBits = 0;
symbolTT[symbolValue].deltaFindState = 0;
return 0;
}
static size_t FSE_compress_usingCTable_generic (void* dst, size_t dstSize,
const void* src, size_t srcSize,
const FSE_CTable* ct, const unsigned fast)
{
const BYTE* const istart = (const BYTE*) src;
const BYTE* const iend = istart + srcSize;
const BYTE* ip=iend;
BIT_CStream_t bitC;
FSE_CState_t CState1, CState2;
/* init */
if (srcSize <= 2) return 0;
{ size_t const initError = BIT_initCStream(&bitC, dst, dstSize);
if (FSE_isError(initError)) return 0; /* not enough space available to write a bitstream */ }
#define FSE_FLUSHBITS(s) (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
if (srcSize & 1) {
FSE_initCState2(&CState1, ct, *--ip);
FSE_initCState2(&CState2, ct, *--ip);
FSE_encodeSymbol(&bitC, &CState1, *--ip);
FSE_FLUSHBITS(&bitC);
} else {
FSE_initCState2(&CState2, ct, *--ip);
FSE_initCState2(&CState1, ct, *--ip);
}
/* join to mod 4 */
srcSize -= 2;
if ((sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) && (srcSize & 2)) { /* test bit 2 */
FSE_encodeSymbol(&bitC, &CState2, *--ip);
FSE_encodeSymbol(&bitC, &CState1, *--ip);
FSE_FLUSHBITS(&bitC);
}
/* 2 or 4 encoding per loop */
while ( ip>istart ) {
FSE_encodeSymbol(&bitC, &CState2, *--ip);
if (sizeof(bitC.bitContainer)*8 < FSE_MAX_TABLELOG*2+7 ) /* this test must be static */
FSE_FLUSHBITS(&bitC);
FSE_encodeSymbol(&bitC, &CState1, *--ip);
if (sizeof(bitC.bitContainer)*8 > FSE_MAX_TABLELOG*4+7 ) { /* this test must be static */
FSE_encodeSymbol(&bitC, &CState2, *--ip);
FSE_encodeSymbol(&bitC, &CState1, *--ip);
}
FSE_FLUSHBITS(&bitC);
}
FSE_flushCState(&bitC, &CState2);
FSE_flushCState(&bitC, &CState1);
return BIT_closeCStream(&bitC);
}
size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
const void* src, size_t srcSize,
const FSE_CTable* ct)
{
unsigned const fast = (dstSize >= FSE_BLOCKBOUND(srcSize));
if (fast)
return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 1);
else
return FSE_compress_usingCTable_generic(dst, dstSize, src, srcSize, ct, 0);
}
size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
/* FSE_compress_wksp() :
* Same as FSE_compress2(), but using an externally allocated scratch buffer (`workSpace`).
* `wkspSize` size must be `(1<<tableLog)`.
*/
size_t FSE_compress_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize)
{
BYTE* const ostart = (BYTE*) dst;
BYTE* op = ostart;
BYTE* const oend = ostart + dstSize;
unsigned count[FSE_MAX_SYMBOL_VALUE+1];
S16 norm[FSE_MAX_SYMBOL_VALUE+1];
FSE_CTable* CTable = (FSE_CTable*)workSpace;
size_t const CTableSize = FSE_CTABLE_SIZE_U32(tableLog, maxSymbolValue);
void* scratchBuffer = (void*)(CTable + CTableSize);
size_t const scratchBufferSize = wkspSize - (CTableSize * sizeof(FSE_CTable));
/* init conditions */
if (wkspSize < FSE_WKSP_SIZE_U32(tableLog, maxSymbolValue)) return ERROR(tableLog_tooLarge);
if (srcSize <= 1) return 0; /* Not compressible */
if (!maxSymbolValue) maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
if (!tableLog) tableLog = FSE_DEFAULT_TABLELOG;
/* Scan input and build symbol stats */
{ CHECK_V_F(maxCount, HIST_count_wksp(count, &maxSymbolValue, src, srcSize, scratchBuffer, scratchBufferSize) );
if (maxCount == srcSize) return 1; /* only a single symbol in src : rle */
if (maxCount == 1) return 0; /* each symbol present maximum once => not compressible */
if (maxCount < (srcSize >> 7)) return 0; /* Heuristic : not compressible enough */
}
tableLog = FSE_optimalTableLog(tableLog, srcSize, maxSymbolValue);
CHECK_F( FSE_normalizeCount(norm, tableLog, count, srcSize, maxSymbolValue) );
/* Write table description header */
{ CHECK_V_F(nc_err, FSE_writeNCount(op, oend-op, norm, maxSymbolValue, tableLog) );
op += nc_err;
}
/* Compress */
CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, scratchBufferSize) );
{ CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, src, srcSize, CTable) );
if (cSize == 0) return 0; /* not enough space for compressed data */
op += cSize;
}
/* check compressibility */
if ( (size_t)(op-ostart) >= srcSize-1 ) return 0;
return op-ostart;
}
typedef struct {
FSE_CTable CTable_max[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
BYTE scratchBuffer[1 << FSE_MAX_TABLELOG];
} fseWkspMax_t;
size_t FSE_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog)
{
fseWkspMax_t scratchBuffer;
DEBUG_STATIC_ASSERT(sizeof(scratchBuffer) >= FSE_WKSP_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)); /* compilation failures here means scratchBuffer is not large enough */
if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
return FSE_compress_wksp(dst, dstCapacity, src, srcSize, maxSymbolValue, tableLog, &scratchBuffer, sizeof(scratchBuffer));
}
size_t FSE_compress (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
{
return FSE_compress2(dst, dstCapacity, src, srcSize, FSE_MAX_SYMBOL_VALUE, FSE_DEFAULT_TABLELOG);
}
#endif /* FSE_COMMONDEFS_ONLY */
module/zstd/lib/compress/hist.c
+183
View File
@@ -0,0 +1,183 @@
/* ******************************************************************
* hist : Histogram functions
* part of Finite State Entropy project
* Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
*
* You can contact the author at :
* - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
* - Public forum : https://groups.google.com/forum/#!forum/lz4c
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
****************************************************************** */
/* --- dependencies --- */
#include "../common/mem.h" /* U32, BYTE, etc. */
#include "../common/debug.h" /* assert, DEBUGLOG */
#include "../common/error_private.h" /* ERROR */
#include "hist.h"
/* --- Error management --- */
unsigned HIST_isError(size_t code) { return ERR_isError(code); }
/*-**************************************************************
* Histogram functions
****************************************************************/
unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize)
{
const BYTE* ip = (const BYTE*)src;
const BYTE* const end = ip + srcSize;
unsigned maxSymbolValue = *maxSymbolValuePtr;
unsigned largestCount=0;
memset(count, 0, (maxSymbolValue+1) * sizeof(*count));
if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
while (ip<end) {
assert(*ip <= maxSymbolValue);
count[*ip++]++;
}
while (!count[maxSymbolValue]) maxSymbolValue--;
*maxSymbolValuePtr = maxSymbolValue;
{ U32 s;
for (s=0; s<=maxSymbolValue; s++)
if (count[s] > largestCount) largestCount = count[s];
}
return largestCount;
}
typedef enum { trustInput, checkMaxSymbolValue } HIST_checkInput_e;
/* HIST_count_parallel_wksp() :
* store histogram into 4 intermediate tables, recombined at the end.
* this design makes better use of OoO cpus,
* and is noticeably faster when some values are heavily repeated.
* But it needs some additional workspace for intermediate tables.
* `workSpace` size must be a table of size >= HIST_WKSP_SIZE_U32.
* @return : largest histogram frequency,
* or an error code (notably when histogram would be larger than *maxSymbolValuePtr). */
static size_t HIST_count_parallel_wksp(
unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize,
HIST_checkInput_e check,
U32* const workSpace)
{
const BYTE* ip = (const BYTE*)source;
const BYTE* const iend = ip+sourceSize;
unsigned maxSymbolValue = *maxSymbolValuePtr;
unsigned max=0;
U32* const Counting1 = workSpace;
U32* const Counting2 = Counting1 + 256;
U32* const Counting3 = Counting2 + 256;
U32* const Counting4 = Counting3 + 256;
memset(workSpace, 0, 4*256*sizeof(unsigned));
/* safety checks */
if (!sourceSize) {
memset(count, 0, maxSymbolValue + 1);
*maxSymbolValuePtr = 0;
return 0;
}
if (!maxSymbolValue) maxSymbolValue = 255; /* 0 == default */
/* by stripes of 16 bytes */
{ U32 cached = MEM_read32(ip); ip += 4;
while (ip < iend-15) {
U32 c = cached; cached = MEM_read32(ip); ip += 4;
Counting1[(BYTE) c ]++;
Counting2[(BYTE)(c>>8) ]++;
Counting3[(BYTE)(c>>16)]++;
Counting4[ c>>24 ]++;
c = cached; cached = MEM_read32(ip); ip += 4;
Counting1[(BYTE) c ]++;
Counting2[(BYTE)(c>>8) ]++;
Counting3[(BYTE)(c>>16)]++;
Counting4[ c>>24 ]++;
c = cached; cached = MEM_read32(ip); ip += 4;
Counting1[(BYTE) c ]++;
Counting2[(BYTE)(c>>8) ]++;
Counting3[(BYTE)(c>>16)]++;
Counting4[ c>>24 ]++;
c = cached; cached = MEM_read32(ip); ip += 4;
Counting1[(BYTE) c ]++;
Counting2[(BYTE)(c>>8) ]++;
Counting3[(BYTE)(c>>16)]++;
Counting4[ c>>24 ]++;
}
ip-=4;
}
/* finish last symbols */
while (ip<iend) Counting1[*ip++]++;
if (check) { /* verify stats will fit into destination table */
U32 s; for (s=255; s>maxSymbolValue; s--) {
Counting1[s] += Counting2[s] + Counting3[s] + Counting4[s];
if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
} }
{ U32 s;
if (maxSymbolValue > 255) maxSymbolValue = 255;
for (s=0; s<=maxSymbolValue; s++) {
count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
if (count[s] > max) max = count[s];
} }
while (!count[maxSymbolValue]) maxSymbolValue--;
*maxSymbolValuePtr = maxSymbolValue;
return (size_t)max;
}
/* HIST_countFast_wksp() :
* Same as HIST_countFast(), but using an externally provided scratch buffer.
* `workSpace` is a writable buffer which must be 4-bytes aligned,
* `workSpaceSize` must be >= HIST_WKSP_SIZE
*/
size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize,
void* workSpace, size_t workSpaceSize)
{
if (sourceSize < 1500) /* heuristic threshold */
return HIST_count_simple(count, maxSymbolValuePtr, source, sourceSize);
if ((size_t)workSpace & 3) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, trustInput, (U32*)workSpace);
}
/* fast variant (unsafe : won't check if src contains values beyond count[] limit) */
size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize)
{
unsigned tmpCounters[HIST_WKSP_SIZE_U32];
return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, tmpCounters, sizeof(tmpCounters));
}
/* HIST_count_wksp() :
* Same as HIST_count(), but using an externally provided scratch buffer.
* `workSpace` size must be table of >= HIST_WKSP_SIZE_U32 unsigned */
size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize,
void* workSpace, size_t workSpaceSize)
{
if ((size_t)workSpace & 3) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (workSpaceSize < HIST_WKSP_SIZE) return ERROR(workSpace_tooSmall);
if (*maxSymbolValuePtr < 255)
return HIST_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, checkMaxSymbolValue, (U32*)workSpace);
*maxSymbolValuePtr = 255;
return HIST_countFast_wksp(count, maxSymbolValuePtr, source, sourceSize, workSpace, workSpaceSize);
}
size_t HIST_count(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize)
{
unsigned tmpCounters[HIST_WKSP_SIZE_U32];
return HIST_count_wksp(count, maxSymbolValuePtr, src, srcSize, tmpCounters, sizeof(tmpCounters));
}
module/zstd/lib/compress/hist.h
+75
View File
@@ -0,0 +1,75 @@
/* ******************************************************************
* hist : Histogram functions
* part of Finite State Entropy project
* Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
*
* You can contact the author at :
* - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
* - Public forum : https://groups.google.com/forum/#!forum/lz4c
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
****************************************************************** */
/* --- dependencies --- */
#include <stddef.h> /* size_t */
/* --- simple histogram functions --- */
/*! HIST_count():
* Provides the precise count of each byte within a table 'count'.
* 'count' is a table of unsigned int, of minimum size (*maxSymbolValuePtr+1).
* Updates *maxSymbolValuePtr with actual largest symbol value detected.
* @return : count of the most frequent symbol (which isn't identified).
* or an error code, which can be tested using HIST_isError().
* note : if return == srcSize, there is only one symbol.
*/
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 */
/* --- advanced histogram functions --- */
#define HIST_WKSP_SIZE_U32 1024
#define HIST_WKSP_SIZE (HIST_WKSP_SIZE_U32 * sizeof(unsigned))
/** HIST_count_wksp() :
* Same as HIST_count(), but using an externally provided scratch buffer.
* Benefit is this function will use very little stack space.
* `workSpace` is a writable buffer which must be 4-bytes aligned,
* `workSpaceSize` must be >= HIST_WKSP_SIZE
*/
size_t HIST_count_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize,
void* workSpace, size_t workSpaceSize);
/** HIST_countFast() :
* same as HIST_count(), but blindly trusts that all byte values within src are <= *maxSymbolValuePtr.
* This function is unsafe, and will segfault if any value within `src` is `> *maxSymbolValuePtr`
*/
size_t HIST_countFast(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize);
/** HIST_countFast_wksp() :
* Same as HIST_countFast(), but using an externally provided scratch buffer.
* `workSpace` is a writable buffer which must be 4-bytes aligned,
* `workSpaceSize` must be >= HIST_WKSP_SIZE
*/
size_t HIST_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize,
void* workSpace, size_t workSpaceSize);
/*! HIST_count_simple() :
* Same as HIST_countFast(), this function is unsafe,
* and will segfault if any value within `src` is `> *maxSymbolValuePtr`.
* It is also a bit slower for large inputs.
* However, it does not need any additional memory (not even on stack).
* @return : count of the most frequent symbol.
* Note this function doesn't produce any error (i.e. it must succeed).
*/
unsigned HIST_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize);
module/zstd/lib/compress/huf_compress.c
+798
View File
@@ -0,0 +1,798 @@
/* ******************************************************************
* Huffman encoder, part of New Generation Entropy library
* Copyright (c) 2013-2020, Yann Collet, Facebook, Inc.
*
* You can contact the author at :
* - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
* - Public forum : https://groups.google.com/forum/#!forum/lz4c
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
****************************************************************** */
/* **************************************************************
* Compiler specifics
****************************************************************/
#ifdef _MSC_VER /* Visual Studio */
# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
#endif
/* **************************************************************
* Includes
****************************************************************/
#include <string.h> /* memcpy, memset */
#include <stdio.h> /* printf (debug) */
#include "../common/compiler.h"
#include "../common/bitstream.h"
#include "hist.h"
#define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */
#include "../common/fse.h" /* header compression */
#define HUF_STATIC_LINKING_ONLY
#include "../common/huf.h"
#include "../common/error_private.h"
/* **************************************************************
* Error Management
****************************************************************/
#define HUF_isError ERR_isError
#define HUF_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c) /* use only *after* variable declarations */
/* **************************************************************
* Utils
****************************************************************/
unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
{
return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
}
/* *******************************************************
* HUF : Huffman block compression
*********************************************************/
/* HUF_compressWeights() :
* Same as FSE_compress(), but dedicated to huff0's weights compression.
* The use case needs much less stack memory.
* Note : all elements within weightTable are supposed to be <= HUF_TABLELOG_MAX.
*/
#define MAX_FSE_TABLELOG_FOR_HUFF_HEADER 6
static size_t HUF_compressWeights (void* dst, size_t dstSize, const void* weightTable, size_t wtSize)
{
BYTE* const ostart = (BYTE*) dst;
BYTE* op = ostart;
BYTE* const oend = ostart + dstSize;
unsigned maxSymbolValue = HUF_TABLELOG_MAX;
U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
FSE_CTable CTable[FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX)];
BYTE scratchBuffer[1<<MAX_FSE_TABLELOG_FOR_HUFF_HEADER];
unsigned count[HUF_TABLELOG_MAX+1];
S16 norm[HUF_TABLELOG_MAX+1];
/* init conditions */
if (wtSize <= 1) return 0; /* Not compressible */
/* Scan input and build symbol stats */
{ unsigned const maxCount = HIST_count_simple(count, &maxSymbolValue, weightTable, wtSize); /* never fails */
if (maxCount == wtSize) return 1; /* only a single symbol in src : rle */
if (maxCount == 1) return 0; /* each symbol present maximum once => not compressible */
}
tableLog = FSE_optimalTableLog(tableLog, wtSize, maxSymbolValue);
CHECK_F( FSE_normalizeCount(norm, tableLog, count, wtSize, maxSymbolValue) );
/* Write table description header */
{ CHECK_V_F(hSize, FSE_writeNCount(op, (size_t)(oend-op), norm, maxSymbolValue, tableLog) );
op += hSize;
}
/* Compress */
CHECK_F( FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, scratchBuffer, sizeof(scratchBuffer)) );
{ CHECK_V_F(cSize, FSE_compress_usingCTable(op, (size_t)(oend - op), weightTable, wtSize, CTable) );
if (cSize == 0) return 0; /* not enough space for compressed data */
op += cSize;
}
return (size_t)(op-ostart);
}
struct HUF_CElt_s {
U16 val;
BYTE nbBits;
}; /* typedef'd to HUF_CElt within "huf.h" */
/*! HUF_writeCTable() :
`CTable` : Huffman tree to save, using huf representation.
@return : size of saved CTable */
size_t HUF_writeCTable (void* dst, size_t maxDstSize,
const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog)
{
BYTE bitsToWeight[HUF_TABLELOG_MAX + 1]; /* precomputed conversion table */
BYTE huffWeight[HUF_SYMBOLVALUE_MAX];
BYTE* op = (BYTE*)dst;
U32 n;
/* check conditions */
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
/* convert to weight */
bitsToWeight[0] = 0;
for (n=1; n<huffLog+1; n++)
bitsToWeight[n] = (BYTE)(huffLog + 1 - n);
for (n=0; n<maxSymbolValue; n++)
huffWeight[n] = bitsToWeight[CTable[n].nbBits];
/* attempt weights compression by FSE */
{ CHECK_V_F(hSize, HUF_compressWeights(op+1, maxDstSize-1, huffWeight, maxSymbolValue) );
if ((hSize>1) & (hSize < maxSymbolValue/2)) { /* FSE compressed */
op[0] = (BYTE)hSize;
return hSize+1;
} }
/* write raw values as 4-bits (max : 15) */
if (maxSymbolValue > (256-128)) return ERROR(GENERIC); /* should not happen : likely means source cannot be compressed */
if (((maxSymbolValue+1)/2) + 1 > maxDstSize) return ERROR(dstSize_tooSmall); /* not enough space within dst buffer */
op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue-1));
huffWeight[maxSymbolValue] = 0; /* to be sure it doesn't cause msan issue in final combination */
for (n=0; n<maxSymbolValue; n+=2)
op[(n/2)+1] = (BYTE)((huffWeight[n] << 4) + huffWeight[n+1]);
return ((maxSymbolValue+1)/2) + 1;
}
size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* hasZeroWeights)
{
BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1]; /* init not required, even though some static analyzer may complain */
U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1]; /* large enough for values from 0 to 16 */
U32 tableLog = 0;
U32 nbSymbols = 0;
/* get symbol weights */
CHECK_V_F(readSize, HUF_readStats(huffWeight, HUF_SYMBOLVALUE_MAX+1, rankVal, &nbSymbols, &tableLog, src, srcSize));
/* check result */
if (tableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
if (nbSymbols > *maxSymbolValuePtr+1) return ERROR(maxSymbolValue_tooSmall);
/* Prepare base value per rank */
{ U32 n, nextRankStart = 0;
for (n=1; n<=tableLog; n++) {
U32 current = nextRankStart;
nextRankStart += (rankVal[n] << (n-1));
rankVal[n] = current;
} }
/* fill nbBits */
*hasZeroWeights = 0;
{ U32 n; for (n=0; n<nbSymbols; n++) {
const U32 w = huffWeight[n];
*hasZeroWeights |= (w == 0);
CTable[n].nbBits = (BYTE)(tableLog + 1 - w) & -(w != 0);
} }
/* fill val */
{ U16 nbPerRank[HUF_TABLELOG_MAX+2] = {0}; /* support w=0=>n=tableLog+1 */
U16 valPerRank[HUF_TABLELOG_MAX+2] = {0};
{ U32 n; for (n=0; n<nbSymbols; n++) nbPerRank[CTable[n].nbBits]++; }
/* determine stating value per rank */
valPerRank[tableLog+1] = 0; /* for w==0 */
{ U16 min = 0;
U32 n; for (n=tableLog; n>0; n--) { /* start at n=tablelog <-> w=1 */
valPerRank[n] = min; /* get starting value within each rank */
min += nbPerRank[n];
min >>= 1;
} }
/* assign value within rank, symbol order */
{ U32 n; for (n=0; n<nbSymbols; n++) CTable[n].val = valPerRank[CTable[n].nbBits]++; }
}
*maxSymbolValuePtr = nbSymbols - 1;
return readSize;
}
U32 HUF_getNbBits(const void* symbolTable, U32 symbolValue)
{
const HUF_CElt* table = (const HUF_CElt*)symbolTable;
assert(symbolValue <= HUF_SYMBOLVALUE_MAX);
return table[symbolValue].nbBits;
}
typedef struct nodeElt_s {
U32 count;
U16 parent;
BYTE byte;
BYTE nbBits;
} nodeElt;
static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
{
const U32 largestBits = huffNode[lastNonNull].nbBits;
if (largestBits <= maxNbBits) return largestBits; /* early exit : no elt > maxNbBits */
/* there are several too large elements (at least >= 2) */
{ int totalCost = 0;
const U32 baseCost = 1 << (largestBits - maxNbBits);
int n = (int)lastNonNull;
while (huffNode[n].nbBits > maxNbBits) {
totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
huffNode[n].nbBits = (BYTE)maxNbBits;
n --;
} /* n stops at huffNode[n].nbBits <= maxNbBits */
while (huffNode[n].nbBits == maxNbBits) n--; /* n end at index of smallest symbol using < maxNbBits */
/* renorm totalCost */
totalCost >>= (largestBits - maxNbBits); /* note : totalCost is necessarily a multiple of baseCost */
/* repay normalized cost */
{ U32 const noSymbol = 0xF0F0F0F0;
U32 rankLast[HUF_TABLELOG_MAX+2];
/* Get pos of last (smallest) symbol per rank */
memset(rankLast, 0xF0, sizeof(rankLast));
{ U32 currentNbBits = maxNbBits;
int pos;
for (pos=n ; pos >= 0; pos--) {
if (huffNode[pos].nbBits >= currentNbBits) continue;
currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */
rankLast[maxNbBits-currentNbBits] = (U32)pos;
} }
while (totalCost > 0) {
U32 nBitsToDecrease = BIT_highbit32((U32)totalCost) + 1;
for ( ; nBitsToDecrease > 1; nBitsToDecrease--) {
U32 const highPos = rankLast[nBitsToDecrease];
U32 const lowPos = rankLast[nBitsToDecrease-1];
if (highPos == noSymbol) continue;
if (lowPos == noSymbol) break;
{ U32 const highTotal = huffNode[highPos].count;
U32 const lowTotal = 2 * huffNode[lowPos].count;
if (highTotal <= lowTotal) break;
} }
/* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
/* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
nBitsToDecrease ++;
totalCost -= 1 << (nBitsToDecrease-1);
if (rankLast[nBitsToDecrease-1] == noSymbol)
rankLast[nBitsToDecrease-1] = rankLast[nBitsToDecrease]; /* this rank is no longer empty */
huffNode[rankLast[nBitsToDecrease]].nbBits ++;
if (rankLast[nBitsToDecrease] == 0) /* special case, reached largest symbol */
rankLast[nBitsToDecrease] = noSymbol;
else {
rankLast[nBitsToDecrease]--;
if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease)
rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */
} } /* while (totalCost > 0) */
while (totalCost < 0) { /* Sometimes, cost correction overshoot */
if (rankLast[1] == noSymbol) { /* special case : no rank 1 symbol (using maxNbBits-1); let's create one from largest rank 0 (using maxNbBits) */
while (huffNode[n].nbBits == maxNbBits) n--;
huffNode[n+1].nbBits--;
assert(n >= 0);
rankLast[1] = (U32)(n+1);
totalCost++;
continue;
}
huffNode[ rankLast[1] + 1 ].nbBits--;
rankLast[1]++;
totalCost ++;
} } } /* there are several too large elements (at least >= 2) */
return maxNbBits;
}
typedef struct {
U32 base;
U32 current;
} rankPos;
typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
#define RANK_POSITION_TABLE_SIZE 32
typedef struct {
huffNodeTable huffNodeTbl;
rankPos rankPosition[RANK_POSITION_TABLE_SIZE];
} HUF_buildCTable_wksp_tables;
static void HUF_sort(nodeElt* huffNode, const unsigned* count, U32 maxSymbolValue, rankPos* rankPosition)
{
U32 n;
memset(rankPosition, 0, sizeof(*rankPosition) * RANK_POSITION_TABLE_SIZE);
for (n=0; n<=maxSymbolValue; n++) {
U32 r = BIT_highbit32(count[n] + 1);
rankPosition[r].base ++;
}
for (n=30; n>0; n--) rankPosition[n-1].base += rankPosition[n].base;
for (n=0; n<32; n++) rankPosition[n].current = rankPosition[n].base;
for (n=0; n<=maxSymbolValue; n++) {
U32 const c = count[n];
U32 const r = BIT_highbit32(c+1) + 1;
U32 pos = rankPosition[r].current++;
while ((pos > rankPosition[r].base) && (c > huffNode[pos-1].count)) {
huffNode[pos] = huffNode[pos-1];
pos--;
}
huffNode[pos].count = c;
huffNode[pos].byte = (BYTE)n;
}
}
/** HUF_buildCTable_wksp() :
* Same as HUF_buildCTable(), but using externally allocated scratch buffer.
* `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as sizeof(HUF_buildCTable_wksp_tables).
*/
#define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
size_t HUF_buildCTable_wksp (HUF_CElt* tree, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
{
HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)workSpace;
nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
nodeElt* const huffNode = huffNode0+1;
int nonNullRank;
int lowS, lowN;
int nodeNb = STARTNODE;
int n, nodeRoot;
/* safety checks */
if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (wkspSize < sizeof(HUF_buildCTable_wksp_tables))
return ERROR(workSpace_tooSmall);
if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
return ERROR(maxSymbolValue_tooLarge);
memset(huffNode0, 0, sizeof(huffNodeTable));
/* sort, decreasing order */
HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
/* init for parents */
nonNullRank = (int)maxSymbolValue;
while(huffNode[nonNullRank].count == 0) nonNullRank--;
lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb;
huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS-1].count;
huffNode[lowS].parent = huffNode[lowS-1].parent = (U16)nodeNb;
nodeNb++; lowS-=2;
for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
huffNode0[0].count = (U32)(1U<<31); /* fake entry, strong barrier */
/* create parents */
while (nodeNb <= nodeRoot) {
int const n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
int const n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count;
huffNode[n1].parent = huffNode[n2].parent = (U16)nodeNb;
nodeNb++;
}
/* distribute weights (unlimited tree height) */
huffNode[nodeRoot].nbBits = 0;
for (n=nodeRoot-1; n>=STARTNODE; n--)
huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
for (n=0; n<=nonNullRank; n++)
huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
/* enforce maxTableLog */
maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
/* fill result into tree (val, nbBits) */
{ U16 nbPerRank[HUF_TABLELOG_MAX+1] = {0};
U16 valPerRank[HUF_TABLELOG_MAX+1] = {0};
int const alphabetSize = (int)(maxSymbolValue + 1);
if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */
for (n=0; n<=nonNullRank; n++)
nbPerRank[huffNode[n].nbBits]++;
/* determine stating value per rank */
{ U16 min = 0;
for (n=(int)maxNbBits; n>0; n--) {
valPerRank[n] = min; /* get starting value within each rank */
min += nbPerRank[n];
min >>= 1;
} }
for (n=0; n<alphabetSize; n++)
tree[huffNode[n].byte].nbBits = huffNode[n].nbBits; /* push nbBits per symbol, symbol order */
for (n=0; n<alphabetSize; n++)
tree[n].val = valPerRank[tree[n].nbBits]++; /* assign value within rank, symbol order */
}
return maxNbBits;
}
/** HUF_buildCTable() :
* @return : maxNbBits
* Note : count is used before tree is written, so they can safely overlap
*/
size_t HUF_buildCTable (HUF_CElt* tree, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits)
{
HUF_buildCTable_wksp_tables workspace;
return HUF_buildCTable_wksp(tree, count, maxSymbolValue, maxNbBits, &workspace, sizeof(workspace));
}
size_t HUF_estimateCompressedSize(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue)
{
size_t nbBits = 0;
int s;
for (s = 0; s <= (int)maxSymbolValue; ++s) {
nbBits += CTable[s].nbBits * count[s];
}
return nbBits >> 3;
}
int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue) {
int bad = 0;
int s;
for (s = 0; s <= (int)maxSymbolValue; ++s) {
bad |= (count[s] != 0) & (CTable[s].nbBits == 0);
}
return !bad;
}
size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
FORCE_INLINE_TEMPLATE void
HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
{
BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
}
#define HUF_FLUSHBITS(s) BIT_flushBits(s)
#define HUF_FLUSHBITS_1(stream) \
if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*2+7) HUF_FLUSHBITS(stream)
#define HUF_FLUSHBITS_2(stream) \
if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*4+7) HUF_FLUSHBITS(stream)
FORCE_INLINE_TEMPLATE size_t
HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
const BYTE* ip = (const BYTE*) src;
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart;
size_t n;
BIT_CStream_t bitC;
/* init */
if (dstSize < 8) return 0; /* not enough space to compress */
{ size_t const initErr = BIT_initCStream(&bitC, op, (size_t)(oend-op));
if (HUF_isError(initErr)) return 0; }
n = srcSize & ~3; /* join to mod 4 */
switch (srcSize & 3)
{
case 3 : HUF_encodeSymbol(&bitC, ip[n+ 2], CTable);
HUF_FLUSHBITS_2(&bitC);
/* fall-through */
case 2 : HUF_encodeSymbol(&bitC, ip[n+ 1], CTable);
HUF_FLUSHBITS_1(&bitC);
/* fall-through */
case 1 : HUF_encodeSymbol(&bitC, ip[n+ 0], CTable);
HUF_FLUSHBITS(&bitC);
/* fall-through */
case 0 : /* fall-through */
default: break;
}
for (; n>0; n-=4) { /* note : n&3==0 at this stage */
HUF_encodeSymbol(&bitC, ip[n- 1], CTable);
HUF_FLUSHBITS_1(&bitC);
HUF_encodeSymbol(&bitC, ip[n- 2], CTable);
HUF_FLUSHBITS_2(&bitC);
HUF_encodeSymbol(&bitC, ip[n- 3], CTable);
HUF_FLUSHBITS_1(&bitC);
HUF_encodeSymbol(&bitC, ip[n- 4], CTable);
HUF_FLUSHBITS(&bitC);
}
return BIT_closeCStream(&bitC);
}
#if DYNAMIC_BMI2
static TARGET_ATTRIBUTE("bmi2") size_t
HUF_compress1X_usingCTable_internal_bmi2(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
static size_t
HUF_compress1X_usingCTable_internal_default(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
static size_t
HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, const int bmi2)
{
if (bmi2) {
return HUF_compress1X_usingCTable_internal_bmi2(dst, dstSize, src, srcSize, CTable);
}
return HUF_compress1X_usingCTable_internal_default(dst, dstSize, src, srcSize, CTable);
}
#else
static size_t
HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, const int bmi2)
{
(void)bmi2;
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
#endif
size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
}
static size_t
HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, int bmi2)
{
size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */
const BYTE* ip = (const BYTE*) src;
const BYTE* const iend = ip + srcSize;
BYTE* const ostart = (BYTE*) dst;
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart;
if (dstSize < 6 + 1 + 1 + 1 + 8) return 0; /* minimum space to compress successfully */
if (srcSize < 12) return 0; /* no saving possible : too small input */
op += 6; /* jumpTable */
assert(op <= oend);
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart, (U16)cSize);
op += cSize;
}
ip += segmentSize;
assert(op <= oend);
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart+2, (U16)cSize);
op += cSize;
}
ip += segmentSize;
assert(op <= oend);
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart+4, (U16)cSize);
op += cSize;
}
ip += segmentSize;
assert(op <= oend);
assert(ip <= iend);
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, (size_t)(oend-op), ip, (size_t)(iend-ip), CTable, bmi2) );
if (cSize==0) return 0;
op += cSize;
}
return (size_t)(op-ostart);
}
size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
{
return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
}
typedef enum { HUF_singleStream, HUF_fourStreams } HUF_nbStreams_e;
static size_t HUF_compressCTable_internal(
BYTE* const ostart, BYTE* op, BYTE* const oend,
const void* src, size_t srcSize,
HUF_nbStreams_e nbStreams, const HUF_CElt* CTable, const int bmi2)
{
size_t const cSize = (nbStreams==HUF_singleStream) ?
HUF_compress1X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2) :
HUF_compress4X_usingCTable_internal(op, (size_t)(oend - op), src, srcSize, CTable, bmi2);
if (HUF_isError(cSize)) { return cSize; }
if (cSize==0) { return 0; } /* uncompressible */
op += cSize;
/* check compressibility */
assert(op >= ostart);
if ((size_t)(op-ostart) >= srcSize-1) { return 0; }
return (size_t)(op-ostart);
}
typedef struct {
unsigned count[HUF_SYMBOLVALUE_MAX + 1];
HUF_CElt CTable[HUF_SYMBOLVALUE_MAX + 1];
HUF_buildCTable_wksp_tables buildCTable_wksp;
} HUF_compress_tables_t;
/* HUF_compress_internal() :
* `workSpace` must a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
static size_t
HUF_compress_internal (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
HUF_nbStreams_e nbStreams,
void* workSpace, size_t wkspSize,
HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat,
const int bmi2)
{
HUF_compress_tables_t* const table = (HUF_compress_tables_t*)workSpace;
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart;
HUF_STATIC_ASSERT(sizeof(*table) <= HUF_WORKSPACE_SIZE);
/* checks & inits */
if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (wkspSize < HUF_WORKSPACE_SIZE) return ERROR(workSpace_tooSmall);
if (!srcSize) return 0; /* Uncompressed */
if (!dstSize) return 0; /* cannot fit anything within dst budget */
if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); /* current block size limit */
if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
/* Heuristic : If old table is valid, use it for small inputs */
if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
nbStreams, oldHufTable, bmi2);
}
/* Scan input and build symbol stats */
{ CHECK_V_F(largest, HIST_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, workSpace, wkspSize) );
if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */
if (largest <= (srcSize >> 7)+4) return 0; /* heuristic : probably not compressible enough */
}
/* Check validity of previous table */
if ( repeat
&& *repeat == HUF_repeat_check
&& !HUF_validateCTable(oldHufTable, table->count, maxSymbolValue)) {
*repeat = HUF_repeat_none;
}
/* Heuristic : use existing table for small inputs */
if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
nbStreams, oldHufTable, bmi2);
}
/* Build Huffman Tree */
huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
{ size_t const maxBits = HUF_buildCTable_wksp(table->CTable, table->count,
maxSymbolValue, huffLog,
&table->buildCTable_wksp, sizeof(table->buildCTable_wksp));
CHECK_F(maxBits);
huffLog = (U32)maxBits;
/* Zero unused symbols in CTable, so we can check it for validity */
memset(table->CTable + (maxSymbolValue + 1), 0,
sizeof(table->CTable) - ((maxSymbolValue + 1) * sizeof(HUF_CElt)));
}
/* Write table description header */
{ CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, table->CTable, maxSymbolValue, huffLog) );
/* Check if using previous huffman table is beneficial */
if (repeat && *repeat != HUF_repeat_none) {
size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, table->count, maxSymbolValue);
size_t const newSize = HUF_estimateCompressedSize(table->CTable, table->count, maxSymbolValue);
if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
nbStreams, oldHufTable, bmi2);
} }
/* Use the new huffman table */
if (hSize + 12ul >= srcSize) { return 0; }
op += hSize;
if (repeat) { *repeat = HUF_repeat_none; }
if (oldHufTable)
memcpy(oldHufTable, table->CTable, sizeof(table->CTable)); /* Save new table */
}
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
nbStreams, table->CTable, bmi2);
}
size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize)
{
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_singleStream,
workSpace, wkspSize,
NULL, NULL, 0, 0 /*bmi2*/);
}
size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
{
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_singleStream,
workSpace, wkspSize, hufTable,
repeat, preferRepeat, bmi2);
}
size_t HUF_compress1X (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog)
{
unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
}
/* HUF_compress4X_repeat():
* compress input using 4 streams.
* provide workspace to generate compression tables */
size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize)
{
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_fourStreams,
workSpace, wkspSize,
NULL, NULL, 0, 0 /*bmi2*/);
}
/* HUF_compress4X_repeat():
* compress input using 4 streams.
* re-use an existing huffman compression table */
size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
{
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_fourStreams,
workSpace, wkspSize,
hufTable, repeat, preferRepeat, bmi2);
}
size_t HUF_compress2 (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog)
{
unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
}
size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
{
return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT);
}
module/zstd/lib/compress/zstd_compress.c
+4278
View File
File diff suppressed because it is too large Load Diff
module/zstd/lib/compress/zstd_compress_internal.h
+1125
View File
File diff suppressed because it is too large Load Diff
module/zstd/lib/compress/zstd_compress_literals.c
+158
View File
@@ -0,0 +1,158 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
/*-*************************************
* Dependencies
***************************************/
#include "zstd_compress_literals.h"
size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
{
BYTE* const ostart = (BYTE* const)dst;
U32 const flSize = 1 + (srcSize>31) + (srcSize>4095);
RETURN_ERROR_IF(srcSize + flSize > dstCapacity, dstSize_tooSmall, "");
switch(flSize)
{
case 1: /* 2 - 1 - 5 */
ostart[0] = (BYTE)((U32)set_basic + (srcSize<<3));
break;
case 2: /* 2 - 2 - 12 */
MEM_writeLE16(ostart, (U16)((U32)set_basic + (1<<2) + (srcSize<<4)));
break;
case 3: /* 2 - 2 - 20 */
MEM_writeLE32(ostart, (U32)((U32)set_basic + (3<<2) + (srcSize<<4)));
break;
default: /* not necessary : flSize is {1,2,3} */
assert(0);
}
memcpy(ostart + flSize, src, srcSize);
DEBUGLOG(5, "Raw literals: %u -> %u", (U32)srcSize, (U32)(srcSize + flSize));
return srcSize + flSize;
}
size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
{
BYTE* const ostart = (BYTE* const)dst;
U32 const flSize = 1 + (srcSize>31) + (srcSize>4095);
(void)dstCapacity; /* dstCapacity already guaranteed to be >=4, hence large enough */
switch(flSize)
{
case 1: /* 2 - 1 - 5 */
ostart[0] = (BYTE)((U32)set_rle + (srcSize<<3));
break;
case 2: /* 2 - 2 - 12 */
MEM_writeLE16(ostart, (U16)((U32)set_rle + (1<<2) + (srcSize<<4)));
break;
case 3: /* 2 - 2 - 20 */
MEM_writeLE32(ostart, (U32)((U32)set_rle + (3<<2) + (srcSize<<4)));
break;
default: /* not necessary : flSize is {1,2,3} */
assert(0);
}
ostart[flSize] = *(const BYTE*)src;
DEBUGLOG(5, "RLE literals: %u -> %u", (U32)srcSize, (U32)flSize + 1);
return flSize+1;
}
size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
ZSTD_hufCTables_t* nextHuf,
ZSTD_strategy strategy, int disableLiteralCompression,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
void* entropyWorkspace, size_t entropyWorkspaceSize,
const int bmi2)
{
size_t const minGain = ZSTD_minGain(srcSize, strategy);
size_t const lhSize = 3 + (srcSize >= 1 KB) + (srcSize >= 16 KB);
BYTE* const ostart = (BYTE*)dst;
U32 singleStream = srcSize < 256;
symbolEncodingType_e hType = set_compressed;
size_t cLitSize;
DEBUGLOG(5,"ZSTD_compressLiterals (disableLiteralCompression=%i srcSize=%u)",
disableLiteralCompression, (U32)srcSize);
/* Prepare nextEntropy assuming reusing the existing table */
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
if (disableLiteralCompression)
return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
/* small ? don't even attempt compression (speed opt) */
# define COMPRESS_LITERALS_SIZE_MIN 63
{ size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
if (srcSize <= minLitSize) return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
}
RETURN_ERROR_IF(dstCapacity < lhSize+1, dstSize_tooSmall, "not enough space for compression");
{ HUF_repeat repeat = prevHuf->repeatMode;
int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
cLitSize = singleStream ?
HUF_compress1X_repeat(
ostart+lhSize, dstCapacity-lhSize, src, srcSize,
HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
(HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2) :
HUF_compress4X_repeat(
ostart+lhSize, dstCapacity-lhSize, src, srcSize,
HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
(HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
if (repeat != HUF_repeat_none) {
/* reused the existing table */
DEBUGLOG(5, "Reusing previous huffman table");
hType = set_repeat;
}
}
if ((cLitSize==0) | (cLitSize >= srcSize - minGain) | ERR_isError(cLitSize)) {
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
}
if (cLitSize==1) {
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
return ZSTD_compressRleLiteralsBlock(dst, dstCapacity, src, srcSize);
}
if (hType == set_compressed) {
/* using a newly constructed table */
nextHuf->repeatMode = HUF_repeat_check;
}
/* Build header */
switch(lhSize)
{
case 3: /* 2 - 2 - 10 - 10 */
{ U32 const lhc = hType + ((!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
MEM_writeLE24(ostart, lhc);
break;
}
case 4: /* 2 - 2 - 14 - 14 */
{ U32 const lhc = hType + (2 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<18);
MEM_writeLE32(ostart, lhc);
break;
}
case 5: /* 2 - 2 - 18 - 18 */
{ U32 const lhc = hType + (3 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<22);
MEM_writeLE32(ostart, lhc);
ostart[4] = (BYTE)(cLitSize >> 10);
break;
}
default: /* not possible : lhSize is {3,4,5} */
assert(0);
}
DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)srcSize, (U32)(lhSize+cLitSize));
return lhSize+cLitSize;
}
module/zstd/lib/compress/zstd_compress_literals.h
+29
View File
@@ -0,0 +1,29 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_COMPRESS_LITERALS_H
#define ZSTD_COMPRESS_LITERALS_H
#include "zstd_compress_internal.h" /* ZSTD_hufCTables_t, ZSTD_minGain() */
size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize);
size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize);
size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
ZSTD_hufCTables_t* nextHuf,
ZSTD_strategy strategy, int disableLiteralCompression,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
void* entropyWorkspace, size_t entropyWorkspaceSize,
const int bmi2);
#endif /* ZSTD_COMPRESS_LITERALS_H */
module/zstd/lib/compress/zstd_compress_sequences.c
+419
View File
@@ -0,0 +1,419 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
/*-*************************************
* Dependencies
***************************************/
#include "zstd_compress_sequences.h"
/**
* -log2(x / 256) lookup table for x in [0, 256).
* If x == 0: Return 0
* Else: Return floor(-log2(x / 256) * 256)
*/
static unsigned const kInverseProbabilityLog256[256] = {
0, 2048, 1792, 1642, 1536, 1453, 1386, 1329, 1280, 1236, 1197, 1162,
1130, 1100, 1073, 1047, 1024, 1001, 980, 960, 941, 923, 906, 889,
874, 859, 844, 830, 817, 804, 791, 779, 768, 756, 745, 734,
724, 714, 704, 694, 685, 676, 667, 658, 650, 642, 633, 626,
618, 610, 603, 595, 588, 581, 574, 567, 561, 554, 548, 542,
535, 529, 523, 517, 512, 506, 500, 495, 489, 484, 478, 473,
468, 463, 458, 453, 448, 443, 438, 434, 429, 424, 420, 415,
411, 407, 402, 398, 394, 390, 386, 382, 377, 373, 370, 366,
362, 358, 354, 350, 347, 343, 339, 336, 332, 329, 325, 322,
318, 315, 311, 308, 305, 302, 298, 295, 292, 289, 286, 282,
279, 276, 273, 270, 267, 264, 261, 258, 256, 253, 250, 247,
244, 241, 239, 236, 233, 230, 228, 225, 222, 220, 217, 215,
212, 209, 207, 204, 202, 199, 197, 194, 192, 190, 187, 185,
182, 180, 178, 175, 173, 171, 168, 166, 164, 162, 159, 157,
155, 153, 151, 149, 146, 144, 142, 140, 138, 136, 134, 132,
130, 128, 126, 123, 121, 119, 117, 115, 114, 112, 110, 108,
106, 104, 102, 100, 98, 96, 94, 93, 91, 89, 87, 85,
83, 82, 80, 78, 76, 74, 73, 71, 69, 67, 66, 64,
62, 61, 59, 57, 55, 54, 52, 50, 49, 47, 46, 44,
42, 41, 39, 37, 36, 34, 33, 31, 30, 28, 26, 25,
23, 22, 20, 19, 17, 16, 14, 13, 11, 10, 8, 7,
5, 4, 2, 1,
};
static unsigned ZSTD_getFSEMaxSymbolValue(FSE_CTable const* ctable) {
void const* ptr = ctable;
U16 const* u16ptr = (U16 const*)ptr;
U32 const maxSymbolValue = MEM_read16(u16ptr + 1);
return maxSymbolValue;
}
/**
* Returns the cost in bytes of encoding the normalized count header.
* Returns an error if any of the helper functions return an error.
*/
static size_t ZSTD_NCountCost(unsigned const* count, unsigned const max,
size_t const nbSeq, unsigned const FSELog)
{
BYTE wksp[FSE_NCOUNTBOUND];
S16 norm[MaxSeq + 1];
const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq, max), "");
return FSE_writeNCount(wksp, sizeof(wksp), norm, max, tableLog);
}
/**
* Returns the cost in bits of encoding the distribution described by count
* using the entropy bound.
*/
static size_t ZSTD_entropyCost(unsigned const* count, unsigned const max, size_t const total)
{
unsigned cost = 0;
unsigned s;
for (s = 0; s <= max; ++s) {
unsigned norm = (unsigned)((256 * count[s]) / total);
if (count[s] != 0 && norm == 0)
norm = 1;
assert(count[s] < total);
cost += count[s] * kInverseProbabilityLog256[norm];
}
return cost >> 8;
}
/**
* Returns the cost in bits of encoding the distribution in count using ctable.
* Returns an error if ctable cannot represent all the symbols in count.
*/
size_t ZSTD_fseBitCost(
FSE_CTable const* ctable,
unsigned const* count,
unsigned const max)
{
unsigned const kAccuracyLog = 8;
size_t cost = 0;
unsigned s;
FSE_CState_t cstate;
FSE_initCState(&cstate, ctable);
if (ZSTD_getFSEMaxSymbolValue(ctable) < max) {
DEBUGLOG(5, "Repeat FSE_CTable has maxSymbolValue %u < %u",
ZSTD_getFSEMaxSymbolValue(ctable), max);
return ERROR(GENERIC);
}
for (s = 0; s <= max; ++s) {
unsigned const tableLog = cstate.stateLog;
unsigned const badCost = (tableLog + 1) << kAccuracyLog;
unsigned const bitCost = FSE_bitCost(cstate.symbolTT, tableLog, s, kAccuracyLog);
if (count[s] == 0)
continue;
if (bitCost >= badCost) {
DEBUGLOG(5, "Repeat FSE_CTable has Prob[%u] == 0", s);
return ERROR(GENERIC);
}
cost += (size_t)count[s] * bitCost;
}
return cost >> kAccuracyLog;
}
/**
* Returns the cost in bits of encoding the distribution in count using the
* table described by norm. The max symbol support by norm is assumed >= max.
* norm must be valid for every symbol with non-zero probability in count.
*/
size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
unsigned const* count, unsigned const max)
{
unsigned const shift = 8 - accuracyLog;
size_t cost = 0;
unsigned s;
assert(accuracyLog <= 8);
for (s = 0; s <= max; ++s) {
unsigned const normAcc = (norm[s] != -1) ? (unsigned)norm[s] : 1;
unsigned const norm256 = normAcc << shift;
assert(norm256 > 0);
assert(norm256 < 256);
cost += count[s] * kInverseProbabilityLog256[norm256];
}
return cost >> 8;
}
symbolEncodingType_e
ZSTD_selectEncodingType(
FSE_repeat* repeatMode, unsigned const* count, unsigned const max,
size_t const mostFrequent, size_t nbSeq, unsigned const FSELog,
FSE_CTable const* prevCTable,
short const* defaultNorm, U32 defaultNormLog,
ZSTD_defaultPolicy_e const isDefaultAllowed,
ZSTD_strategy const strategy)
{
ZSTD_STATIC_ASSERT(ZSTD_defaultDisallowed == 0 && ZSTD_defaultAllowed != 0);
if (mostFrequent == nbSeq) {
*repeatMode = FSE_repeat_none;
if (isDefaultAllowed && nbSeq <= 2) {
/* Prefer set_basic over set_rle when there are 2 or less symbols,
* since RLE uses 1 byte, but set_basic uses 5-6 bits per symbol.
* If basic encoding isn't possible, always choose RLE.
*/
DEBUGLOG(5, "Selected set_basic");
return set_basic;
}
DEBUGLOG(5, "Selected set_rle");
return set_rle;
}
if (strategy < ZSTD_lazy) {
if (isDefaultAllowed) {
size_t const staticFse_nbSeq_max = 1000;
size_t const mult = 10 - strategy;
size_t const baseLog = 3;
size_t const dynamicFse_nbSeq_min = (((size_t)1 << defaultNormLog) * mult) >> baseLog; /* 28-36 for offset, 56-72 for lengths */
assert(defaultNormLog >= 5 && defaultNormLog <= 6); /* xx_DEFAULTNORMLOG */
assert(mult <= 9 && mult >= 7);
if ( (*repeatMode == FSE_repeat_valid)
&& (nbSeq < staticFse_nbSeq_max) ) {
DEBUGLOG(5, "Selected set_repeat");
return set_repeat;
}
if ( (nbSeq < dynamicFse_nbSeq_min)
|| (mostFrequent < (nbSeq >> (defaultNormLog-1))) ) {
DEBUGLOG(5, "Selected set_basic");
/* The format allows default tables to be repeated, but it isn't useful.
* When using simple heuristics to select encoding type, we don't want
* to confuse these tables with dictionaries. When running more careful
* analysis, we don't need to waste time checking both repeating tables
* and default tables.
*/
*repeatMode = FSE_repeat_none;
return set_basic;
}
}
} else {
size_t const basicCost = isDefaultAllowed ? ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, count, max) : ERROR(GENERIC);
size_t const repeatCost = *repeatMode != FSE_repeat_none ? ZSTD_fseBitCost(prevCTable, count, max) : ERROR(GENERIC);
size_t const NCountCost = ZSTD_NCountCost(count, max, nbSeq, FSELog);
size_t const compressedCost = (NCountCost << 3) + ZSTD_entropyCost(count, max, nbSeq);
if (isDefaultAllowed) {
assert(!ZSTD_isError(basicCost));
assert(!(*repeatMode == FSE_repeat_valid && ZSTD_isError(repeatCost)));
}
assert(!ZSTD_isError(NCountCost));
assert(compressedCost < ERROR(maxCode));
DEBUGLOG(5, "Estimated bit costs: basic=%u\trepeat=%u\tcompressed=%u",
(unsigned)basicCost, (unsigned)repeatCost, (unsigned)compressedCost);
if (basicCost <= repeatCost && basicCost <= compressedCost) {
DEBUGLOG(5, "Selected set_basic");
assert(isDefaultAllowed);
*repeatMode = FSE_repeat_none;
return set_basic;
}
if (repeatCost <= compressedCost) {
DEBUGLOG(5, "Selected set_repeat");
assert(!ZSTD_isError(repeatCost));
return set_repeat;
}
assert(compressedCost < basicCost && compressedCost < repeatCost);
}
DEBUGLOG(5, "Selected set_compressed");
*repeatMode = FSE_repeat_check;
return set_compressed;
}
size_t
ZSTD_buildCTable(void* dst, size_t dstCapacity,
FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
unsigned* count, U32 max,
const BYTE* codeTable, size_t nbSeq,
const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
const FSE_CTable* prevCTable, size_t prevCTableSize,
void* entropyWorkspace, size_t entropyWorkspaceSize)
{
BYTE* op = (BYTE*)dst;
const BYTE* const oend = op + dstCapacity;
DEBUGLOG(6, "ZSTD_buildCTable (dstCapacity=%u)", (unsigned)dstCapacity);
switch (type) {
case set_rle:
FORWARD_IF_ERROR(FSE_buildCTable_rle(nextCTable, (BYTE)max), "");
RETURN_ERROR_IF(dstCapacity==0, dstSize_tooSmall, "not enough space");
*op = codeTable[0];
return 1;
case set_repeat:
memcpy(nextCTable, prevCTable, prevCTableSize);
return 0;
case set_basic:
FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize), ""); /* note : could be pre-calculated */
return 0;
case set_compressed: {
S16 norm[MaxSeq + 1];
size_t nbSeq_1 = nbSeq;
const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
if (count[codeTable[nbSeq-1]] > 1) {
count[codeTable[nbSeq-1]]--;
nbSeq_1--;
}
assert(nbSeq_1 > 1);
FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max), "");
{ size_t const NCountSize = FSE_writeNCount(op, oend - op, norm, max, tableLog); /* overflow protected */
FORWARD_IF_ERROR(NCountSize, "FSE_writeNCount failed");
FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, entropyWorkspace, entropyWorkspaceSize), "");
return NCountSize;
}
}
default: assert(0); RETURN_ERROR(GENERIC, "impossible to reach");
}
}
FORCE_INLINE_TEMPLATE size_t
ZSTD_encodeSequences_body(
void* dst, size_t dstCapacity,
FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
seqDef const* sequences, size_t nbSeq, int longOffsets)
{
BIT_CStream_t blockStream;
FSE_CState_t stateMatchLength;
FSE_CState_t stateOffsetBits;
FSE_CState_t stateLitLength;
RETURN_ERROR_IF(
ERR_isError(BIT_initCStream(&blockStream, dst, dstCapacity)),
dstSize_tooSmall, "not enough space remaining");
DEBUGLOG(6, "available space for bitstream : %i (dstCapacity=%u)",
(int)(blockStream.endPtr - blockStream.startPtr),
(unsigned)dstCapacity);
/* first symbols */
FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
FSE_initCState2(&stateOffsetBits, CTable_OffsetBits, ofCodeTable[nbSeq-1]);
FSE_initCState2(&stateLitLength, CTable_LitLength, llCodeTable[nbSeq-1]);
BIT_addBits(&blockStream, sequences[nbSeq-1].litLength, LL_bits[llCodeTable[nbSeq-1]]);
if (MEM_32bits()) BIT_flushBits(&blockStream);
BIT_addBits(&blockStream, sequences[nbSeq-1].matchLength, ML_bits[mlCodeTable[nbSeq-1]]);
if (MEM_32bits()) BIT_flushBits(&blockStream);
if (longOffsets) {
U32 const ofBits = ofCodeTable[nbSeq-1];
unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
if (extraBits) {
BIT_addBits(&blockStream, sequences[nbSeq-1].offset, extraBits);
BIT_flushBits(&blockStream);
}
BIT_addBits(&blockStream, sequences[nbSeq-1].offset >> extraBits,
ofBits - extraBits);
} else {
BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
}
BIT_flushBits(&blockStream);
{ size_t n;
for (n=nbSeq-2 ; n<nbSeq ; n--) { /* intentional underflow */
BYTE const llCode = llCodeTable[n];
BYTE const ofCode = ofCodeTable[n];
BYTE const mlCode = mlCodeTable[n];
U32 const llBits = LL_bits[llCode];
U32 const ofBits = ofCode;
U32 const mlBits = ML_bits[mlCode];
DEBUGLOG(6, "encoding: litlen:%2u - matchlen:%2u - offCode:%7u",
(unsigned)sequences[n].litLength,
(unsigned)sequences[n].matchLength + MINMATCH,
(unsigned)sequences[n].offset);
/* 32b*/ /* 64b*/
/* (7)*/ /* (7)*/
FSE_encodeSymbol(&blockStream, &stateOffsetBits, ofCode); /* 15 */ /* 15 */
FSE_encodeSymbol(&blockStream, &stateMatchLength, mlCode); /* 24 */ /* 24 */
if (MEM_32bits()) BIT_flushBits(&blockStream); /* (7)*/
FSE_encodeSymbol(&blockStream, &stateLitLength, llCode); /* 16 */ /* 33 */
if (MEM_32bits() || (ofBits+mlBits+llBits >= 64-7-(LLFSELog+MLFSELog+OffFSELog)))
BIT_flushBits(&blockStream); /* (7)*/
BIT_addBits(&blockStream, sequences[n].litLength, llBits);
if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
if (MEM_32bits() || (ofBits+mlBits+llBits > 56)) BIT_flushBits(&blockStream);
if (longOffsets) {
unsigned const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
if (extraBits) {
BIT_addBits(&blockStream, sequences[n].offset, extraBits);
BIT_flushBits(&blockStream); /* (7)*/
}
BIT_addBits(&blockStream, sequences[n].offset >> extraBits,
ofBits - extraBits); /* 31 */
} else {
BIT_addBits(&blockStream, sequences[n].offset, ofBits); /* 31 */
}
BIT_flushBits(&blockStream); /* (7)*/
DEBUGLOG(7, "remaining space : %i", (int)(blockStream.endPtr - blockStream.ptr));
} }
DEBUGLOG(6, "ZSTD_encodeSequences: flushing ML state with %u bits", stateMatchLength.stateLog);
FSE_flushCState(&blockStream, &stateMatchLength);
DEBUGLOG(6, "ZSTD_encodeSequences: flushing Off state with %u bits", stateOffsetBits.stateLog);
FSE_flushCState(&blockStream, &stateOffsetBits);
DEBUGLOG(6, "ZSTD_encodeSequences: flushing LL state with %u bits", stateLitLength.stateLog);
FSE_flushCState(&blockStream, &stateLitLength);
{ size_t const streamSize = BIT_closeCStream(&blockStream);
RETURN_ERROR_IF(streamSize==0, dstSize_tooSmall, "not enough space");
return streamSize;
}
}
static size_t
ZSTD_encodeSequences_default(
void* dst, size_t dstCapacity,
FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
seqDef const* sequences, size_t nbSeq, int longOffsets)
{
return ZSTD_encodeSequences_body(dst, dstCapacity,
CTable_MatchLength, mlCodeTable,
CTable_OffsetBits, ofCodeTable,
CTable_LitLength, llCodeTable,
sequences, nbSeq, longOffsets);
}
#if DYNAMIC_BMI2
static TARGET_ATTRIBUTE("bmi2") size_t
ZSTD_encodeSequences_bmi2(
void* dst, size_t dstCapacity,
FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
seqDef const* sequences, size_t nbSeq, int longOffsets)
{
return ZSTD_encodeSequences_body(dst, dstCapacity,
CTable_MatchLength, mlCodeTable,
CTable_OffsetBits, ofCodeTable,
CTable_LitLength, llCodeTable,
sequences, nbSeq, longOffsets);
}
#endif
size_t ZSTD_encodeSequences(
void* dst, size_t dstCapacity,
FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2)
{
DEBUGLOG(5, "ZSTD_encodeSequences: dstCapacity = %u", (unsigned)dstCapacity);
#if DYNAMIC_BMI2
if (bmi2) {
return ZSTD_encodeSequences_bmi2(dst, dstCapacity,
CTable_MatchLength, mlCodeTable,
CTable_OffsetBits, ofCodeTable,
CTable_LitLength, llCodeTable,
sequences, nbSeq, longOffsets);
}
#endif
(void)bmi2;
return ZSTD_encodeSequences_default(dst, dstCapacity,
CTable_MatchLength, mlCodeTable,
CTable_OffsetBits, ofCodeTable,
CTable_LitLength, llCodeTable,
sequences, nbSeq, longOffsets);
}
module/zstd/lib/compress/zstd_compress_sequences.h
+54
View File
@@ -0,0 +1,54 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_COMPRESS_SEQUENCES_H
#define ZSTD_COMPRESS_SEQUENCES_H
#include "../common/fse.h" /* FSE_repeat, FSE_CTable */
#include "../common/zstd_internal.h" /* symbolEncodingType_e, ZSTD_strategy */
typedef enum {
ZSTD_defaultDisallowed = 0,
ZSTD_defaultAllowed = 1
} ZSTD_defaultPolicy_e;
symbolEncodingType_e
ZSTD_selectEncodingType(
FSE_repeat* repeatMode, unsigned const* count, unsigned const max,
size_t const mostFrequent, size_t nbSeq, unsigned const FSELog,
FSE_CTable const* prevCTable,
short const* defaultNorm, U32 defaultNormLog,
ZSTD_defaultPolicy_e const isDefaultAllowed,
ZSTD_strategy const strategy);
size_t
ZSTD_buildCTable(void* dst, size_t dstCapacity,
FSE_CTable* nextCTable, U32 FSELog, symbolEncodingType_e type,
unsigned* count, U32 max,
const BYTE* codeTable, size_t nbSeq,
const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
const FSE_CTable* prevCTable, size_t prevCTableSize,
void* entropyWorkspace, size_t entropyWorkspaceSize);
size_t ZSTD_encodeSequences(
void* dst, size_t dstCapacity,
FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
seqDef const* sequences, size_t nbSeq, int longOffsets, int bmi2);
size_t ZSTD_fseBitCost(
FSE_CTable const* ctable,
unsigned const* count,
unsigned const max);
size_t ZSTD_crossEntropyCost(short const* norm, unsigned accuracyLog,
unsigned const* count, unsigned const max);
#endif /* ZSTD_COMPRESS_SEQUENCES_H */
module/zstd/lib/compress/zstd_compress_superblock.c
+845
View File
@@ -0,0 +1,845 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
/*-*************************************
* Dependencies
***************************************/
#include "zstd_compress_superblock.h"
#include "../common/zstd_internal.h" /* ZSTD_getSequenceLength */
#include "hist.h" /* HIST_countFast_wksp */
#include "zstd_compress_internal.h"
#include "zstd_compress_sequences.h"
#include "zstd_compress_literals.h"
/*-*************************************
* Superblock entropy buffer structs
***************************************/
/** ZSTD_hufCTablesMetadata_t :
* Stores Literals Block Type for a super-block in hType, and
* huffman tree description in hufDesBuffer.
* hufDesSize refers to the size of huffman tree description in bytes.
* This metadata is populated in ZSTD_buildSuperBlockEntropy_literal() */
typedef struct {
symbolEncodingType_e hType;
BYTE hufDesBuffer[500]; /* TODO give name to this value */
size_t hufDesSize;
} ZSTD_hufCTablesMetadata_t;
/** ZSTD_fseCTablesMetadata_t :
* Stores symbol compression modes for a super-block in {ll, ol, ml}Type, and
* fse tables in fseTablesBuffer.
* fseTablesSize refers to the size of fse tables in bytes.
* This metadata is populated in ZSTD_buildSuperBlockEntropy_sequences() */
typedef struct {
symbolEncodingType_e llType;
symbolEncodingType_e ofType;
symbolEncodingType_e mlType;
BYTE fseTablesBuffer[500]; /* TODO give name to this value */
size_t fseTablesSize;
size_t lastCountSize; /* This is to account for bug in 1.3.4. More detail in ZSTD_compressSubBlock_sequences() */
} ZSTD_fseCTablesMetadata_t;
typedef struct {
ZSTD_hufCTablesMetadata_t hufMetadata;
ZSTD_fseCTablesMetadata_t fseMetadata;
} ZSTD_entropyCTablesMetadata_t;
/** ZSTD_buildSuperBlockEntropy_literal() :
* Builds entropy for the super-block literals.
* Stores literals block type (raw, rle, compressed, repeat) and
* huffman description table to hufMetadata.
* @return : size of huffman description table or error code */
static size_t ZSTD_buildSuperBlockEntropy_literal(void* const src, size_t srcSize,
const ZSTD_hufCTables_t* prevHuf,
ZSTD_hufCTables_t* nextHuf,
ZSTD_hufCTablesMetadata_t* hufMetadata,
const int disableLiteralsCompression,
void* workspace, size_t wkspSize)
{
BYTE* const wkspStart = (BYTE*)workspace;
BYTE* const wkspEnd = wkspStart + wkspSize;
BYTE* const countWkspStart = wkspStart;
unsigned* const countWksp = (unsigned*)workspace;
const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned);
BYTE* const nodeWksp = countWkspStart + countWkspSize;
const size_t nodeWkspSize = wkspEnd-nodeWksp;
unsigned maxSymbolValue = 255;
unsigned huffLog = HUF_TABLELOG_DEFAULT;
HUF_repeat repeat = prevHuf->repeatMode;
DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_literal (srcSize=%zu)", srcSize);
/* Prepare nextEntropy assuming reusing the existing table */
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
if (disableLiteralsCompression) {
DEBUGLOG(5, "set_basic - disabled");
hufMetadata->hType = set_basic;
return 0;
}
/* small ? don't even attempt compression (speed opt) */
# define COMPRESS_LITERALS_SIZE_MIN 63
{ size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
if (srcSize <= minLitSize) {
DEBUGLOG(5, "set_basic - too small");
hufMetadata->hType = set_basic;
return 0;
}
}
/* Scan input and build symbol stats */
{ size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)src, srcSize, workspace, wkspSize);
FORWARD_IF_ERROR(largest, "HIST_count_wksp failed");
if (largest == srcSize) {
DEBUGLOG(5, "set_rle");
hufMetadata->hType = set_rle;
return 0;
}
if (largest <= (srcSize >> 7)+4) {
DEBUGLOG(5, "set_basic - no gain");
hufMetadata->hType = set_basic;
return 0;
}
}
/* Validate the previous Huffman table */
if (repeat == HUF_repeat_check && !HUF_validateCTable((HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue)) {
repeat = HUF_repeat_none;
}
/* Build Huffman Tree */
memset(nextHuf->CTable, 0, sizeof(nextHuf->CTable));
huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
{ size_t const maxBits = HUF_buildCTable_wksp((HUF_CElt*)nextHuf->CTable, countWksp,
maxSymbolValue, huffLog,
nodeWksp, nodeWkspSize);
FORWARD_IF_ERROR(maxBits, "HUF_buildCTable_wksp");
huffLog = (U32)maxBits;
{ /* Build and write the CTable */
size_t const newCSize = HUF_estimateCompressedSize(
(HUF_CElt*)nextHuf->CTable, countWksp, maxSymbolValue);
size_t const hSize = HUF_writeCTable(
hufMetadata->hufDesBuffer, sizeof(hufMetadata->hufDesBuffer),
(HUF_CElt*)nextHuf->CTable, maxSymbolValue, huffLog);
/* Check against repeating the previous CTable */
if (repeat != HUF_repeat_none) {
size_t const oldCSize = HUF_estimateCompressedSize(
(HUF_CElt const*)prevHuf->CTable, countWksp, maxSymbolValue);
if (oldCSize < srcSize && (oldCSize <= hSize + newCSize || hSize + 12 >= srcSize)) {
DEBUGLOG(5, "set_repeat - smaller");
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
hufMetadata->hType = set_repeat;
return 0;
}
}
if (newCSize + hSize >= srcSize) {
DEBUGLOG(5, "set_basic - no gains");
memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
hufMetadata->hType = set_basic;
return 0;
}
DEBUGLOG(5, "set_compressed (hSize=%u)", (U32)hSize);
hufMetadata->hType = set_compressed;
nextHuf->repeatMode = HUF_repeat_check;
return hSize;
}
}
}
/** ZSTD_buildSuperBlockEntropy_sequences() :
* Builds entropy for the super-block sequences.
* Stores symbol compression modes and fse table to fseMetadata.
* @return : size of fse tables or error code */
static size_t ZSTD_buildSuperBlockEntropy_sequences(seqStore_t* seqStorePtr,
const ZSTD_fseCTables_t* prevEntropy,
ZSTD_fseCTables_t* nextEntropy,
const ZSTD_CCtx_params* cctxParams,
ZSTD_fseCTablesMetadata_t* fseMetadata,
void* workspace, size_t wkspSize)
{
BYTE* const wkspStart = (BYTE*)workspace;
BYTE* const wkspEnd = wkspStart + wkspSize;
BYTE* const countWkspStart = wkspStart;
unsigned* const countWksp = (unsigned*)workspace;
const size_t countWkspSize = (MaxSeq + 1) * sizeof(unsigned);
BYTE* const cTableWksp = countWkspStart + countWkspSize;
const size_t cTableWkspSize = wkspEnd-cTableWksp;
ZSTD_strategy const strategy = cctxParams->cParams.strategy;
FSE_CTable* CTable_LitLength = nextEntropy->litlengthCTable;
FSE_CTable* CTable_OffsetBits = nextEntropy->offcodeCTable;
FSE_CTable* CTable_MatchLength = nextEntropy->matchlengthCTable;
const BYTE* const ofCodeTable = seqStorePtr->ofCode;
const BYTE* const llCodeTable = seqStorePtr->llCode;
const BYTE* const mlCodeTable = seqStorePtr->mlCode;
size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
BYTE* const ostart = fseMetadata->fseTablesBuffer;
BYTE* const oend = ostart + sizeof(fseMetadata->fseTablesBuffer);
BYTE* op = ostart;
assert(cTableWkspSize >= (1 << MaxFSELog) * sizeof(FSE_FUNCTION_TYPE));
DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy_sequences (nbSeq=%zu)", nbSeq);
memset(workspace, 0, wkspSize);
fseMetadata->lastCountSize = 0;
/* convert length/distances into codes */
ZSTD_seqToCodes(seqStorePtr);
/* build CTable for Literal Lengths */
{ U32 LLtype;
unsigned max = MaxLL;
size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, llCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
DEBUGLOG(5, "Building LL table");
nextEntropy->litlength_repeatMode = prevEntropy->litlength_repeatMode;
LLtype = ZSTD_selectEncodingType(&nextEntropy->litlength_repeatMode,
countWksp, max, mostFrequent, nbSeq,
LLFSELog, prevEntropy->litlengthCTable,
LL_defaultNorm, LL_defaultNormLog,
ZSTD_defaultAllowed, strategy);
assert(set_basic < set_compressed && set_rle < set_compressed);
assert(!(LLtype < set_compressed && nextEntropy->litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
{ size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
countWksp, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
prevEntropy->litlengthCTable, sizeof(prevEntropy->litlengthCTable),
cTableWksp, cTableWkspSize);
FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for LitLens failed");
if (LLtype == set_compressed)
fseMetadata->lastCountSize = countSize;
op += countSize;
fseMetadata->llType = (symbolEncodingType_e) LLtype;
} }
/* build CTable for Offsets */
{ U32 Offtype;
unsigned max = MaxOff;
size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, ofCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
/* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
DEBUGLOG(5, "Building OF table");
nextEntropy->offcode_repeatMode = prevEntropy->offcode_repeatMode;
Offtype = ZSTD_selectEncodingType(&nextEntropy->offcode_repeatMode,
countWksp, max, mostFrequent, nbSeq,
OffFSELog, prevEntropy->offcodeCTable,
OF_defaultNorm, OF_defaultNormLog,
defaultPolicy, strategy);
assert(!(Offtype < set_compressed && nextEntropy->offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
{ size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
countWksp, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
prevEntropy->offcodeCTable, sizeof(prevEntropy->offcodeCTable),
cTableWksp, cTableWkspSize);
FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for Offsets failed");
if (Offtype == set_compressed)
fseMetadata->lastCountSize = countSize;
op += countSize;
fseMetadata->ofType = (symbolEncodingType_e) Offtype;
} }
/* build CTable for MatchLengths */
{ U32 MLtype;
unsigned max = MaxML;
size_t const mostFrequent = HIST_countFast_wksp(countWksp, &max, mlCodeTable, nbSeq, workspace, wkspSize); /* can't fail */
DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
nextEntropy->matchlength_repeatMode = prevEntropy->matchlength_repeatMode;
MLtype = ZSTD_selectEncodingType(&nextEntropy->matchlength_repeatMode,
countWksp, max, mostFrequent, nbSeq,
MLFSELog, prevEntropy->matchlengthCTable,
ML_defaultNorm, ML_defaultNormLog,
ZSTD_defaultAllowed, strategy);
assert(!(MLtype < set_compressed && nextEntropy->matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
{ size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
countWksp, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
prevEntropy->matchlengthCTable, sizeof(prevEntropy->matchlengthCTable),
cTableWksp, cTableWkspSize);
FORWARD_IF_ERROR(countSize, "ZSTD_buildCTable for MatchLengths failed");
if (MLtype == set_compressed)
fseMetadata->lastCountSize = countSize;
op += countSize;
fseMetadata->mlType = (symbolEncodingType_e) MLtype;
} }
assert((size_t) (op-ostart) <= sizeof(fseMetadata->fseTablesBuffer));
return op-ostart;
}
/** ZSTD_buildSuperBlockEntropy() :
* Builds entropy for the super-block.
* @return : 0 on success or error code */
static size_t
ZSTD_buildSuperBlockEntropy(seqStore_t* seqStorePtr,
const ZSTD_entropyCTables_t* prevEntropy,
ZSTD_entropyCTables_t* nextEntropy,
const ZSTD_CCtx_params* cctxParams,
ZSTD_entropyCTablesMetadata_t* entropyMetadata,
void* workspace, size_t wkspSize)
{
size_t const litSize = seqStorePtr->lit - seqStorePtr->litStart;
DEBUGLOG(5, "ZSTD_buildSuperBlockEntropy");
entropyMetadata->hufMetadata.hufDesSize =
ZSTD_buildSuperBlockEntropy_literal(seqStorePtr->litStart, litSize,
&prevEntropy->huf, &nextEntropy->huf,
&entropyMetadata->hufMetadata,
ZSTD_disableLiteralsCompression(cctxParams),
workspace, wkspSize);
FORWARD_IF_ERROR(entropyMetadata->hufMetadata.hufDesSize, "ZSTD_buildSuperBlockEntropy_literal failed");
entropyMetadata->fseMetadata.fseTablesSize =
ZSTD_buildSuperBlockEntropy_sequences(seqStorePtr,
&prevEntropy->fse, &nextEntropy->fse,
cctxParams,
&entropyMetadata->fseMetadata,
workspace, wkspSize);
FORWARD_IF_ERROR(entropyMetadata->fseMetadata.fseTablesSize, "ZSTD_buildSuperBlockEntropy_sequences failed");
return 0;
}
/** ZSTD_compressSubBlock_literal() :
* Compresses literals section for a sub-block.
* When we have to write the Huffman table we will sometimes choose a header
* size larger than necessary. This is because we have to pick the header size
* before we know the table size + compressed size, so we have a bound on the
* table size. If we guessed incorrectly, we fall back to uncompressed literals.
*
* We write the header when writeEntropy=1 and set entropyWrriten=1 when we succeeded
* in writing the header, otherwise it is set to 0.
*
* hufMetadata->hType has literals block type info.
* If it is set_basic, all sub-blocks literals section will be Raw_Literals_Block.
* If it is set_rle, all sub-blocks literals section will be RLE_Literals_Block.
* If it is set_compressed, first sub-block's literals section will be Compressed_Literals_Block
* If it is set_compressed, first sub-block's literals section will be Treeless_Literals_Block
* and the following sub-blocks' literals sections will be Treeless_Literals_Block.
* @return : compressed size of literals section of a sub-block
* Or 0 if it unable to compress.
* Or error code */
static size_t ZSTD_compressSubBlock_literal(const HUF_CElt* hufTable,
const ZSTD_hufCTablesMetadata_t* hufMetadata,
const BYTE* literals, size_t litSize,
void* dst, size_t dstSize,
const int bmi2, int writeEntropy, int* entropyWritten)
{
size_t const header = writeEntropy ? 200 : 0;
size_t const lhSize = 3 + (litSize >= (1 KB - header)) + (litSize >= (16 KB - header));
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart + lhSize;
U32 const singleStream = lhSize == 3;
symbolEncodingType_e hType = writeEntropy ? hufMetadata->hType : set_repeat;
size_t cLitSize = 0;
(void)bmi2; /* TODO bmi2... */
DEBUGLOG(5, "ZSTD_compressSubBlock_literal (litSize=%zu, lhSize=%zu, writeEntropy=%d)", litSize, lhSize, writeEntropy);
*entropyWritten = 0;
if (litSize == 0 || hufMetadata->hType == set_basic) {
DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal");
return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
} else if (hufMetadata->hType == set_rle) {
DEBUGLOG(5, "ZSTD_compressSubBlock_literal using rle literal");
return ZSTD_compressRleLiteralsBlock(dst, dstSize, literals, litSize);
}
assert(litSize > 0);
assert(hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat);
if (writeEntropy && hufMetadata->hType == set_compressed) {
memcpy(op, hufMetadata->hufDesBuffer, hufMetadata->hufDesSize);
op += hufMetadata->hufDesSize;
cLitSize += hufMetadata->hufDesSize;
DEBUGLOG(5, "ZSTD_compressSubBlock_literal (hSize=%zu)", hufMetadata->hufDesSize);
}
/* TODO bmi2 */
{ const size_t cSize = singleStream ? HUF_compress1X_usingCTable(op, oend-op, literals, litSize, hufTable)
: HUF_compress4X_usingCTable(op, oend-op, literals, litSize, hufTable);
op += cSize;
cLitSize += cSize;
if (cSize == 0 || ERR_isError(cSize)) {
DEBUGLOG(5, "Failed to write entropy tables %s", ZSTD_getErrorName(cSize));
return 0;
}
/* If we expand and we aren't writing a header then emit uncompressed */
if (!writeEntropy && cLitSize >= litSize) {
DEBUGLOG(5, "ZSTD_compressSubBlock_literal using raw literal because uncompressible");
return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
}
/* If we are writing headers then allow expansion that doesn't change our header size. */
if (lhSize < (size_t)(3 + (cLitSize >= 1 KB) + (cLitSize >= 16 KB))) {
assert(cLitSize > litSize);
DEBUGLOG(5, "Literals expanded beyond allowed header size");
return ZSTD_noCompressLiterals(dst, dstSize, literals, litSize);
}
DEBUGLOG(5, "ZSTD_compressSubBlock_literal (cSize=%zu)", cSize);
}
/* Build header */
switch(lhSize)
{
case 3: /* 2 - 2 - 10 - 10 */
{ U32 const lhc = hType + ((!singleStream) << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<14);
MEM_writeLE24(ostart, lhc);
break;
}
case 4: /* 2 - 2 - 14 - 14 */
{ U32 const lhc = hType + (2 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<18);
MEM_writeLE32(ostart, lhc);
break;
}
case 5: /* 2 - 2 - 18 - 18 */
{ U32 const lhc = hType + (3 << 2) + ((U32)litSize<<4) + ((U32)cLitSize<<22);
MEM_writeLE32(ostart, lhc);
ostart[4] = (BYTE)(cLitSize >> 10);
break;
}
default: /* not possible : lhSize is {3,4,5} */
assert(0);
}
*entropyWritten = 1;
DEBUGLOG(5, "Compressed literals: %u -> %u", (U32)litSize, (U32)(op-ostart));
return op-ostart;
}
static size_t ZSTD_seqDecompressedSize(seqStore_t const* seqStore, const seqDef* sequences, size_t nbSeq, size_t litSize, int lastSequence) {
const seqDef* const sstart = sequences;
const seqDef* const send = sequences + nbSeq;
const seqDef* sp = sstart;
size_t matchLengthSum = 0;
size_t litLengthSum = 0;
while (send-sp > 0) {
ZSTD_sequenceLength const seqLen = ZSTD_getSequenceLength(seqStore, sp);
litLengthSum += seqLen.litLength;
matchLengthSum += seqLen.matchLength;
sp++;
}
assert(litLengthSum <= litSize);
if (!lastSequence) {
assert(litLengthSum == litSize);
}
return matchLengthSum + litSize;
}
/** ZSTD_compressSubBlock_sequences() :
* Compresses sequences section for a sub-block.
* fseMetadata->llType, fseMetadata->ofType, and fseMetadata->mlType have
* symbol compression modes for the super-block.
* The first successfully compressed block will have these in its header.
* We set entropyWritten=1 when we succeed in compressing the sequences.
* The following sub-blocks will always have repeat mode.
* @return : compressed size of sequences section of a sub-block
* Or 0 if it is unable to compress
* Or error code. */
static size_t ZSTD_compressSubBlock_sequences(const ZSTD_fseCTables_t* fseTables,
const ZSTD_fseCTablesMetadata_t* fseMetadata,
const seqDef* sequences, size_t nbSeq,
const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
const ZSTD_CCtx_params* cctxParams,
void* dst, size_t dstCapacity,
const int bmi2, int writeEntropy, int* entropyWritten)
{
const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstCapacity;
BYTE* op = ostart;
BYTE* seqHead;
DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (nbSeq=%zu, writeEntropy=%d, longOffsets=%d)", nbSeq, writeEntropy, longOffsets);
*entropyWritten = 0;
/* Sequences Header */
RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
dstSize_tooSmall, "");
if (nbSeq < 0x7F)
*op++ = (BYTE)nbSeq;
else if (nbSeq < LONGNBSEQ)
op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
else
op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
if (nbSeq==0) {
return op - ostart;
}
/* seqHead : flags for FSE encoding type */
seqHead = op++;
DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (seqHeadSize=%u)", (unsigned)(op-ostart));
if (writeEntropy) {
const U32 LLtype = fseMetadata->llType;
const U32 Offtype = fseMetadata->ofType;
const U32 MLtype = fseMetadata->mlType;
DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (fseTablesSize=%zu)", fseMetadata->fseTablesSize);
*seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
memcpy(op, fseMetadata->fseTablesBuffer, fseMetadata->fseTablesSize);
op += fseMetadata->fseTablesSize;
} else {
const U32 repeat = set_repeat;
*seqHead = (BYTE)((repeat<<6) + (repeat<<4) + (repeat<<2));
}
{ size_t const bitstreamSize = ZSTD_encodeSequences(
op, oend - op,
fseTables->matchlengthCTable, mlCode,
fseTables->offcodeCTable, ofCode,
fseTables->litlengthCTable, llCode,
sequences, nbSeq,
longOffsets, bmi2);
FORWARD_IF_ERROR(bitstreamSize, "ZSTD_encodeSequences failed");
op += bitstreamSize;
/* zstd versions <= 1.3.4 mistakenly report corruption when
* FSE_readNCount() receives a buffer < 4 bytes.
* Fixed by https://github.com/facebook/zstd/pull/1146.
* This can happen when the last set_compressed table present is 2
* bytes and the bitstream is only one byte.
* In this exceedingly rare case, we will simply emit an uncompressed
* block, since it isn't worth optimizing.
*/
#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
if (writeEntropy && fseMetadata->lastCountSize && fseMetadata->lastCountSize + bitstreamSize < 4) {
/* NCountSize >= 2 && bitstreamSize > 0 ==> lastCountSize == 3 */
assert(fseMetadata->lastCountSize + bitstreamSize == 3);
DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.3.4 by "
"emitting an uncompressed block.");
return 0;
}
#endif
DEBUGLOG(5, "ZSTD_compressSubBlock_sequences (bitstreamSize=%zu)", bitstreamSize);
}
/* zstd versions <= 1.4.0 mistakenly report error when
* sequences section body size is less than 3 bytes.
* Fixed by https://github.com/facebook/zstd/pull/1664.
* This can happen when the previous sequences section block is compressed
* with rle mode and the current block's sequences section is compressed
* with repeat mode where sequences section body size can be 1 byte.
*/
#ifndef FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION
if (op-seqHead < 4) {
DEBUGLOG(5, "Avoiding bug in zstd decoder in versions <= 1.4.0 by emitting "
"an uncompressed block when sequences are < 4 bytes");
return 0;
}
#endif
*entropyWritten = 1;
return op - ostart;
}
/** ZSTD_compressSubBlock() :
* Compresses a single sub-block.
* @return : compressed size of the sub-block
* Or 0 if it failed to compress. */
static size_t ZSTD_compressSubBlock(const ZSTD_entropyCTables_t* entropy,
const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
const seqDef* sequences, size_t nbSeq,
const BYTE* literals, size_t litSize,
const BYTE* llCode, const BYTE* mlCode, const BYTE* ofCode,
const ZSTD_CCtx_params* cctxParams,
void* dst, size_t dstCapacity,
const int bmi2,
int writeLitEntropy, int writeSeqEntropy,
int* litEntropyWritten, int* seqEntropyWritten,
U32 lastBlock)
{
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstCapacity;
BYTE* op = ostart + ZSTD_blockHeaderSize;
DEBUGLOG(5, "ZSTD_compressSubBlock (litSize=%zu, nbSeq=%zu, writeLitEntropy=%d, writeSeqEntropy=%d, lastBlock=%d)",
litSize, nbSeq, writeLitEntropy, writeSeqEntropy, lastBlock);
{ size_t cLitSize = ZSTD_compressSubBlock_literal((const HUF_CElt*)entropy->huf.CTable,
&entropyMetadata->hufMetadata, literals, litSize,
op, oend-op, bmi2, writeLitEntropy, litEntropyWritten);
FORWARD_IF_ERROR(cLitSize, "ZSTD_compressSubBlock_literal failed");
if (cLitSize == 0) return 0;
op += cLitSize;
}
{ size_t cSeqSize = ZSTD_compressSubBlock_sequences(&entropy->fse,
&entropyMetadata->fseMetadata,
sequences, nbSeq,
llCode, mlCode, ofCode,
cctxParams,
op, oend-op,
bmi2, writeSeqEntropy, seqEntropyWritten);
FORWARD_IF_ERROR(cSeqSize, "ZSTD_compressSubBlock_sequences failed");
if (cSeqSize == 0) return 0;
op += cSeqSize;
}
/* Write block header */
{ size_t cSize = (op-ostart)-ZSTD_blockHeaderSize;
U32 const cBlockHeader24 = lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
MEM_writeLE24(ostart, cBlockHeader24);
}
return op-ostart;
}
static size_t ZSTD_estimateSubBlockSize_literal(const BYTE* literals, size_t litSize,
const ZSTD_hufCTables_t* huf,
const ZSTD_hufCTablesMetadata_t* hufMetadata,
void* workspace, size_t wkspSize,
int writeEntropy)
{
unsigned* const countWksp = (unsigned*)workspace;
unsigned maxSymbolValue = 255;
size_t literalSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
if (hufMetadata->hType == set_basic) return litSize;
else if (hufMetadata->hType == set_rle) return 1;
else if (hufMetadata->hType == set_compressed || hufMetadata->hType == set_repeat) {
size_t const largest = HIST_count_wksp (countWksp, &maxSymbolValue, (const BYTE*)literals, litSize, workspace, wkspSize);
if (ZSTD_isError(largest)) return litSize;
{ size_t cLitSizeEstimate = HUF_estimateCompressedSize((const HUF_CElt*)huf->CTable, countWksp, maxSymbolValue);
if (writeEntropy) cLitSizeEstimate += hufMetadata->hufDesSize;
return cLitSizeEstimate + literalSectionHeaderSize;
} }
assert(0); /* impossible */
return 0;
}
static size_t ZSTD_estimateSubBlockSize_symbolType(symbolEncodingType_e type,
const BYTE* codeTable, unsigned maxCode,
size_t nbSeq, const FSE_CTable* fseCTable,
const U32* additionalBits,
short const* defaultNorm, U32 defaultNormLog,
void* workspace, size_t wkspSize)
{
unsigned* const countWksp = (unsigned*)workspace;
const BYTE* ctp = codeTable;
const BYTE* const ctStart = ctp;
const BYTE* const ctEnd = ctStart + nbSeq;
size_t cSymbolTypeSizeEstimateInBits = 0;
unsigned max = maxCode;
HIST_countFast_wksp(countWksp, &max, codeTable, nbSeq, workspace, wkspSize); /* can't fail */
if (type == set_basic) {
cSymbolTypeSizeEstimateInBits = ZSTD_crossEntropyCost(defaultNorm, defaultNormLog, countWksp, max);
} else if (type == set_rle) {
cSymbolTypeSizeEstimateInBits = 0;
} else if (type == set_compressed || type == set_repeat) {
cSymbolTypeSizeEstimateInBits = ZSTD_fseBitCost(fseCTable, countWksp, max);
}
if (ZSTD_isError(cSymbolTypeSizeEstimateInBits)) return nbSeq * 10;
while (ctp < ctEnd) {
if (additionalBits) cSymbolTypeSizeEstimateInBits += additionalBits[*ctp];
else cSymbolTypeSizeEstimateInBits += *ctp; /* for offset, offset code is also the number of additional bits */
ctp++;
}
return cSymbolTypeSizeEstimateInBits / 8;
}
static size_t ZSTD_estimateSubBlockSize_sequences(const BYTE* ofCodeTable,
const BYTE* llCodeTable,
const BYTE* mlCodeTable,
size_t nbSeq,
const ZSTD_fseCTables_t* fseTables,
const ZSTD_fseCTablesMetadata_t* fseMetadata,
void* workspace, size_t wkspSize,
int writeEntropy)
{
size_t sequencesSectionHeaderSize = 3; /* Use hard coded size of 3 bytes */
size_t cSeqSizeEstimate = 0;
cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->ofType, ofCodeTable, MaxOff,
nbSeq, fseTables->offcodeCTable, NULL,
OF_defaultNorm, OF_defaultNormLog,
workspace, wkspSize);
cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->llType, llCodeTable, MaxLL,
nbSeq, fseTables->litlengthCTable, LL_bits,
LL_defaultNorm, LL_defaultNormLog,
workspace, wkspSize);
cSeqSizeEstimate += ZSTD_estimateSubBlockSize_symbolType(fseMetadata->mlType, mlCodeTable, MaxML,
nbSeq, fseTables->matchlengthCTable, ML_bits,
ML_defaultNorm, ML_defaultNormLog,
workspace, wkspSize);
if (writeEntropy) cSeqSizeEstimate += fseMetadata->fseTablesSize;
return cSeqSizeEstimate + sequencesSectionHeaderSize;
}
static size_t ZSTD_estimateSubBlockSize(const BYTE* literals, size_t litSize,
const BYTE* ofCodeTable,
const BYTE* llCodeTable,
const BYTE* mlCodeTable,
size_t nbSeq,
const ZSTD_entropyCTables_t* entropy,
const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
void* workspace, size_t wkspSize,
int writeLitEntropy, int writeSeqEntropy) {
size_t cSizeEstimate = 0;
cSizeEstimate += ZSTD_estimateSubBlockSize_literal(literals, litSize,
&entropy->huf, &entropyMetadata->hufMetadata,
workspace, wkspSize, writeLitEntropy);
cSizeEstimate += ZSTD_estimateSubBlockSize_sequences(ofCodeTable, llCodeTable, mlCodeTable,
nbSeq, &entropy->fse, &entropyMetadata->fseMetadata,
workspace, wkspSize, writeSeqEntropy);
return cSizeEstimate + ZSTD_blockHeaderSize;
}
static int ZSTD_needSequenceEntropyTables(ZSTD_fseCTablesMetadata_t const* fseMetadata)
{
if (fseMetadata->llType == set_compressed || fseMetadata->llType == set_rle)
return 1;
if (fseMetadata->mlType == set_compressed || fseMetadata->mlType == set_rle)
return 1;
if (fseMetadata->ofType == set_compressed || fseMetadata->ofType == set_rle)
return 1;
return 0;
}
/** ZSTD_compressSubBlock_multi() :
* Breaks super-block into multiple sub-blocks and compresses them.
* Entropy will be written to the first block.
* The following blocks will use repeat mode to compress.
* All sub-blocks are compressed blocks (no raw or rle blocks).
* @return : compressed size of the super block (which is multiple ZSTD blocks)
* Or 0 if it failed to compress. */
static size_t ZSTD_compressSubBlock_multi(const seqStore_t* seqStorePtr,
const ZSTD_compressedBlockState_t* prevCBlock,
ZSTD_compressedBlockState_t* nextCBlock,
const ZSTD_entropyCTablesMetadata_t* entropyMetadata,
const ZSTD_CCtx_params* cctxParams,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
const int bmi2, U32 lastBlock,
void* workspace, size_t wkspSize)
{
const seqDef* const sstart = seqStorePtr->sequencesStart;
const seqDef* const send = seqStorePtr->sequences;
const seqDef* sp = sstart;
const BYTE* const lstart = seqStorePtr->litStart;
const BYTE* const lend = seqStorePtr->lit;
const BYTE* lp = lstart;
BYTE const* ip = (BYTE const*)src;
BYTE const* const iend = ip + srcSize;
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstCapacity;
BYTE* op = ostart;
const BYTE* llCodePtr = seqStorePtr->llCode;
const BYTE* mlCodePtr = seqStorePtr->mlCode;
const BYTE* ofCodePtr = seqStorePtr->ofCode;
size_t targetCBlockSize = cctxParams->targetCBlockSize;
size_t litSize, seqCount;
int writeLitEntropy = entropyMetadata->hufMetadata.hType == set_compressed;
int writeSeqEntropy = 1;
int lastSequence = 0;
DEBUGLOG(5, "ZSTD_compressSubBlock_multi (litSize=%u, nbSeq=%u)",
(unsigned)(lend-lp), (unsigned)(send-sstart));
litSize = 0;
seqCount = 0;
do {
size_t cBlockSizeEstimate = 0;
if (sstart == send) {
lastSequence = 1;
} else {
const seqDef* const sequence = sp + seqCount;
lastSequence = sequence == send - 1;
litSize += ZSTD_getSequenceLength(seqStorePtr, sequence).litLength;
seqCount++;
}
if (lastSequence) {
assert(lp <= lend);
assert(litSize <= (size_t)(lend - lp));
litSize = (size_t)(lend - lp);
}
/* I think there is an optimization opportunity here.
* Calling ZSTD_estimateSubBlockSize for every sequence can be wasteful
* since it recalculates estimate from scratch.
* For example, it would recount literal distribution and symbol codes everytime.
*/
cBlockSizeEstimate = ZSTD_estimateSubBlockSize(lp, litSize, ofCodePtr, llCodePtr, mlCodePtr, seqCount,
&nextCBlock->entropy, entropyMetadata,
workspace, wkspSize, writeLitEntropy, writeSeqEntropy);
if (cBlockSizeEstimate > targetCBlockSize || lastSequence) {
int litEntropyWritten = 0;
int seqEntropyWritten = 0;
const size_t decompressedSize = ZSTD_seqDecompressedSize(seqStorePtr, sp, seqCount, litSize, lastSequence);
const size_t cSize = ZSTD_compressSubBlock(&nextCBlock->entropy, entropyMetadata,
sp, seqCount,
lp, litSize,
llCodePtr, mlCodePtr, ofCodePtr,
cctxParams,
op, oend-op,
bmi2, writeLitEntropy, writeSeqEntropy,
&litEntropyWritten, &seqEntropyWritten,
lastBlock && lastSequence);
FORWARD_IF_ERROR(cSize, "ZSTD_compressSubBlock failed");
if (cSize > 0 && cSize < decompressedSize) {
DEBUGLOG(5, "Committed the sub-block");
assert(ip + decompressedSize <= iend);
ip += decompressedSize;
sp += seqCount;
lp += litSize;
op += cSize;
llCodePtr += seqCount;
mlCodePtr += seqCount;
ofCodePtr += seqCount;
litSize = 0;
seqCount = 0;
/* Entropy only needs to be written once */
if (litEntropyWritten) {
writeLitEntropy = 0;
}
if (seqEntropyWritten) {
writeSeqEntropy = 0;
}
}
}
} while (!lastSequence);
if (writeLitEntropy) {
DEBUGLOG(5, "ZSTD_compressSubBlock_multi has literal entropy tables unwritten");
memcpy(&nextCBlock->entropy.huf, &prevCBlock->entropy.huf, sizeof(prevCBlock->entropy.huf));
}
if (writeSeqEntropy && ZSTD_needSequenceEntropyTables(&entropyMetadata->fseMetadata)) {
/* If we haven't written our entropy tables, then we've violated our contract and
* must emit an uncompressed block.
*/
DEBUGLOG(5, "ZSTD_compressSubBlock_multi has sequence entropy tables unwritten");
return 0;
}
if (ip < iend) {
size_t const cSize = ZSTD_noCompressBlock(op, oend - op, ip, iend - ip, lastBlock);
DEBUGLOG(5, "ZSTD_compressSubBlock_multi last sub-block uncompressed, %zu bytes", (size_t)(iend - ip));
FORWARD_IF_ERROR(cSize, "ZSTD_noCompressBlock failed");
assert(cSize != 0);
op += cSize;
/* We have to regenerate the repcodes because we've skipped some sequences */
if (sp < send) {
seqDef const* seq;
repcodes_t rep;
memcpy(&rep, prevCBlock->rep, sizeof(rep));
for (seq = sstart; seq < sp; ++seq) {
rep = ZSTD_updateRep(rep.rep, seq->offset - 1, ZSTD_getSequenceLength(seqStorePtr, seq).litLength == 0);
}
memcpy(nextCBlock->rep, &rep, sizeof(rep));
}
}
DEBUGLOG(5, "ZSTD_compressSubBlock_multi compressed");
return op-ostart;
}
size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
void* dst, size_t dstCapacity,
void const* src, size_t srcSize,
unsigned lastBlock) {
ZSTD_entropyCTablesMetadata_t entropyMetadata;
FORWARD_IF_ERROR(ZSTD_buildSuperBlockEntropy(&zc->seqStore,
&zc->blockState.prevCBlock->entropy,
&zc->blockState.nextCBlock->entropy,
&zc->appliedParams,
&entropyMetadata,
zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */), "");
return ZSTD_compressSubBlock_multi(&zc->seqStore,
zc->blockState.prevCBlock,
zc->blockState.nextCBlock,
&entropyMetadata,
&zc->appliedParams,
dst, dstCapacity,
src, srcSize,
zc->bmi2, lastBlock,
zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */);
}
module/zstd/lib/compress/zstd_compress_superblock.h
+32
View File
@@ -0,0 +1,32 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_COMPRESS_ADVANCED_H
#define ZSTD_COMPRESS_ADVANCED_H
/*-*************************************
* Dependencies
***************************************/
#include "../zstd.h" /* ZSTD_CCtx */
/*-*************************************
* Target Compressed Block Size
***************************************/
/* ZSTD_compressSuperBlock() :
* Used to compress a super block when targetCBlockSize is being used.
* The given block will be compressed into multiple sub blocks that are around targetCBlockSize. */
size_t ZSTD_compressSuperBlock(ZSTD_CCtx* zc,
void* dst, size_t dstCapacity,
void const* src, size_t srcSize,
unsigned lastBlock);
#endif /* ZSTD_COMPRESS_ADVANCED_H */
module/zstd/lib/compress/zstd_cwksp.h
+525
View File
@@ -0,0 +1,525 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_CWKSP_H
#define ZSTD_CWKSP_H
/*-*************************************
* Dependencies
***************************************/
#include "../common/zstd_internal.h"
#if defined (__cplusplus)
extern "C" {
#endif
/*-*************************************
* Constants
***************************************/
/* Since the workspace is effectively its own little malloc implementation /
* arena, when we run under ASAN, we should similarly insert redzones between
* each internal element of the workspace, so ASAN will catch overruns that
* reach outside an object but that stay inside the workspace.
*
* This defines the size of that redzone.
*/
#ifndef ZSTD_CWKSP_ASAN_REDZONE_SIZE
#define ZSTD_CWKSP_ASAN_REDZONE_SIZE 128
#endif
/*-*************************************
* Structures
***************************************/
typedef enum {
ZSTD_cwksp_alloc_objects,
ZSTD_cwksp_alloc_buffers,
ZSTD_cwksp_alloc_aligned
} ZSTD_cwksp_alloc_phase_e;
/**
* Zstd fits all its internal datastructures into a single continuous buffer,
* so that it only needs to perform a single OS allocation (or so that a buffer
* can be provided to it and it can perform no allocations at all). This buffer
* is called the workspace.
*
* Several optimizations complicate that process of allocating memory ranges
* from this workspace for each internal datastructure:
*
* - These different internal datastructures have different setup requirements:
*
* - The static objects need to be cleared once and can then be trivially
* reused for each compression.
*
* - Various buffers don't need to be initialized at all--they are always
* written into before they're read.
*
* - The matchstate tables have a unique requirement that they don't need
* their memory to be totally cleared, but they do need the memory to have
* some bound, i.e., a guarantee that all values in the memory they've been
* allocated is less than some maximum value (which is the starting value
* for the indices that they will then use for compression). When this
* guarantee is provided to them, they can use the memory without any setup
* work. When it can't, they have to clear the area.
*
* - These buffers also have different alignment requirements.
*
* - We would like to reuse the objects in the workspace for multiple
* compressions without having to perform any expensive reallocation or
* reinitialization work.
*
* - We would like to be able to efficiently reuse the workspace across
* multiple compressions **even when the compression parameters change** and
* we need to resize some of the objects (where possible).
*
* To attempt to manage this buffer, given these constraints, the ZSTD_cwksp
* abstraction was created. It works as follows:
*
* Workspace Layout:
*
* [ ... workspace ... ]
* [objects][tables ... ->] free space [<- ... aligned][<- ... buffers]
*
* The various objects that live in the workspace are divided into the
* following categories, and are allocated separately:
*
* - Static objects: this is optionally the enclosing ZSTD_CCtx or ZSTD_CDict,
* so that literally everything fits in a single buffer. Note: if present,
* this must be the first object in the workspace, since ZSTD_free{CCtx,
* CDict}() rely on a pointer comparison to see whether one or two frees are
* required.
*
* - Fixed size objects: these are fixed-size, fixed-count objects that are
* nonetheless "dynamically" allocated in the workspace so that we can
* control how they're initialized separately from the broader ZSTD_CCtx.
* Examples:
* - Entropy Workspace
* - 2 x ZSTD_compressedBlockState_t
* - CDict dictionary contents
*
* - Tables: these are any of several different datastructures (hash tables,
* chain tables, binary trees) that all respect a common format: they are
* uint32_t arrays, all of whose values are between 0 and (nextSrc - base).
* Their sizes depend on the cparams.
*
* - Aligned: these buffers are used for various purposes that require 4 byte
* alignment, but don't require any initialization before they're used.
*
* - Buffers: these buffers are used for various purposes that don't require
* any alignment or initialization before they're used. This means they can
* be moved around at no cost for a new compression.
*
* Allocating Memory:
*
* The various types of objects must be allocated in order, so they can be
* correctly packed into the workspace buffer. That order is:
*
* 1. Objects
* 2. Buffers
* 3. Aligned
* 4. Tables
*
* Attempts to reserve objects of different types out of order will fail.
*/
typedef struct {
void* workspace;
void* workspaceEnd;
void* objectEnd;
void* tableEnd;
void* tableValidEnd;
void* allocStart;
int allocFailed;
int workspaceOversizedDuration;
ZSTD_cwksp_alloc_phase_e phase;
} ZSTD_cwksp;
/*-*************************************
* Functions
***************************************/
MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws);
MEM_STATIC void ZSTD_cwksp_assert_internal_consistency(ZSTD_cwksp* ws) {
(void)ws;
assert(ws->workspace <= ws->objectEnd);
assert(ws->objectEnd <= ws->tableEnd);
assert(ws->objectEnd <= ws->tableValidEnd);
assert(ws->tableEnd <= ws->allocStart);
assert(ws->tableValidEnd <= ws->allocStart);
assert(ws->allocStart <= ws->workspaceEnd);
}
/**
* Align must be a power of 2.
*/
MEM_STATIC size_t ZSTD_cwksp_align(size_t size, size_t const align) {
size_t const mask = align - 1;
assert((align & mask) == 0);
return (size + mask) & ~mask;
}
/**
* Use this to determine how much space in the workspace we will consume to
* allocate this object. (Normally it should be exactly the size of the object,
* but under special conditions, like ASAN, where we pad each object, it might
* be larger.)
*
* Since tables aren't currently redzoned, you don't need to call through this
* to figure out how much space you need for the matchState tables. Everything
* else is though.
*/
MEM_STATIC size_t ZSTD_cwksp_alloc_size(size_t size) {
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
return size + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
#else
return size;
#endif
}
MEM_STATIC void ZSTD_cwksp_internal_advance_phase(
ZSTD_cwksp* ws, ZSTD_cwksp_alloc_phase_e phase) {
assert(phase >= ws->phase);
if (phase > ws->phase) {
if (ws->phase < ZSTD_cwksp_alloc_buffers &&
phase >= ZSTD_cwksp_alloc_buffers) {
ws->tableValidEnd = ws->objectEnd;
}
if (ws->phase < ZSTD_cwksp_alloc_aligned &&
phase >= ZSTD_cwksp_alloc_aligned) {
/* If unaligned allocations down from a too-large top have left us
* unaligned, we need to realign our alloc ptr. Technically, this
* can consume space that is unaccounted for in the neededSpace
* calculation. However, I believe this can only happen when the
* workspace is too large, and specifically when it is too large
* by a larger margin than the space that will be consumed. */
/* TODO: cleaner, compiler warning friendly way to do this??? */
ws->allocStart = (BYTE*)ws->allocStart - ((size_t)ws->allocStart & (sizeof(U32)-1));
if (ws->allocStart < ws->tableValidEnd) {
ws->tableValidEnd = ws->allocStart;
}
}
ws->phase = phase;
}
}
/**
* Returns whether this object/buffer/etc was allocated in this workspace.
*/
MEM_STATIC int ZSTD_cwksp_owns_buffer(const ZSTD_cwksp* ws, const void* ptr) {
return (ptr != NULL) && (ws->workspace <= ptr) && (ptr <= ws->workspaceEnd);
}
/**
* Internal function. Do not use directly.
*/
MEM_STATIC void* ZSTD_cwksp_reserve_internal(
ZSTD_cwksp* ws, size_t bytes, ZSTD_cwksp_alloc_phase_e phase) {
void* alloc;
void* bottom = ws->tableEnd;
ZSTD_cwksp_internal_advance_phase(ws, phase);
alloc = (BYTE *)ws->allocStart - bytes;
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
/* over-reserve space */
alloc = (BYTE *)alloc - 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
#endif
DEBUGLOG(5, "cwksp: reserving %p %zd bytes, %zd bytes remaining",
alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
ZSTD_cwksp_assert_internal_consistency(ws);
assert(alloc >= bottom);
if (alloc < bottom) {
DEBUGLOG(4, "cwksp: alloc failed!");
ws->allocFailed = 1;
return NULL;
}
if (alloc < ws->tableValidEnd) {
ws->tableValidEnd = alloc;
}
ws->allocStart = alloc;
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
/* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
* either size. */
alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
__asan_unpoison_memory_region(alloc, bytes);
#endif
return alloc;
}
/**
* Reserves and returns unaligned memory.
*/
MEM_STATIC BYTE* ZSTD_cwksp_reserve_buffer(ZSTD_cwksp* ws, size_t bytes) {
return (BYTE*)ZSTD_cwksp_reserve_internal(ws, bytes, ZSTD_cwksp_alloc_buffers);
}
/**
* Reserves and returns memory sized on and aligned on sizeof(unsigned).
*/
MEM_STATIC void* ZSTD_cwksp_reserve_aligned(ZSTD_cwksp* ws, size_t bytes) {
assert((bytes & (sizeof(U32)-1)) == 0);
return ZSTD_cwksp_reserve_internal(ws, ZSTD_cwksp_align(bytes, sizeof(U32)), ZSTD_cwksp_alloc_aligned);
}
/**
* Aligned on sizeof(unsigned). These buffers have the special property that
* their values remain constrained, allowing us to re-use them without
* memset()-ing them.
*/
MEM_STATIC void* ZSTD_cwksp_reserve_table(ZSTD_cwksp* ws, size_t bytes) {
const ZSTD_cwksp_alloc_phase_e phase = ZSTD_cwksp_alloc_aligned;
void* alloc = ws->tableEnd;
void* end = (BYTE *)alloc + bytes;
void* top = ws->allocStart;
DEBUGLOG(5, "cwksp: reserving %p table %zd bytes, %zd bytes remaining",
alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
assert((bytes & (sizeof(U32)-1)) == 0);
ZSTD_cwksp_internal_advance_phase(ws, phase);
ZSTD_cwksp_assert_internal_consistency(ws);
assert(end <= top);
if (end > top) {
DEBUGLOG(4, "cwksp: table alloc failed!");
ws->allocFailed = 1;
return NULL;
}
ws->tableEnd = end;
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
__asan_unpoison_memory_region(alloc, bytes);
#endif
return alloc;
}
/**
* Aligned on sizeof(void*).
*/
MEM_STATIC void* ZSTD_cwksp_reserve_object(ZSTD_cwksp* ws, size_t bytes) {
size_t roundedBytes = ZSTD_cwksp_align(bytes, sizeof(void*));
void* alloc = ws->objectEnd;
void* end = (BYTE*)alloc + roundedBytes;
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
/* over-reserve space */
end = (BYTE *)end + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
#endif
DEBUGLOG(5,
"cwksp: reserving %p object %zd bytes (rounded to %zd), %zd bytes remaining",
alloc, bytes, roundedBytes, ZSTD_cwksp_available_space(ws) - roundedBytes);
assert(((size_t)alloc & (sizeof(void*)-1)) == 0);
assert((bytes & (sizeof(void*)-1)) == 0);
ZSTD_cwksp_assert_internal_consistency(ws);
/* we must be in the first phase, no advance is possible */
if (ws->phase != ZSTD_cwksp_alloc_objects || end > ws->workspaceEnd) {
DEBUGLOG(4, "cwksp: object alloc failed!");
ws->allocFailed = 1;
return NULL;
}
ws->objectEnd = end;
ws->tableEnd = end;
ws->tableValidEnd = end;
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
/* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
* either size. */
alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
__asan_unpoison_memory_region(alloc, bytes);
#endif
return alloc;
}
MEM_STATIC void ZSTD_cwksp_mark_tables_dirty(ZSTD_cwksp* ws) {
DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_dirty");
#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
/* To validate that the table re-use logic is sound, and that we don't
* access table space that we haven't cleaned, we re-"poison" the table
* space every time we mark it dirty. */
{
size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
assert(__msan_test_shadow(ws->objectEnd, size) == -1);
__msan_poison(ws->objectEnd, size);
}
#endif
assert(ws->tableValidEnd >= ws->objectEnd);
assert(ws->tableValidEnd <= ws->allocStart);
ws->tableValidEnd = ws->objectEnd;
ZSTD_cwksp_assert_internal_consistency(ws);
}
MEM_STATIC void ZSTD_cwksp_mark_tables_clean(ZSTD_cwksp* ws) {
DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_clean");
assert(ws->tableValidEnd >= ws->objectEnd);
assert(ws->tableValidEnd <= ws->allocStart);
if (ws->tableValidEnd < ws->tableEnd) {
ws->tableValidEnd = ws->tableEnd;
}
ZSTD_cwksp_assert_internal_consistency(ws);
}
/**
* Zero the part of the allocated tables not already marked clean.
*/
MEM_STATIC void ZSTD_cwksp_clean_tables(ZSTD_cwksp* ws) {
DEBUGLOG(4, "cwksp: ZSTD_cwksp_clean_tables");
assert(ws->tableValidEnd >= ws->objectEnd);
assert(ws->tableValidEnd <= ws->allocStart);
if (ws->tableValidEnd < ws->tableEnd) {
memset(ws->tableValidEnd, 0, (BYTE*)ws->tableEnd - (BYTE*)ws->tableValidEnd);
}
ZSTD_cwksp_mark_tables_clean(ws);
}
/**
* Invalidates table allocations.
* All other allocations remain valid.
*/
MEM_STATIC void ZSTD_cwksp_clear_tables(ZSTD_cwksp* ws) {
DEBUGLOG(4, "cwksp: clearing tables!");
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
{
size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
__asan_poison_memory_region(ws->objectEnd, size);
}
#endif
ws->tableEnd = ws->objectEnd;
ZSTD_cwksp_assert_internal_consistency(ws);
}
/**
* Invalidates all buffer, aligned, and table allocations.
* Object allocations remain valid.
*/
MEM_STATIC void ZSTD_cwksp_clear(ZSTD_cwksp* ws) {
DEBUGLOG(4, "cwksp: clearing!");
#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
/* To validate that the context re-use logic is sound, and that we don't
* access stuff that this compression hasn't initialized, we re-"poison"
* the workspace (or at least the non-static, non-table parts of it)
* every time we start a new compression. */
{
size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->tableValidEnd;
__msan_poison(ws->tableValidEnd, size);
}
#endif
#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
{
size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->objectEnd;
__asan_poison_memory_region(ws->objectEnd, size);
}
#endif
ws->tableEnd = ws->objectEnd;
ws->allocStart = ws->workspaceEnd;
ws->allocFailed = 0;
if (ws->phase > ZSTD_cwksp_alloc_buffers) {
ws->phase = ZSTD_cwksp_alloc_buffers;
}
ZSTD_cwksp_assert_internal_consistency(ws);
}
/**
* The provided workspace takes ownership of the buffer [start, start+size).
* Any existing values in the workspace are ignored (the previously managed
* buffer, if present, must be separately freed).
*/
MEM_STATIC void ZSTD_cwksp_init(ZSTD_cwksp* ws, void* start, size_t size) {
DEBUGLOG(4, "cwksp: init'ing workspace with %zd bytes", size);
assert(((size_t)start & (sizeof(void*)-1)) == 0); /* ensure correct alignment */
ws->workspace = start;
ws->workspaceEnd = (BYTE*)start + size;
ws->objectEnd = ws->workspace;
ws->tableValidEnd = ws->objectEnd;
ws->phase = ZSTD_cwksp_alloc_objects;
ZSTD_cwksp_clear(ws);
ws->workspaceOversizedDuration = 0;
ZSTD_cwksp_assert_internal_consistency(ws);
}
MEM_STATIC size_t ZSTD_cwksp_create(ZSTD_cwksp* ws, size_t size, ZSTD_customMem customMem) {
void* workspace = ZSTD_malloc(size, customMem);
DEBUGLOG(4, "cwksp: creating new workspace with %zd bytes", size);
RETURN_ERROR_IF(workspace == NULL, memory_allocation, "NULL pointer!");
ZSTD_cwksp_init(ws, workspace, size);
return 0;
}
MEM_STATIC void ZSTD_cwksp_free(ZSTD_cwksp* ws, ZSTD_customMem customMem) {
void *ptr = ws->workspace;
DEBUGLOG(4, "cwksp: freeing workspace");
memset(ws, 0, sizeof(ZSTD_cwksp));
ZSTD_free(ptr, customMem);
}
/**
* Moves the management of a workspace from one cwksp to another. The src cwksp
* is left in an invalid state (src must be re-init()'ed before its used again).
*/
MEM_STATIC void ZSTD_cwksp_move(ZSTD_cwksp* dst, ZSTD_cwksp* src) {
*dst = *src;
memset(src, 0, sizeof(ZSTD_cwksp));
}
MEM_STATIC size_t ZSTD_cwksp_sizeof(const ZSTD_cwksp* ws) {
return (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->workspace);
}
MEM_STATIC int ZSTD_cwksp_reserve_failed(const ZSTD_cwksp* ws) {
return ws->allocFailed;
}
/*-*************************************
* Functions Checking Free Space
***************************************/
MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws) {
return (size_t)((BYTE*)ws->allocStart - (BYTE*)ws->tableEnd);
}
MEM_STATIC int ZSTD_cwksp_check_available(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
return ZSTD_cwksp_available_space(ws) >= additionalNeededSpace;
}
MEM_STATIC int ZSTD_cwksp_check_too_large(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
return ZSTD_cwksp_check_available(
ws, additionalNeededSpace * ZSTD_WORKSPACETOOLARGE_FACTOR);
}
MEM_STATIC int ZSTD_cwksp_check_wasteful(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
return ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)
&& ws->workspaceOversizedDuration > ZSTD_WORKSPACETOOLARGE_MAXDURATION;
}
MEM_STATIC void ZSTD_cwksp_bump_oversized_duration(
ZSTD_cwksp* ws, size_t additionalNeededSpace) {
if (ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)) {
ws->workspaceOversizedDuration++;
} else {
ws->workspaceOversizedDuration = 0;
}
}
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_CWKSP_H */
module/zstd/lib/compress/zstd_double_fast.c
+521
View File
@@ -0,0 +1,521 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#include "zstd_compress_internal.h"
#include "zstd_double_fast.h"
void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
void const* end, ZSTD_dictTableLoadMethod_e dtlm)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashLarge = ms->hashTable;
U32 const hBitsL = cParams->hashLog;
U32 const mls = cParams->minMatch;
U32* const hashSmall = ms->chainTable;
U32 const hBitsS = cParams->chainLog;
const BYTE* const base = ms->window.base;
const BYTE* ip = base + ms->nextToUpdate;
const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
const U32 fastHashFillStep = 3;
/* Always insert every fastHashFillStep position into the hash tables.
* Insert the other positions into the large hash table if their entry
* is empty.
*/
for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
U32 const current = (U32)(ip - base);
U32 i;
for (i = 0; i < fastHashFillStep; ++i) {
size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
if (i == 0)
hashSmall[smHash] = current + i;
if (i == 0 || hashLarge[lgHash] == 0)
hashLarge[lgHash] = current + i;
/* Only load extra positions for ZSTD_dtlm_full */
if (dtlm == ZSTD_dtlm_fast)
break;
} }
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_doubleFast_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const mls /* template */, ZSTD_dictMode_e const dictMode)
{
ZSTD_compressionParameters const* cParams = &ms->cParams;
U32* const hashLong = ms->hashTable;
const U32 hBitsL = cParams->hashLog;
U32* const hashSmall = ms->chainTable;
const U32 hBitsS = cParams->chainLog;
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
/* presumes that, if there is a dictionary, it must be using Attach mode */
const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
const BYTE* const prefixLowest = base + prefixLowestIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - HASH_READ_SIZE;
U32 offset_1=rep[0], offset_2=rep[1];
U32 offsetSaved = 0;
const ZSTD_matchState_t* const dms = ms->dictMatchState;
const ZSTD_compressionParameters* const dictCParams =
dictMode == ZSTD_dictMatchState ?
&dms->cParams : NULL;
const U32* const dictHashLong = dictMode == ZSTD_dictMatchState ?
dms->hashTable : NULL;
const U32* const dictHashSmall = dictMode == ZSTD_dictMatchState ?
dms->chainTable : NULL;
const U32 dictStartIndex = dictMode == ZSTD_dictMatchState ?
dms->window.dictLimit : 0;
const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
dms->window.base : NULL;
const BYTE* const dictStart = dictMode == ZSTD_dictMatchState ?
dictBase + dictStartIndex : NULL;
const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
dms->window.nextSrc : NULL;
const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
prefixLowestIndex - (U32)(dictEnd - dictBase) :
0;
const U32 dictHBitsL = dictMode == ZSTD_dictMatchState ?
dictCParams->hashLog : hBitsL;
const U32 dictHBitsS = dictMode == ZSTD_dictMatchState ?
dictCParams->chainLog : hBitsS;
const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictStart));
DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_generic");
assert(dictMode == ZSTD_noDict || dictMode == ZSTD_dictMatchState);
/* if a dictionary is attached, it must be within window range */
if (dictMode == ZSTD_dictMatchState) {
assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
}
/* init */
ip += (dictAndPrefixLength == 0);
if (dictMode == ZSTD_noDict) {
U32 const current = (U32)(ip - base);
U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
U32 const maxRep = current - windowLow;
if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
}
if (dictMode == ZSTD_dictMatchState) {
/* dictMatchState repCode checks don't currently handle repCode == 0
* disabling. */
assert(offset_1 <= dictAndPrefixLength);
assert(offset_2 <= dictAndPrefixLength);
}
/* Main Search Loop */
while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
size_t mLength;
U32 offset;
size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8);
size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls);
U32 const current = (U32)(ip-base);
U32 const matchIndexL = hashLong[h2];
U32 matchIndexS = hashSmall[h];
const BYTE* matchLong = base + matchIndexL;
const BYTE* match = base + matchIndexS;
const U32 repIndex = current + 1 - offset_1;
const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
&& repIndex < prefixLowestIndex) ?
dictBase + (repIndex - dictIndexDelta) :
base + repIndex;
hashLong[h2] = hashSmall[h] = current; /* update hash tables */
/* check dictMatchState repcode */
if (dictMode == ZSTD_dictMatchState
&& ((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
ip++;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
goto _match_stored;
}
/* check noDict repcode */
if ( dictMode == ZSTD_noDict
&& ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
ip++;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
goto _match_stored;
}
if (matchIndexL > prefixLowestIndex) {
/* check prefix long match */
if (MEM_read64(matchLong) == MEM_read64(ip)) {
mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
offset = (U32)(ip-matchLong);
while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
goto _match_found;
}
} else if (dictMode == ZSTD_dictMatchState) {
/* check dictMatchState long match */
U32 const dictMatchIndexL = dictHashLong[dictHL];
const BYTE* dictMatchL = dictBase + dictMatchIndexL;
assert(dictMatchL < dictEnd);
if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) {
mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8;
offset = (U32)(current - dictMatchIndexL - dictIndexDelta);
while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */
goto _match_found;
} }
if (matchIndexS > prefixLowestIndex) {
/* check prefix short match */
if (MEM_read32(match) == MEM_read32(ip)) {
goto _search_next_long;
}
} else if (dictMode == ZSTD_dictMatchState) {
/* check dictMatchState short match */
U32 const dictMatchIndexS = dictHashSmall[dictHS];
match = dictBase + dictMatchIndexS;
matchIndexS = dictMatchIndexS + dictIndexDelta;
if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) {
goto _search_next_long;
} }
ip += ((ip-anchor) >> kSearchStrength) + 1;
#if defined(__aarch64__)
PREFETCH_L1(ip+256);
#endif
continue;
_search_next_long:
{ size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8);
U32 const matchIndexL3 = hashLong[hl3];
const BYTE* matchL3 = base + matchIndexL3;
hashLong[hl3] = current + 1;
/* check prefix long +1 match */
if (matchIndexL3 > prefixLowestIndex) {
if (MEM_read64(matchL3) == MEM_read64(ip+1)) {
mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
ip++;
offset = (U32)(ip-matchL3);
while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
goto _match_found;
}
} else if (dictMode == ZSTD_dictMatchState) {
/* check dict long +1 match */
U32 const dictMatchIndexL3 = dictHashLong[dictHLNext];
const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3;
assert(dictMatchL3 < dictEnd);
if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) {
mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8;
ip++;
offset = (U32)(current + 1 - dictMatchIndexL3 - dictIndexDelta);
while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */
goto _match_found;
} } }
/* if no long +1 match, explore the short match we found */
if (dictMode == ZSTD_dictMatchState && matchIndexS < prefixLowestIndex) {
mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4;
offset = (U32)(current - matchIndexS);
while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
} else {
mLength = ZSTD_count(ip+4, match+4, iend) + 4;
offset = (U32)(ip - match);
while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
}
/* fall-through */
_match_found:
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
_match_stored:
/* match found */
ip += mLength;
anchor = ip;
if (ip <= ilimit) {
/* Complementary insertion */
/* done after iLimit test, as candidates could be > iend-8 */
{ U32 const indexToInsert = current+2;
hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
}
/* check immediate repcode */
if (dictMode == ZSTD_dictMatchState) {
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex2 = current2 - offset_2;
const BYTE* repMatch2 = dictMode == ZSTD_dictMatchState
&& repIndex2 < prefixLowestIndex ?
dictBase + repIndex2 - dictIndexDelta :
base + repIndex2;
if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
&& (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4;
U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
ip += repLength2;
anchor = ip;
continue;
}
break;
} }
if (dictMode == ZSTD_noDict) {
while ( (ip <= ilimit)
&& ( (offset_2>0)
& (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
/* store sequence */
size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, rLength-MINMATCH);
ip += rLength;
anchor = ip;
continue; /* faster when present ... (?) */
} } }
} /* while (ip < ilimit) */
/* save reps for next block */
rep[0] = offset_1 ? offset_1 : offsetSaved;
rep[1] = offset_2 ? offset_2 : offsetSaved;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_doubleFast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
const U32 mls = ms->cParams.minMatch;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_noDict);
case 5 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_noDict);
case 6 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_noDict);
case 7 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_noDict);
}
}
size_t ZSTD_compressBlock_doubleFast_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
const U32 mls = ms->cParams.minMatch;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_dictMatchState);
case 5 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_dictMatchState);
case 6 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_dictMatchState);
case 7 :
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_dictMatchState);
}
}
static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const mls /* template */)
{
ZSTD_compressionParameters const* cParams = &ms->cParams;
U32* const hashLong = ms->hashTable;
U32 const hBitsL = cParams->hashLog;
U32* const hashSmall = ms->chainTable;
U32 const hBitsS = cParams->chainLog;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
const BYTE* const base = ms->window.base;
const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
const U32 dictStartIndex = lowLimit;
const U32 dictLimit = ms->window.dictLimit;
const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit;
const BYTE* const prefixStart = base + prefixStartIndex;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const dictStart = dictBase + dictStartIndex;
const BYTE* const dictEnd = dictBase + prefixStartIndex;
U32 offset_1=rep[0], offset_2=rep[1];
DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize);
/* if extDict is invalidated due to maxDistance, switch to "regular" variant */
if (prefixStartIndex == dictStartIndex)
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, mls, ZSTD_noDict);
/* Search Loop */
while (ip < ilimit) { /* < instead of <=, because (ip+1) */
const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
const U32 matchIndex = hashSmall[hSmall];
const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
const BYTE* match = matchBase + matchIndex;
const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
const U32 matchLongIndex = hashLong[hLong];
const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base;
const BYTE* matchLong = matchLongBase + matchLongIndex;
const U32 current = (U32)(ip-base);
const U32 repIndex = current + 1 - offset_1; /* offset_1 expected <= current +1 */
const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
size_t mLength;
hashSmall[hSmall] = hashLong[hLong] = current; /* update hash table */
if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */
& (repIndex > dictStartIndex))
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
ip++;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
} else {
if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend;
const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart;
U32 offset;
mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8;
offset = current - matchLongIndex;
while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) {
size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
U32 const matchIndex3 = hashLong[h3];
const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base;
const BYTE* match3 = match3Base + matchIndex3;
U32 offset;
hashLong[h3] = current + 1;
if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend;
const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart;
mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8;
ip++;
offset = current+1 - matchIndex3;
while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
} else {
const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
offset = current - matchIndex;
while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
}
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} else {
ip += ((ip-anchor) >> kSearchStrength) + 1;
continue;
} }
/* move to next sequence start */
ip += mLength;
anchor = ip;
if (ip <= ilimit) {
/* Complementary insertion */
/* done after iLimit test, as candidates could be > iend-8 */
{ U32 const indexToInsert = current+2;
hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
}
/* check immediate repcode */
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex2 = current2 - offset_2;
const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */
& (repIndex2 > dictStartIndex))
&& (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
ip += repLength2;
anchor = ip;
continue;
}
break;
} } }
/* save reps for next block */
rep[0] = offset_1;
rep[1] = offset_2;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_doubleFast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
U32 const mls = ms->cParams.minMatch;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 7);
}
}
module/zstd/lib/compress/zstd_double_fast.h
+38
View File
@@ -0,0 +1,38 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_DOUBLE_FAST_H
#define ZSTD_DOUBLE_FAST_H
#if defined (__cplusplus)
extern "C" {
#endif
#include "../common/mem.h" /* U32 */
#include "zstd_compress_internal.h" /* ZSTD_CCtx, size_t */
void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
void const* end, ZSTD_dictTableLoadMethod_e dtlm);
size_t ZSTD_compressBlock_doubleFast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_doubleFast_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_doubleFast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_DOUBLE_FAST_H */
module/zstd/lib/compress/zstd_fast.c
+496
View File
@@ -0,0 +1,496 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#include "zstd_compress_internal.h" /* ZSTD_hashPtr, ZSTD_count, ZSTD_storeSeq */
#include "zstd_fast.h"
void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
const void* const end,
ZSTD_dictTableLoadMethod_e dtlm)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hBits = cParams->hashLog;
U32 const mls = cParams->minMatch;
const BYTE* const base = ms->window.base;
const BYTE* ip = base + ms->nextToUpdate;
const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
const U32 fastHashFillStep = 3;
/* Always insert every fastHashFillStep position into the hash table.
* Insert the other positions if their hash entry is empty.
*/
for ( ; ip + fastHashFillStep < iend + 2; ip += fastHashFillStep) {
U32 const current = (U32)(ip - base);
size_t const hash0 = ZSTD_hashPtr(ip, hBits, mls);
hashTable[hash0] = current;
if (dtlm == ZSTD_dtlm_fast) continue;
/* Only load extra positions for ZSTD_dtlm_full */
{ U32 p;
for (p = 1; p < fastHashFillStep; ++p) {
size_t const hash = ZSTD_hashPtr(ip + p, hBits, mls);
if (hashTable[hash] == 0) { /* not yet filled */
hashTable[hash] = current + p;
} } } }
}
FORCE_INLINE_TEMPLATE size_t
ZSTD_compressBlock_fast_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const mls)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hlog = cParams->hashLog;
/* support stepSize of 0 */
size_t const stepSize = cParams->targetLength + !(cParams->targetLength) + 1;
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
/* We check ip0 (ip + 0) and ip1 (ip + 1) each loop */
const BYTE* ip0 = istart;
const BYTE* ip1;
const BYTE* anchor = istart;
const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
const U32 prefixStartIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
const BYTE* const prefixStart = base + prefixStartIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - HASH_READ_SIZE;
U32 offset_1=rep[0], offset_2=rep[1];
U32 offsetSaved = 0;
/* init */
DEBUGLOG(5, "ZSTD_compressBlock_fast_generic");
ip0 += (ip0 == prefixStart);
ip1 = ip0 + 1;
{ U32 const current = (U32)(ip0 - base);
U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
U32 const maxRep = current - windowLow;
if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
}
/* Main Search Loop */
#ifdef __INTEL_COMPILER
/* From intel 'The vector pragma indicates that the loop should be
* vectorized if it is legal to do so'. Can be used together with
* #pragma ivdep (but have opted to exclude that because intel
* warns against using it).*/
#pragma vector always
#endif
while (ip1 < ilimit) { /* < instead of <=, because check at ip0+2 */
size_t mLength;
BYTE const* ip2 = ip0 + 2;
size_t const h0 = ZSTD_hashPtr(ip0, hlog, mls);
U32 const val0 = MEM_read32(ip0);
size_t const h1 = ZSTD_hashPtr(ip1, hlog, mls);
U32 const val1 = MEM_read32(ip1);
U32 const current0 = (U32)(ip0-base);
U32 const current1 = (U32)(ip1-base);
U32 const matchIndex0 = hashTable[h0];
U32 const matchIndex1 = hashTable[h1];
BYTE const* repMatch = ip2 - offset_1;
const BYTE* match0 = base + matchIndex0;
const BYTE* match1 = base + matchIndex1;
U32 offcode;
#if defined(__aarch64__)
PREFETCH_L1(ip0+256);
#endif
hashTable[h0] = current0; /* update hash table */
hashTable[h1] = current1; /* update hash table */
assert(ip0 + 1 == ip1);
if ((offset_1 > 0) & (MEM_read32(repMatch) == MEM_read32(ip2))) {
mLength = (ip2[-1] == repMatch[-1]) ? 1 : 0;
ip0 = ip2 - mLength;
match0 = repMatch - mLength;
mLength += 4;
offcode = 0;
goto _match;
}
if ((matchIndex0 > prefixStartIndex) && MEM_read32(match0) == val0) {
/* found a regular match */
goto _offset;
}
if ((matchIndex1 > prefixStartIndex) && MEM_read32(match1) == val1) {
/* found a regular match after one literal */
ip0 = ip1;
match0 = match1;
goto _offset;
}
{ size_t const step = ((size_t)(ip0-anchor) >> (kSearchStrength - 1)) + stepSize;
assert(step >= 2);
ip0 += step;
ip1 += step;
continue;
}
_offset: /* Requires: ip0, match0 */
/* Compute the offset code */
offset_2 = offset_1;
offset_1 = (U32)(ip0-match0);
offcode = offset_1 + ZSTD_REP_MOVE;
mLength = 4;
/* Count the backwards match length */
while (((ip0>anchor) & (match0>prefixStart))
&& (ip0[-1] == match0[-1])) { ip0--; match0--; mLength++; } /* catch up */
_match: /* Requires: ip0, match0, offcode */
/* Count the forward length */
mLength += ZSTD_count(ip0+mLength, match0+mLength, iend);
ZSTD_storeSeq(seqStore, (size_t)(ip0-anchor), anchor, iend, offcode, mLength-MINMATCH);
/* match found */
ip0 += mLength;
anchor = ip0;
if (ip0 <= ilimit) {
/* Fill Table */
assert(base+current0+2 > istart); /* check base overflow */
hashTable[ZSTD_hashPtr(base+current0+2, hlog, mls)] = current0+2; /* here because current+2 could be > iend-8 */
hashTable[ZSTD_hashPtr(ip0-2, hlog, mls)] = (U32)(ip0-2-base);
if (offset_2 > 0) { /* offset_2==0 means offset_2 is invalidated */
while ( (ip0 <= ilimit) && (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) ) {
/* store sequence */
size_t const rLength = ZSTD_count(ip0+4, ip0+4-offset_2, iend) + 4;
{ U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
ip0 += rLength;
ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, 0 /*offCode*/, rLength-MINMATCH);
anchor = ip0;
continue; /* faster when present (confirmed on gcc-8) ... (?) */
} } }
ip1 = ip0 + 1;
}
/* save reps for next block */
rep[0] = offset_1 ? offset_1 : offsetSaved;
rep[1] = offset_2 ? offset_2 : offsetSaved;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_fast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
U32 const mls = ms->cParams.minMatch;
assert(ms->dictMatchState == NULL);
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, 7);
}
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_fast_dictMatchState_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize, U32 const mls)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hlog = cParams->hashLog;
/* support stepSize of 0 */
U32 const stepSize = cParams->targetLength + !(cParams->targetLength);
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 prefixStartIndex = ms->window.dictLimit;
const BYTE* const prefixStart = base + prefixStartIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - HASH_READ_SIZE;
U32 offset_1=rep[0], offset_2=rep[1];
U32 offsetSaved = 0;
const ZSTD_matchState_t* const dms = ms->dictMatchState;
const ZSTD_compressionParameters* const dictCParams = &dms->cParams ;
const U32* const dictHashTable = dms->hashTable;
const U32 dictStartIndex = dms->window.dictLimit;
const BYTE* const dictBase = dms->window.base;
const BYTE* const dictStart = dictBase + dictStartIndex;
const BYTE* const dictEnd = dms->window.nextSrc;
const U32 dictIndexDelta = prefixStartIndex - (U32)(dictEnd - dictBase);
const U32 dictAndPrefixLength = (U32)(ip - prefixStart + dictEnd - dictStart);
const U32 dictHLog = dictCParams->hashLog;
/* if a dictionary is still attached, it necessarily means that
* it is within window size. So we just check it. */
const U32 maxDistance = 1U << cParams->windowLog;
const U32 endIndex = (U32)((size_t)(ip - base) + srcSize);
assert(endIndex - prefixStartIndex <= maxDistance);
(void)maxDistance; (void)endIndex; /* these variables are not used when assert() is disabled */
/* ensure there will be no no underflow
* when translating a dict index into a local index */
assert(prefixStartIndex >= (U32)(dictEnd - dictBase));
/* init */
DEBUGLOG(5, "ZSTD_compressBlock_fast_dictMatchState_generic");
ip += (dictAndPrefixLength == 0);
/* dictMatchState repCode checks don't currently handle repCode == 0
* disabling. */
assert(offset_1 <= dictAndPrefixLength);
assert(offset_2 <= dictAndPrefixLength);
/* Main Search Loop */
while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
size_t mLength;
size_t const h = ZSTD_hashPtr(ip, hlog, mls);
U32 const current = (U32)(ip-base);
U32 const matchIndex = hashTable[h];
const BYTE* match = base + matchIndex;
const U32 repIndex = current + 1 - offset_1;
const BYTE* repMatch = (repIndex < prefixStartIndex) ?
dictBase + (repIndex - dictIndexDelta) :
base + repIndex;
hashTable[h] = current; /* update hash table */
if ( ((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex isn't overlapping dict + prefix */
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
ip++;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
} else if ( (matchIndex <= prefixStartIndex) ) {
size_t const dictHash = ZSTD_hashPtr(ip, dictHLog, mls);
U32 const dictMatchIndex = dictHashTable[dictHash];
const BYTE* dictMatch = dictBase + dictMatchIndex;
if (dictMatchIndex <= dictStartIndex ||
MEM_read32(dictMatch) != MEM_read32(ip)) {
assert(stepSize >= 1);
ip += ((ip-anchor) >> kSearchStrength) + stepSize;
continue;
} else {
/* found a dict match */
U32 const offset = (U32)(current-dictMatchIndex-dictIndexDelta);
mLength = ZSTD_count_2segments(ip+4, dictMatch+4, iend, dictEnd, prefixStart) + 4;
while (((ip>anchor) & (dictMatch>dictStart))
&& (ip[-1] == dictMatch[-1])) {
ip--; dictMatch--; mLength++;
} /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
}
} else if (MEM_read32(match) != MEM_read32(ip)) {
/* it's not a match, and we're not going to check the dictionary */
assert(stepSize >= 1);
ip += ((ip-anchor) >> kSearchStrength) + stepSize;
continue;
} else {
/* found a regular match */
U32 const offset = (U32)(ip-match);
mLength = ZSTD_count(ip+4, match+4, iend) + 4;
while (((ip>anchor) & (match>prefixStart))
&& (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
}
/* match found */
ip += mLength;
anchor = ip;
if (ip <= ilimit) {
/* Fill Table */
assert(base+current+2 > istart); /* check base overflow */
hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2; /* here because current+2 could be > iend-8 */
hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
/* check immediate repcode */
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex2 = current2 - offset_2;
const BYTE* repMatch2 = repIndex2 < prefixStartIndex ?
dictBase - dictIndexDelta + repIndex2 :
base + repIndex2;
if ( ((U32)((prefixStartIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
&& (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
ip += repLength2;
anchor = ip;
continue;
}
break;
}
}
}
/* save reps for next block */
rep[0] = offset_1 ? offset_1 : offsetSaved;
rep[1] = offset_2 ? offset_2 : offsetSaved;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_fast_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
U32 const mls = ms->cParams.minMatch;
assert(ms->dictMatchState != NULL);
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_fast_dictMatchState_generic(ms, seqStore, rep, src, srcSize, 7);
}
}
static size_t ZSTD_compressBlock_fast_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize, U32 const mls)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hlog = cParams->hashLog;
/* support stepSize of 0 */
U32 const stepSize = cParams->targetLength + !(cParams->targetLength);
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
const U32 dictStartIndex = lowLimit;
const BYTE* const dictStart = dictBase + dictStartIndex;
const U32 dictLimit = ms->window.dictLimit;
const U32 prefixStartIndex = dictLimit < lowLimit ? lowLimit : dictLimit;
const BYTE* const prefixStart = base + prefixStartIndex;
const BYTE* const dictEnd = dictBase + prefixStartIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
U32 offset_1=rep[0], offset_2=rep[1];
DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1);
/* switch to "regular" variant if extDict is invalidated due to maxDistance */
if (prefixStartIndex == dictStartIndex)
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, mls);
/* Search Loop */
while (ip < ilimit) { /* < instead of <=, because (ip+1) */
const size_t h = ZSTD_hashPtr(ip, hlog, mls);
const U32 matchIndex = hashTable[h];
const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
const BYTE* match = matchBase + matchIndex;
const U32 current = (U32)(ip-base);
const U32 repIndex = current + 1 - offset_1;
const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
hashTable[h] = current; /* update hash table */
DEBUGLOG(7, "offset_1 = %u , current = %u", offset_1, current);
assert(offset_1 <= current +1); /* check repIndex */
if ( (((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > dictStartIndex))
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
size_t const rLength = ZSTD_count_2segments(ip+1 +4, repMatch +4, iend, repMatchEnd, prefixStart) + 4;
ip++;
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, rLength-MINMATCH);
ip += rLength;
anchor = ip;
} else {
if ( (matchIndex < dictStartIndex) ||
(MEM_read32(match) != MEM_read32(ip)) ) {
assert(stepSize >= 1);
ip += ((ip-anchor) >> kSearchStrength) + stepSize;
continue;
}
{ const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
U32 const offset = current - matchIndex;
size_t mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
offset_2 = offset_1; offset_1 = offset; /* update offset history */
ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
ip += mLength;
anchor = ip;
} }
if (ip <= ilimit) {
/* Fill Table */
hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2;
hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
/* check immediate repcode */
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex2 = current2 - offset_2;
const BYTE* const repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) & (repIndex2 > dictStartIndex)) /* intentional overflow */
&& (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
{ U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; } /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStore, 0 /*litlen*/, anchor, iend, 0 /*offcode*/, repLength2-MINMATCH);
hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
ip += repLength2;
anchor = ip;
continue;
}
break;
} } }
/* save reps for next block */
rep[0] = offset_1;
rep[1] = offset_2;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_fast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
U32 const mls = ms->cParams.minMatch;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, 7);
}
}
module/zstd/lib/compress/zstd_fast.h
+37
View File
@@ -0,0 +1,37 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_FAST_H
#define ZSTD_FAST_H
#if defined (__cplusplus)
extern "C" {
#endif
#include "../common/mem.h" /* U32 */
#include "zstd_compress_internal.h"
void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
void const* end, ZSTD_dictTableLoadMethod_e dtlm);
size_t ZSTD_compressBlock_fast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_fast_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_fast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_FAST_H */
module/zstd/lib/compress/zstd_lazy.c
+1138
View File
File diff suppressed because it is too large Load Diff
module/zstd/lib/compress/zstd_lazy.h
+67
View File
@@ -0,0 +1,67 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_LAZY_H
#define ZSTD_LAZY_H
#if defined (__cplusplus)
extern "C" {
#endif
#include "zstd_compress_internal.h"
U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip);
void ZSTD_preserveUnsortedMark (U32* const table, U32 const size, U32 const reducerValue); /*! used in ZSTD_reduceIndex(). preemptively increase value of ZSTD_DUBT_UNSORTED_MARK */
size_t ZSTD_compressBlock_btlazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btlazy2_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btlazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_LAZY_H */
module/zstd/lib/compress/zstd_ldm.c
+619
View File
@@ -0,0 +1,619 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#include "zstd_ldm.h"
#include "../common/debug.h"
#include "zstd_fast.h" /* ZSTD_fillHashTable() */
#include "zstd_double_fast.h" /* ZSTD_fillDoubleHashTable() */
#define LDM_BUCKET_SIZE_LOG 3
#define LDM_MIN_MATCH_LENGTH 64
#define LDM_HASH_RLOG 7
#define LDM_HASH_CHAR_OFFSET 10
void ZSTD_ldm_adjustParameters(ldmParams_t* params,
ZSTD_compressionParameters const* cParams)
{
params->windowLog = cParams->windowLog;
ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
if (cParams->strategy >= ZSTD_btopt) {
/* Get out of the way of the optimal parser */
U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
params->minMatchLength = minMatch;
}
if (params->hashLog == 0) {
params->hashLog = MAX(ZSTD_HASHLOG_MIN, params->windowLog - LDM_HASH_RLOG);
assert(params->hashLog <= ZSTD_HASHLOG_MAX);
}
if (params->hashRateLog == 0) {
params->hashRateLog = params->windowLog < params->hashLog
? 0
: params->windowLog - params->hashLog;
}
params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
}
size_t ZSTD_ldm_getTableSize(ldmParams_t params)
{
size_t const ldmHSize = ((size_t)1) << params.hashLog;
size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog);
size_t const ldmBucketSize = ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
size_t const totalSize = ZSTD_cwksp_alloc_size(ldmBucketSize)
+ ZSTD_cwksp_alloc_size(ldmHSize * sizeof(ldmEntry_t));
return params.enableLdm ? totalSize : 0;
}
size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize)
{
return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
}
/** ZSTD_ldm_getSmallHash() :
* numBits should be <= 32
* If numBits==0, returns 0.
* @return : the most significant numBits of value. */
static U32 ZSTD_ldm_getSmallHash(U64 value, U32 numBits)
{
assert(numBits <= 32);
return numBits == 0 ? 0 : (U32)(value >> (64 - numBits));
}
/** ZSTD_ldm_getChecksum() :
* numBitsToDiscard should be <= 32
* @return : the next most significant 32 bits after numBitsToDiscard */
static U32 ZSTD_ldm_getChecksum(U64 hash, U32 numBitsToDiscard)
{
assert(numBitsToDiscard <= 32);
return (hash >> (64 - 32 - numBitsToDiscard)) & 0xFFFFFFFF;
}
/** ZSTD_ldm_getTag() ;
* Given the hash, returns the most significant numTagBits bits
* after (32 + hbits) bits.
*
* If there are not enough bits remaining, return the last
* numTagBits bits. */
static U32 ZSTD_ldm_getTag(U64 hash, U32 hbits, U32 numTagBits)
{
assert(numTagBits < 32 && hbits <= 32);
if (32 - hbits < numTagBits) {
return hash & (((U32)1 << numTagBits) - 1);
} else {
return (hash >> (32 - hbits - numTagBits)) & (((U32)1 << numTagBits) - 1);
}
}
/** ZSTD_ldm_getBucket() :
* Returns a pointer to the start of the bucket associated with hash. */
static ldmEntry_t* ZSTD_ldm_getBucket(
ldmState_t* ldmState, size_t hash, ldmParams_t const ldmParams)
{
return ldmState->hashTable + (hash << ldmParams.bucketSizeLog);
}
/** ZSTD_ldm_insertEntry() :
* Insert the entry with corresponding hash into the hash table */
static void ZSTD_ldm_insertEntry(ldmState_t* ldmState,
size_t const hash, const ldmEntry_t entry,
ldmParams_t const ldmParams)
{
BYTE* const bucketOffsets = ldmState->bucketOffsets;
*(ZSTD_ldm_getBucket(ldmState, hash, ldmParams) + bucketOffsets[hash]) = entry;
bucketOffsets[hash]++;
bucketOffsets[hash] &= ((U32)1 << ldmParams.bucketSizeLog) - 1;
}
/** ZSTD_ldm_makeEntryAndInsertByTag() :
*
* Gets the small hash, checksum, and tag from the rollingHash.
*
* If the tag matches (1 << ldmParams.hashRateLog)-1, then
* creates an ldmEntry from the offset, and inserts it into the hash table.
*
* hBits is the length of the small hash, which is the most significant hBits
* of rollingHash. The checksum is the next 32 most significant bits, followed
* by ldmParams.hashRateLog bits that make up the tag. */
static void ZSTD_ldm_makeEntryAndInsertByTag(ldmState_t* ldmState,
U64 const rollingHash,
U32 const hBits,
U32 const offset,
ldmParams_t const ldmParams)
{
U32 const tag = ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashRateLog);
U32 const tagMask = ((U32)1 << ldmParams.hashRateLog) - 1;
if (tag == tagMask) {
U32 const hash = ZSTD_ldm_getSmallHash(rollingHash, hBits);
U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
ldmEntry_t entry;
entry.offset = offset;
entry.checksum = checksum;
ZSTD_ldm_insertEntry(ldmState, hash, entry, ldmParams);
}
}
/** ZSTD_ldm_countBackwardsMatch() :
* Returns the number of bytes that match backwards before pIn and pMatch.
*
* We count only bytes where pMatch >= pBase and pIn >= pAnchor. */
static size_t ZSTD_ldm_countBackwardsMatch(
const BYTE* pIn, const BYTE* pAnchor,
const BYTE* pMatch, const BYTE* pBase)
{
size_t matchLength = 0;
while (pIn > pAnchor && pMatch > pBase && pIn[-1] == pMatch[-1]) {
pIn--;
pMatch--;
matchLength++;
}
return matchLength;
}
/** ZSTD_ldm_fillFastTables() :
*
* Fills the relevant tables for the ZSTD_fast and ZSTD_dfast strategies.
* This is similar to ZSTD_loadDictionaryContent.
*
* The tables for the other strategies are filled within their
* block compressors. */
static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t* ms,
void const* end)
{
const BYTE* const iend = (const BYTE*)end;
switch(ms->cParams.strategy)
{
case ZSTD_fast:
ZSTD_fillHashTable(ms, iend, ZSTD_dtlm_fast);
break;
case ZSTD_dfast:
ZSTD_fillDoubleHashTable(ms, iend, ZSTD_dtlm_fast);
break;
case ZSTD_greedy:
case ZSTD_lazy:
case ZSTD_lazy2:
case ZSTD_btlazy2:
case ZSTD_btopt:
case ZSTD_btultra:
case ZSTD_btultra2:
break;
default:
assert(0); /* not possible : not a valid strategy id */
}
return 0;
}
/** ZSTD_ldm_fillLdmHashTable() :
*
* Fills hashTable from (lastHashed + 1) to iend (non-inclusive).
* lastHash is the rolling hash that corresponds to lastHashed.
*
* Returns the rolling hash corresponding to position iend-1. */
static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
U64 lastHash, const BYTE* lastHashed,
const BYTE* iend, const BYTE* base,
U32 hBits, ldmParams_t const ldmParams)
{
U64 rollingHash = lastHash;
const BYTE* cur = lastHashed + 1;
while (cur < iend) {
rollingHash = ZSTD_rollingHash_rotate(rollingHash, cur[-1],
cur[ldmParams.minMatchLength-1],
state->hashPower);
ZSTD_ldm_makeEntryAndInsertByTag(state,
rollingHash, hBits,
(U32)(cur - base), ldmParams);
++cur;
}
return rollingHash;
}
void ZSTD_ldm_fillHashTable(
ldmState_t* state, const BYTE* ip,
const BYTE* iend, ldmParams_t const* params)
{
DEBUGLOG(5, "ZSTD_ldm_fillHashTable");
if ((size_t)(iend - ip) >= params->minMatchLength) {
U64 startingHash = ZSTD_rollingHash_compute(ip, params->minMatchLength);
ZSTD_ldm_fillLdmHashTable(
state, startingHash, ip, iend - params->minMatchLength, state->window.base,
params->hashLog - params->bucketSizeLog,
*params);
}
}
/** ZSTD_ldm_limitTableUpdate() :
*
* Sets cctx->nextToUpdate to a position corresponding closer to anchor
* if it is far way
* (after a long match, only update tables a limited amount). */
static void ZSTD_ldm_limitTableUpdate(ZSTD_matchState_t* ms, const BYTE* anchor)
{
U32 const current = (U32)(anchor - ms->window.base);
if (current > ms->nextToUpdate + 1024) {
ms->nextToUpdate =
current - MIN(512, current - ms->nextToUpdate - 1024);
}
}
static size_t ZSTD_ldm_generateSequences_internal(
ldmState_t* ldmState, rawSeqStore_t* rawSeqStore,
ldmParams_t const* params, void const* src, size_t srcSize)
{
/* LDM parameters */
int const extDict = ZSTD_window_hasExtDict(ldmState->window);
U32 const minMatchLength = params->minMatchLength;
U64 const hashPower = ldmState->hashPower;
U32 const hBits = params->hashLog - params->bucketSizeLog;
U32 const ldmBucketSize = 1U << params->bucketSizeLog;
U32 const hashRateLog = params->hashRateLog;
U32 const ldmTagMask = (1U << params->hashRateLog) - 1;
/* Prefix and extDict parameters */
U32 const dictLimit = ldmState->window.dictLimit;
U32 const lowestIndex = extDict ? ldmState->window.lowLimit : dictLimit;
BYTE const* const base = ldmState->window.base;
BYTE const* const dictBase = extDict ? ldmState->window.dictBase : NULL;
BYTE const* const dictStart = extDict ? dictBase + lowestIndex : NULL;
BYTE const* const dictEnd = extDict ? dictBase + dictLimit : NULL;
BYTE const* const lowPrefixPtr = base + dictLimit;
/* Input bounds */
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
BYTE const* const ilimit = iend - MAX(minMatchLength, HASH_READ_SIZE);
/* Input positions */
BYTE const* anchor = istart;
BYTE const* ip = istart;
/* Rolling hash */
BYTE const* lastHashed = NULL;
U64 rollingHash = 0;
while (ip <= ilimit) {
size_t mLength;
U32 const current = (U32)(ip - base);
size_t forwardMatchLength = 0, backwardMatchLength = 0;
ldmEntry_t* bestEntry = NULL;
if (ip != istart) {
rollingHash = ZSTD_rollingHash_rotate(rollingHash, lastHashed[0],
lastHashed[minMatchLength],
hashPower);
} else {
rollingHash = ZSTD_rollingHash_compute(ip, minMatchLength);
}
lastHashed = ip;
/* Do not insert and do not look for a match */
if (ZSTD_ldm_getTag(rollingHash, hBits, hashRateLog) != ldmTagMask) {
ip++;
continue;
}
/* Get the best entry and compute the match lengths */
{
ldmEntry_t* const bucket =
ZSTD_ldm_getBucket(ldmState,
ZSTD_ldm_getSmallHash(rollingHash, hBits),
*params);
ldmEntry_t* cur;
size_t bestMatchLength = 0;
U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
size_t curForwardMatchLength, curBackwardMatchLength,
curTotalMatchLength;
if (cur->checksum != checksum || cur->offset <= lowestIndex) {
continue;
}
if (extDict) {
BYTE const* const curMatchBase =
cur->offset < dictLimit ? dictBase : base;
BYTE const* const pMatch = curMatchBase + cur->offset;
BYTE const* const matchEnd =
cur->offset < dictLimit ? dictEnd : iend;
BYTE const* const lowMatchPtr =
cur->offset < dictLimit ? dictStart : lowPrefixPtr;
curForwardMatchLength = ZSTD_count_2segments(
ip, pMatch, iend,
matchEnd, lowPrefixPtr);
if (curForwardMatchLength < minMatchLength) {
continue;
}
curBackwardMatchLength =
ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
lowMatchPtr);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
} else { /* !extDict */
BYTE const* const pMatch = base + cur->offset;
curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
if (curForwardMatchLength < minMatchLength) {
continue;
}
curBackwardMatchLength =
ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
lowPrefixPtr);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
}
if (curTotalMatchLength > bestMatchLength) {
bestMatchLength = curTotalMatchLength;
forwardMatchLength = curForwardMatchLength;
backwardMatchLength = curBackwardMatchLength;
bestEntry = cur;
}
}
}
/* No match found -- continue searching */
if (bestEntry == NULL) {
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash,
hBits, current,
*params);
ip++;
continue;
}
/* Match found */
mLength = forwardMatchLength + backwardMatchLength;
ip -= backwardMatchLength;
{
/* Store the sequence:
* ip = current - backwardMatchLength
* The match is at (bestEntry->offset - backwardMatchLength)
*/
U32 const matchIndex = bestEntry->offset;
U32 const offset = current - matchIndex;
rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
/* Out of sequence storage */
if (rawSeqStore->size == rawSeqStore->capacity)
return ERROR(dstSize_tooSmall);
seq->litLength = (U32)(ip - anchor);
seq->matchLength = (U32)mLength;
seq->offset = offset;
rawSeqStore->size++;
}
/* Insert the current entry into the hash table */
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
(U32)(lastHashed - base),
*params);
assert(ip + backwardMatchLength == lastHashed);
/* Fill the hash table from lastHashed+1 to ip+mLength*/
/* Heuristic: don't need to fill the entire table at end of block */
if (ip + mLength <= ilimit) {
rollingHash = ZSTD_ldm_fillLdmHashTable(
ldmState, rollingHash, lastHashed,
ip + mLength, base, hBits, *params);
lastHashed = ip + mLength - 1;
}
ip += mLength;
anchor = ip;
}
return iend - anchor;
}
/*! ZSTD_ldm_reduceTable() :
* reduce table indexes by `reducerValue` */
static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
U32 const reducerValue)
{
U32 u;
for (u = 0; u < size; u++) {
if (table[u].offset < reducerValue) table[u].offset = 0;
else table[u].offset -= reducerValue;
}
}
size_t ZSTD_ldm_generateSequences(
ldmState_t* ldmState, rawSeqStore_t* sequences,
ldmParams_t const* params, void const* src, size_t srcSize)
{
U32 const maxDist = 1U << params->windowLog;
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
size_t const kMaxChunkSize = 1 << 20;
size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
size_t chunk;
size_t leftoverSize = 0;
assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
/* Check that ZSTD_window_update() has been called for this chunk prior
* to passing it to this function.
*/
assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize);
/* The input could be very large (in zstdmt), so it must be broken up into
* chunks to enforce the maximum distance and handle overflow correction.
*/
assert(sequences->pos <= sequences->size);
assert(sequences->size <= sequences->capacity);
for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
size_t const remaining = (size_t)(iend - chunkStart);
BYTE const *const chunkEnd =
(remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
size_t const chunkSize = chunkEnd - chunkStart;
size_t newLeftoverSize;
size_t const prevSize = sequences->size;
assert(chunkStart < iend);
/* 1. Perform overflow correction if necessary. */
if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
U32 const ldmHSize = 1U << params->hashLog;
U32 const correction = ZSTD_window_correctOverflow(
&ldmState->window, /* cycleLog */ 0, maxDist, chunkStart);
ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
/* invalidate dictionaries on overflow correction */
ldmState->loadedDictEnd = 0;
}
/* 2. We enforce the maximum offset allowed.
*
* kMaxChunkSize should be small enough that we don't lose too much of
* the window through early invalidation.
* TODO: * Test the chunk size.
* * Try invalidation after the sequence generation and test the
* the offset against maxDist directly.
*
* NOTE: Because of dictionaries + sequence splitting we MUST make sure
* that any offset used is valid at the END of the sequence, since it may
* be split into two sequences. This condition holds when using
* ZSTD_window_enforceMaxDist(), but if we move to checking offsets
* against maxDist directly, we'll have to carefully handle that case.
*/
ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, &ldmState->loadedDictEnd, NULL);
/* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
newLeftoverSize = ZSTD_ldm_generateSequences_internal(
ldmState, sequences, params, chunkStart, chunkSize);
if (ZSTD_isError(newLeftoverSize))
return newLeftoverSize;
/* 4. We add the leftover literals from previous iterations to the first
* newly generated sequence, or add the `newLeftoverSize` if none are
* generated.
*/
/* Prepend the leftover literals from the last call */
if (prevSize < sequences->size) {
sequences->seq[prevSize].litLength += (U32)leftoverSize;
leftoverSize = newLeftoverSize;
} else {
assert(newLeftoverSize == chunkSize);
leftoverSize += chunkSize;
}
}
return 0;
}
void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize, U32 const minMatch) {
while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
if (srcSize <= seq->litLength) {
/* Skip past srcSize literals */
seq->litLength -= (U32)srcSize;
return;
}
srcSize -= seq->litLength;
seq->litLength = 0;
if (srcSize < seq->matchLength) {
/* Skip past the first srcSize of the match */
seq->matchLength -= (U32)srcSize;
if (seq->matchLength < minMatch) {
/* The match is too short, omit it */
if (rawSeqStore->pos + 1 < rawSeqStore->size) {
seq[1].litLength += seq[0].matchLength;
}
rawSeqStore->pos++;
}
return;
}
srcSize -= seq->matchLength;
seq->matchLength = 0;
rawSeqStore->pos++;
}
}
/**
* If the sequence length is longer than remaining then the sequence is split
* between this block and the next.
*
* Returns the current sequence to handle, or if the rest of the block should
* be literals, it returns a sequence with offset == 0.
*/
static rawSeq maybeSplitSequence(rawSeqStore_t* rawSeqStore,
U32 const remaining, U32 const minMatch)
{
rawSeq sequence = rawSeqStore->seq[rawSeqStore->pos];
assert(sequence.offset > 0);
/* Likely: No partial sequence */
if (remaining >= sequence.litLength + sequence.matchLength) {
rawSeqStore->pos++;
return sequence;
}
/* Cut the sequence short (offset == 0 ==> rest is literals). */
if (remaining <= sequence.litLength) {
sequence.offset = 0;
} else if (remaining < sequence.litLength + sequence.matchLength) {
sequence.matchLength = remaining - sequence.litLength;
if (sequence.matchLength < minMatch) {
sequence.offset = 0;
}
}
/* Skip past `remaining` bytes for the future sequences. */
ZSTD_ldm_skipSequences(rawSeqStore, remaining, minMatch);
return sequence;
}
size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
unsigned const minMatch = cParams->minMatch;
ZSTD_blockCompressor const blockCompressor =
ZSTD_selectBlockCompressor(cParams->strategy, ZSTD_matchState_dictMode(ms));
/* Input bounds */
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
/* Input positions */
BYTE const* ip = istart;
DEBUGLOG(5, "ZSTD_ldm_blockCompress: srcSize=%zu", srcSize);
assert(rawSeqStore->pos <= rawSeqStore->size);
assert(rawSeqStore->size <= rawSeqStore->capacity);
/* Loop through each sequence and apply the block compressor to the lits */
while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
/* maybeSplitSequence updates rawSeqStore->pos */
rawSeq const sequence = maybeSplitSequence(rawSeqStore,
(U32)(iend - ip), minMatch);
int i;
/* End signal */
if (sequence.offset == 0)
break;
assert(ip + sequence.litLength + sequence.matchLength <= iend);
/* Fill tables for block compressor */
ZSTD_ldm_limitTableUpdate(ms, ip);
ZSTD_ldm_fillFastTables(ms, ip);
/* Run the block compressor */
DEBUGLOG(5, "pos %u : calling block compressor on segment of size %u", (unsigned)(ip-istart), sequence.litLength);
{
size_t const newLitLength =
blockCompressor(ms, seqStore, rep, ip, sequence.litLength);
ip += sequence.litLength;
/* Update the repcodes */
for (i = ZSTD_REP_NUM - 1; i > 0; i--)
rep[i] = rep[i-1];
rep[0] = sequence.offset;
/* Store the sequence */
ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, iend,
sequence.offset + ZSTD_REP_MOVE,
sequence.matchLength - MINMATCH);
ip += sequence.matchLength;
}
}
/* Fill the tables for the block compressor */
ZSTD_ldm_limitTableUpdate(ms, ip);
ZSTD_ldm_fillFastTables(ms, ip);
/* Compress the last literals */
return blockCompressor(ms, seqStore, rep, ip, iend - ip);
}
module/zstd/lib/compress/zstd_ldm.h
+110
View File
@@ -0,0 +1,110 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_LDM_H
#define ZSTD_LDM_H
#if defined (__cplusplus)
extern "C" {
#endif
#include "zstd_compress_internal.h" /* ldmParams_t, U32 */
#include "../zstd.h" /* ZSTD_CCtx, size_t */
/*-*************************************
* Long distance matching
***************************************/
#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_LIMIT_DEFAULT
void ZSTD_ldm_fillHashTable(
ldmState_t* state, const BYTE* ip,
const BYTE* iend, ldmParams_t const* params);
/**
* ZSTD_ldm_generateSequences():
*
* Generates the sequences using the long distance match finder.
* Generates long range matching sequences in `sequences`, which parse a prefix
* of the source. `sequences` must be large enough to store every sequence,
* which can be checked with `ZSTD_ldm_getMaxNbSeq()`.
* @returns 0 or an error code.
*
* NOTE: The user must have called ZSTD_window_update() for all of the input
* they have, even if they pass it to ZSTD_ldm_generateSequences() in chunks.
* NOTE: This function returns an error if it runs out of space to store
* sequences.
*/
size_t ZSTD_ldm_generateSequences(
ldmState_t* ldms, rawSeqStore_t* sequences,
ldmParams_t const* params, void const* src, size_t srcSize);
/**
* ZSTD_ldm_blockCompress():
*
* Compresses a block using the predefined sequences, along with a secondary
* block compressor. The literals section of every sequence is passed to the
* secondary block compressor, and those sequences are interspersed with the
* predefined sequences. Returns the length of the last literals.
* Updates `rawSeqStore.pos` to indicate how many sequences have been consumed.
* `rawSeqStore.seq` may also be updated to split the last sequence between two
* blocks.
* @return The length of the last literals.
*
* NOTE: The source must be at most the maximum block size, but the predefined
* sequences can be any size, and may be longer than the block. In the case that
* they are longer than the block, the last sequences may need to be split into
* two. We handle that case correctly, and update `rawSeqStore` appropriately.
* NOTE: This function does not return any errors.
*/
size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
/**
* ZSTD_ldm_skipSequences():
*
* Skip past `srcSize` bytes worth of sequences in `rawSeqStore`.
* Avoids emitting matches less than `minMatch` bytes.
* Must be called for data with is not passed to ZSTD_ldm_blockCompress().
*/
void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize,
U32 const minMatch);
/** ZSTD_ldm_getTableSize() :
* Estimate the space needed for long distance matching tables or 0 if LDM is
* disabled.
*/
size_t ZSTD_ldm_getTableSize(ldmParams_t params);
/** ZSTD_ldm_getSeqSpace() :
* Return an upper bound on the number of sequences that can be produced by
* the long distance matcher, or 0 if LDM is disabled.
*/
size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
/** ZSTD_ldm_adjustParameters() :
* If the params->hashRateLog is not set, set it to its default value based on
* windowLog and params->hashLog.
*
* Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
* params->hashLog if it is not).
*
* Ensures that the minMatchLength >= targetLength during optimal parsing.
*/
void ZSTD_ldm_adjustParameters(ldmParams_t* params,
ZSTD_compressionParameters const* cParams);
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_FAST_H */
module/zstd/lib/compress/zstd_opt.c
+1200
View File
File diff suppressed because it is too large Load Diff
module/zstd/lib/compress/zstd_opt.h
+56
View File
@@ -0,0 +1,56 @@
/*
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#ifndef ZSTD_OPT_H
#define ZSTD_OPT_H
#if defined (__cplusplus)
extern "C" {
#endif
#include "zstd_compress_internal.h"
/* used in ZSTD_loadDictionaryContent() */
void ZSTD_updateTree(ZSTD_matchState_t* ms, const BYTE* ip, const BYTE* iend);
size_t ZSTD_compressBlock_btopt(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btopt_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btopt_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize);
/* note : no btultra2 variant for extDict nor dictMatchState,
* because btultra2 is not meant to work with dictionaries
* and is only specific for the first block (no prefix) */
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_OPT_H */