tracy/zstd/zstd_lazy.c

1413 lines
62 KiB
C
Raw Normal View History

2020-02-08 13:56:59 +00:00
/*
2020-06-04 19:03:27 +00:00
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
2020-02-08 13:56:59 +00:00
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
*/
#include "zstd_compress_internal.h"
#include "zstd_lazy.h"
/*-*************************************
* Binary Tree search
***************************************/
static void
ZSTD_updateDUBT(ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* iend,
U32 mls)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hashLog = cParams->hashLog;
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
const BYTE* const base = ms->window.base;
U32 const target = (U32)(ip - base);
U32 idx = ms->nextToUpdate;
if (idx != target)
DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
idx, target, ms->window.dictLimit);
assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */
(void)iend;
assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */
for ( ; idx < target ; idx++) {
size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */
U32 const matchIndex = hashTable[h];
U32* const nextCandidatePtr = bt + 2*(idx&btMask);
U32* const sortMarkPtr = nextCandidatePtr + 1;
DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
hashTable[h] = idx; /* Update Hash Table */
*nextCandidatePtr = matchIndex; /* update BT like a chain */
*sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
}
ms->nextToUpdate = target;
}
/** ZSTD_insertDUBT1() :
* sort one already inserted but unsorted position
2020-12-18 14:46:27 +00:00
* assumption : curr >= btlow == (curr - btmask)
2020-02-08 13:56:59 +00:00
* doesn't fail */
static void
ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
2020-12-18 14:46:27 +00:00
U32 curr, const BYTE* inputEnd,
2020-02-08 13:56:59 +00:00
U32 nbCompares, U32 btLow,
const ZSTD_dictMode_e dictMode)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
2020-12-18 14:46:27 +00:00
const BYTE* const ip = (curr>=dictLimit) ? base + curr : dictBase + curr;
const BYTE* const iend = (curr>=dictLimit) ? inputEnd : dictBase + dictLimit;
2020-02-08 13:56:59 +00:00
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* match;
2020-12-18 14:46:27 +00:00
U32* smallerPtr = bt + 2*(curr&btMask);
2020-02-08 13:56:59 +00:00
U32* largerPtr = smallerPtr + 1;
U32 matchIndex = *smallerPtr; /* this candidate is unsorted : next sorted candidate is reached through *smallerPtr, while *largerPtr contains previous unsorted candidate (which is already saved and can be overwritten) */
U32 dummy32; /* to be nullified at the end */
U32 const windowValid = ms->window.lowLimit;
U32 const maxDistance = 1U << cParams->windowLog;
2020-12-18 14:46:27 +00:00
U32 const windowLow = (curr - windowValid > maxDistance) ? curr - maxDistance : windowValid;
2020-02-08 13:56:59 +00:00
DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
2020-12-18 14:46:27 +00:00
curr, dictLimit, windowLow);
assert(curr >= btLow);
2020-02-08 13:56:59 +00:00
assert(ip < iend); /* condition for ZSTD_count */
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
2020-12-18 14:46:27 +00:00
assert(matchIndex < curr);
2020-02-08 13:56:59 +00:00
/* note : all candidates are now supposed sorted,
* but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
* when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
if ( (dictMode != ZSTD_extDict)
|| (matchIndex+matchLength >= dictLimit) /* both in current segment*/
2020-12-18 14:46:27 +00:00
|| (curr < dictLimit) /* both in extDict */) {
2020-02-08 13:56:59 +00:00
const BYTE* const mBase = ( (dictMode != ZSTD_extDict)
|| (matchIndex+matchLength >= dictLimit)) ?
base : dictBase;
assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */
2020-12-18 14:46:27 +00:00
|| (curr < dictLimit) );
2020-02-08 13:56:59 +00:00
match = mBase + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (matchIndex+matchLength >= dictLimit)
match = base + matchIndex; /* preparation for next read of match[matchLength] */
}
DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
2020-12-18 14:46:27 +00:00
curr, matchIndex, (U32)matchLength);
2020-02-08 13:56:59 +00:00
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
}
if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */
/* match is smaller than current */
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
matchIndex, btLow, nextPtr[1]);
smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
} else {
/* match is larger than current */
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
matchIndex, btLow, nextPtr[0]);
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
}
static size_t
ZSTD_DUBT_findBetterDictMatch (
ZSTD_matchState_t* ms,
const BYTE* const ip, const BYTE* const iend,
size_t* offsetPtr,
size_t bestLength,
U32 nbCompares,
U32 const mls,
const ZSTD_dictMode_e dictMode)
{
const ZSTD_matchState_t * const dms = ms->dictMatchState;
const ZSTD_compressionParameters* const dmsCParams = &dms->cParams;
const U32 * const dictHashTable = dms->hashTable;
U32 const hashLog = dmsCParams->hashLog;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32 dictMatchIndex = dictHashTable[h];
const BYTE* const base = ms->window.base;
const BYTE* const prefixStart = base + ms->window.dictLimit;
2020-12-18 14:46:27 +00:00
U32 const curr = (U32)(ip-base);
2020-02-08 13:56:59 +00:00
const BYTE* const dictBase = dms->window.base;
const BYTE* const dictEnd = dms->window.nextSrc;
U32 const dictHighLimit = (U32)(dms->window.nextSrc - dms->window.base);
U32 const dictLowLimit = dms->window.lowLimit;
U32 const dictIndexDelta = ms->window.lowLimit - dictHighLimit;
U32* const dictBt = dms->chainTable;
U32 const btLog = dmsCParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
U32 const btLow = (btMask >= dictHighLimit - dictLowLimit) ? dictLowLimit : dictHighLimit - btMask;
size_t commonLengthSmaller=0, commonLengthLarger=0;
(void)dictMode;
assert(dictMode == ZSTD_dictMatchState);
while (nbCompares-- && (dictMatchIndex > dictLowLimit)) {
U32* const nextPtr = dictBt + 2*(dictMatchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
const BYTE* match = dictBase + dictMatchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (dictMatchIndex+matchLength >= dictHighLimit)
match = base + dictMatchIndex + dictIndexDelta; /* to prepare for next usage of match[matchLength] */
if (matchLength > bestLength) {
U32 matchIndex = dictMatchIndex + dictIndexDelta;
2020-12-18 14:46:27 +00:00
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) {
2020-02-08 13:56:59 +00:00
DEBUGLOG(9, "ZSTD_DUBT_findBetterDictMatch(%u) : found better match length %u -> %u and offsetCode %u -> %u (dictMatchIndex %u, matchIndex %u)",
2020-12-18 14:46:27 +00:00
curr, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, ZSTD_REP_MOVE + curr - matchIndex, dictMatchIndex, matchIndex);
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + curr - matchIndex;
2020-02-08 13:56:59 +00:00
}
if (ip+matchLength == iend) { /* reached end of input : ip[matchLength] is not valid, no way to know if it's larger or smaller than match */
break; /* drop, to guarantee consistency (miss a little bit of compression) */
}
}
if (match[matchLength] < ip[matchLength]) {
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
dictMatchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
} else {
/* match is larger than current */
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
commonLengthLarger = matchLength;
dictMatchIndex = nextPtr[0];
}
}
if (bestLength >= MINMATCH) {
2020-12-18 14:46:27 +00:00
U32 const mIndex = curr - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
2020-02-08 13:56:59 +00:00
DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
2020-12-18 14:46:27 +00:00
curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
2020-02-08 13:56:59 +00:00
}
return bestLength;
}
static size_t
ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
const BYTE* const ip, const BYTE* const iend,
size_t* offsetPtr,
U32 const mls,
const ZSTD_dictMode_e dictMode)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hashLog = cParams->hashLog;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32 matchIndex = hashTable[h];
const BYTE* const base = ms->window.base;
2020-12-18 14:46:27 +00:00
U32 const curr = (U32)(ip-base);
U32 const windowLow = ZSTD_getLowestMatchIndex(ms, curr, cParams->windowLog);
2020-02-08 13:56:59 +00:00
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
2020-12-18 14:46:27 +00:00
U32 const btLow = (btMask >= curr) ? 0 : curr - btMask;
2020-02-08 13:56:59 +00:00
U32 const unsortLimit = MAX(btLow, windowLow);
U32* nextCandidate = bt + 2*(matchIndex&btMask);
U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1;
U32 nbCompares = 1U << cParams->searchLog;
U32 nbCandidates = nbCompares;
U32 previousCandidate = 0;
2020-12-18 14:46:27 +00:00
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", curr);
2020-02-08 13:56:59 +00:00
assert(ip <= iend-8); /* required for h calculation */
2020-12-18 14:46:27 +00:00
assert(dictMode != ZSTD_dedicatedDictSearch);
2020-02-08 13:56:59 +00:00
/* reach end of unsorted candidates list */
while ( (matchIndex > unsortLimit)
&& (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
&& (nbCandidates > 1) ) {
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
matchIndex);
*unsortedMark = previousCandidate; /* the unsortedMark becomes a reversed chain, to move up back to original position */
previousCandidate = matchIndex;
matchIndex = *nextCandidate;
nextCandidate = bt + 2*(matchIndex&btMask);
unsortedMark = bt + 2*(matchIndex&btMask) + 1;
nbCandidates --;
}
/* nullify last candidate if it's still unsorted
* simplification, detrimental to compression ratio, beneficial for speed */
if ( (matchIndex > unsortLimit)
&& (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
matchIndex);
*nextCandidate = *unsortedMark = 0;
}
/* batch sort stacked candidates */
matchIndex = previousCandidate;
while (matchIndex) { /* will end on matchIndex == 0 */
U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
U32 const nextCandidateIdx = *nextCandidateIdxPtr;
ZSTD_insertDUBT1(ms, matchIndex, iend,
nbCandidates, unsortLimit, dictMode);
matchIndex = nextCandidateIdx;
nbCandidates++;
}
/* find longest match */
{ size_t commonLengthSmaller = 0, commonLengthLarger = 0;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
2020-12-18 14:46:27 +00:00
U32* smallerPtr = bt + 2*(curr&btMask);
U32* largerPtr = bt + 2*(curr&btMask) + 1;
U32 matchEndIdx = curr + 8 + 1;
2020-02-08 13:56:59 +00:00
U32 dummy32; /* to be nullified at the end */
size_t bestLength = 0;
matchIndex = hashTable[h];
2020-12-18 14:46:27 +00:00
hashTable[h] = curr; /* Update Hash Table */
2020-02-08 13:56:59 +00:00
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
const BYTE* match;
if ((dictMode != ZSTD_extDict) || (matchIndex+matchLength >= dictLimit)) {
match = base + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (matchIndex+matchLength >= dictLimit)
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
}
if (matchLength > bestLength) {
if (matchLength > matchEndIdx - matchIndex)
matchEndIdx = matchIndex + (U32)matchLength;
2020-12-18 14:46:27 +00:00
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + curr - matchIndex;
2020-02-08 13:56:59 +00:00
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
if (dictMode == ZSTD_dictMatchState) {
nbCompares = 0; /* in addition to avoiding checking any
* further in this loop, make sure we
* skip checking in the dictionary. */
}
break; /* drop, to guarantee consistency (miss a little bit of compression) */
}
}
if (match[matchLength] < ip[matchLength]) {
/* match is smaller than current */
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
} else {
/* match is larger than current */
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
if (dictMode == ZSTD_dictMatchState && nbCompares) {
bestLength = ZSTD_DUBT_findBetterDictMatch(
ms, ip, iend,
offsetPtr, bestLength, nbCompares,
mls, dictMode);
}
2020-12-18 14:46:27 +00:00
assert(matchEndIdx > curr+8); /* ensure nextToUpdate is increased */
2020-02-08 13:56:59 +00:00
ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
if (bestLength >= MINMATCH) {
2020-12-18 14:46:27 +00:00
U32 const mIndex = curr - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
2020-02-08 13:56:59 +00:00
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
2020-12-18 14:46:27 +00:00
curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
2020-02-08 13:56:59 +00:00
}
return bestLength;
}
}
/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
FORCE_INLINE_TEMPLATE size_t
ZSTD_BtFindBestMatch( ZSTD_matchState_t* ms,
const BYTE* const ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 mls /* template */,
const ZSTD_dictMode_e dictMode)
{
DEBUGLOG(7, "ZSTD_BtFindBestMatch");
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
ZSTD_updateDUBT(ms, ip, iLimit, mls);
return ZSTD_DUBT_findBestMatch(ms, ip, iLimit, offsetPtr, mls, dictMode);
}
static size_t
ZSTD_BtFindBestMatch_selectMLS ( ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
case 7 :
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
}
}
static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
case 7 :
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
}
}
static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
case 7 :
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
}
}
/* *********************************
* Hash Chain
***********************************/
#define NEXT_IN_CHAIN(d, mask) chainTable[(d) & (mask)]
/* Update chains up to ip (excluded)
Assumption : always within prefix (i.e. not within extDict) */
2020-12-18 14:46:27 +00:00
FORCE_INLINE_TEMPLATE U32 ZSTD_insertAndFindFirstIndex_internal(
2020-02-08 13:56:59 +00:00
ZSTD_matchState_t* ms,
const ZSTD_compressionParameters* const cParams,
const BYTE* ip, U32 const mls)
{
U32* const hashTable = ms->hashTable;
const U32 hashLog = cParams->hashLog;
U32* const chainTable = ms->chainTable;
const U32 chainMask = (1 << cParams->chainLog) - 1;
const BYTE* const base = ms->window.base;
const U32 target = (U32)(ip - base);
U32 idx = ms->nextToUpdate;
while(idx < target) { /* catch up */
size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
hashTable[h] = idx;
idx++;
}
ms->nextToUpdate = target;
return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
}
U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
const ZSTD_compressionParameters* const cParams = &ms->cParams;
return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
}
2020-12-18 14:46:27 +00:00
void ZSTD_dedicatedDictSearch_lazy_loadDictionary(ZSTD_matchState_t* ms, const BYTE* const ip)
{
const BYTE* const base = ms->window.base;
U32 const target = (U32)(ip - base);
U32* const hashTable = ms->hashTable;
U32* const chainTable = ms->chainTable;
U32 const chainSize = 1 << ms->cParams.chainLog;
U32 idx = ms->nextToUpdate;
U32 const minChain = chainSize < target ? target - chainSize : idx;
U32 const bucketSize = 1 << ZSTD_LAZY_DDSS_BUCKET_LOG;
U32 const cacheSize = bucketSize - 1;
U32 const chainAttempts = (1 << ms->cParams.searchLog) - cacheSize;
U32 const chainLimit = chainAttempts > 255 ? 255 : chainAttempts;
/* We know the hashtable is oversized by a factor of `bucketSize`.
* We are going to temporarily pretend `bucketSize == 1`, keeping only a
* single entry. We will use the rest of the space to construct a temporary
* chaintable.
*/
U32 const hashLog = ms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG;
U32* const tmpHashTable = hashTable;
U32* const tmpChainTable = hashTable + ((size_t)1 << hashLog);
U32 const tmpChainSize = ((1 << ZSTD_LAZY_DDSS_BUCKET_LOG) - 1) << hashLog;
U32 const tmpMinChain = tmpChainSize < target ? target - tmpChainSize : idx;
U32 hashIdx;
assert(ms->cParams.chainLog <= 24);
assert(ms->cParams.hashLog >= ms->cParams.chainLog);
assert(idx != 0);
assert(tmpMinChain <= minChain);
/* fill conventional hash table and conventional chain table */
for ( ; idx < target; idx++) {
U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch);
if (idx >= tmpMinChain) {
tmpChainTable[idx - tmpMinChain] = hashTable[h];
}
tmpHashTable[h] = idx;
}
/* sort chains into ddss chain table */
{
U32 chainPos = 0;
for (hashIdx = 0; hashIdx < (1U << hashLog); hashIdx++) {
U32 count;
U32 countBeyondMinChain = 0;
U32 i = tmpHashTable[hashIdx];
for (count = 0; i >= tmpMinChain && count < cacheSize; count++) {
/* skip through the chain to the first position that won't be
* in the hash cache bucket */
if (i < minChain) {
countBeyondMinChain++;
}
i = tmpChainTable[i - tmpMinChain];
}
if (count == cacheSize) {
for (count = 0; count < chainLimit;) {
if (i < minChain) {
if (!i || countBeyondMinChain++ > cacheSize) {
/* only allow pulling `cacheSize` number of entries
* into the cache or chainTable beyond `minChain`,
* to replace the entries pulled out of the
* chainTable into the cache. This lets us reach
* back further without increasing the total number
* of entries in the chainTable, guaranteeing the
* DDSS chain table will fit into the space
* allocated for the regular one. */
break;
}
}
chainTable[chainPos++] = i;
count++;
if (i < tmpMinChain) {
break;
}
i = tmpChainTable[i - tmpMinChain];
}
} else {
count = 0;
}
if (count) {
tmpHashTable[hashIdx] = ((chainPos - count) << 8) + count;
} else {
tmpHashTable[hashIdx] = 0;
}
}
assert(chainPos <= chainSize); /* I believe this is guaranteed... */
}
/* move chain pointers into the last entry of each hash bucket */
for (hashIdx = (1 << hashLog); hashIdx; ) {
U32 const bucketIdx = --hashIdx << ZSTD_LAZY_DDSS_BUCKET_LOG;
U32 const chainPackedPointer = tmpHashTable[hashIdx];
U32 i;
for (i = 0; i < cacheSize; i++) {
hashTable[bucketIdx + i] = 0;
}
hashTable[bucketIdx + bucketSize - 1] = chainPackedPointer;
}
/* fill the buckets of the hash table */
for (idx = ms->nextToUpdate; idx < target; idx++) {
U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch)
<< ZSTD_LAZY_DDSS_BUCKET_LOG;
U32 i;
/* Shift hash cache down 1. */
for (i = cacheSize - 1; i; i--)
hashTable[h + i] = hashTable[h + i - 1];
hashTable[h] = idx;
}
ms->nextToUpdate = target;
}
2020-02-08 13:56:59 +00:00
/* inlining is important to hardwire a hot branch (template emulation) */
FORCE_INLINE_TEMPLATE
size_t ZSTD_HcFindBestMatch_generic (
ZSTD_matchState_t* ms,
const BYTE* const ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 mls, const ZSTD_dictMode_e dictMode)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const chainTable = ms->chainTable;
const U32 chainSize = (1 << cParams->chainLog);
const U32 chainMask = chainSize-1;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
2020-12-18 14:46:27 +00:00
const U32 curr = (U32)(ip-base);
2020-02-08 13:56:59 +00:00
const U32 maxDistance = 1U << cParams->windowLog;
const U32 lowestValid = ms->window.lowLimit;
2020-12-18 14:46:27 +00:00
const U32 withinMaxDistance = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
2020-02-08 13:56:59 +00:00
const U32 isDictionary = (ms->loadedDictEnd != 0);
const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
2020-12-18 14:46:27 +00:00
const U32 minChain = curr > chainSize ? curr - chainSize : 0;
2020-02-08 13:56:59 +00:00
U32 nbAttempts = 1U << cParams->searchLog;
size_t ml=4-1;
2020-12-18 14:46:27 +00:00
const ZSTD_matchState_t* const dms = ms->dictMatchState;
const U32 ddsHashLog = dictMode == ZSTD_dedicatedDictSearch
? dms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
const size_t ddsIdx = dictMode == ZSTD_dedicatedDictSearch
? ZSTD_hashPtr(ip, ddsHashLog, mls) << ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
U32 matchIndex;
if (dictMode == ZSTD_dedicatedDictSearch) {
const U32* entry = &dms->hashTable[ddsIdx];
PREFETCH_L1(entry);
}
2020-02-08 13:56:59 +00:00
/* HC4 match finder */
2020-12-18 14:46:27 +00:00
matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
2020-02-08 13:56:59 +00:00
2020-12-18 14:46:27 +00:00
for ( ; (matchIndex>=lowLimit) & (nbAttempts>0) ; nbAttempts--) {
2020-02-08 13:56:59 +00:00
size_t currentMl=0;
if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
const BYTE* const match = base + matchIndex;
assert(matchIndex >= dictLimit); /* ensures this is true if dictMode != ZSTD_extDict */
if (match[ml] == ip[ml]) /* potentially better */
currentMl = ZSTD_count(ip, match, iLimit);
} else {
const BYTE* const match = dictBase + matchIndex;
assert(match+4 <= dictEnd);
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
}
/* save best solution */
if (currentMl > ml) {
ml = currentMl;
2020-12-18 14:46:27 +00:00
*offsetPtr = curr - matchIndex + ZSTD_REP_MOVE;
2020-02-08 13:56:59 +00:00
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
}
if (matchIndex <= minChain) break;
matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
}
2020-12-18 14:46:27 +00:00
if (dictMode == ZSTD_dedicatedDictSearch) {
const U32 ddsLowestIndex = dms->window.dictLimit;
const BYTE* const ddsBase = dms->window.base;
const BYTE* const ddsEnd = dms->window.nextSrc;
const U32 ddsSize = (U32)(ddsEnd - ddsBase);
const U32 ddsIndexDelta = dictLimit - ddsSize;
const U32 bucketSize = (1 << ZSTD_LAZY_DDSS_BUCKET_LOG);
const U32 bucketLimit = nbAttempts < bucketSize - 1 ? nbAttempts : bucketSize - 1;
U32 ddsAttempt;
for (ddsAttempt = 0; ddsAttempt < bucketSize - 1; ddsAttempt++) {
PREFETCH_L1(ddsBase + dms->hashTable[ddsIdx + ddsAttempt]);
}
{
U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
U32 const chainIndex = chainPackedPointer >> 8;
PREFETCH_L1(&dms->chainTable[chainIndex]);
}
for (ddsAttempt = 0; ddsAttempt < bucketLimit; ddsAttempt++) {
size_t currentMl=0;
const BYTE* match;
matchIndex = dms->hashTable[ddsIdx + ddsAttempt];
match = ddsBase + matchIndex;
if (!matchIndex) {
return ml;
}
/* guaranteed by table construction */
(void)ddsLowestIndex;
assert(matchIndex >= ddsLowestIndex);
assert(match+4 <= ddsEnd);
if (MEM_read32(match) == MEM_read32(ip)) {
/* assumption : matchIndex <= dictLimit-4 (by table construction) */
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
}
/* save best solution */
if (currentMl > ml) {
ml = currentMl;
*offsetPtr = curr - (matchIndex + ddsIndexDelta) + ZSTD_REP_MOVE;
if (ip+currentMl == iLimit) {
/* best possible, avoids read overflow on next attempt */
return ml;
}
}
}
{
U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
U32 chainIndex = chainPackedPointer >> 8;
U32 const chainLength = chainPackedPointer & 0xFF;
U32 const chainAttempts = nbAttempts - ddsAttempt;
U32 const chainLimit = chainAttempts > chainLength ? chainLength : chainAttempts;
U32 chainAttempt;
for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++) {
PREFETCH_L1(ddsBase + dms->chainTable[chainIndex + chainAttempt]);
}
for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++, chainIndex++) {
size_t currentMl=0;
const BYTE* match;
matchIndex = dms->chainTable[chainIndex];
match = ddsBase + matchIndex;
/* guaranteed by table construction */
assert(matchIndex >= ddsLowestIndex);
assert(match+4 <= ddsEnd);
if (MEM_read32(match) == MEM_read32(ip)) {
/* assumption : matchIndex <= dictLimit-4 (by table construction) */
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
}
/* save best solution */
if (currentMl > ml) {
ml = currentMl;
*offsetPtr = curr - (matchIndex + ddsIndexDelta) + ZSTD_REP_MOVE;
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
}
}
}
} else if (dictMode == ZSTD_dictMatchState) {
2020-02-08 13:56:59 +00:00
const U32* const dmsChainTable = dms->chainTable;
const U32 dmsChainSize = (1 << dms->cParams.chainLog);
const U32 dmsChainMask = dmsChainSize - 1;
const U32 dmsLowestIndex = dms->window.dictLimit;
const BYTE* const dmsBase = dms->window.base;
const BYTE* const dmsEnd = dms->window.nextSrc;
const U32 dmsSize = (U32)(dmsEnd - dmsBase);
const U32 dmsIndexDelta = dictLimit - dmsSize;
const U32 dmsMinChain = dmsSize > dmsChainSize ? dmsSize - dmsChainSize : 0;
matchIndex = dms->hashTable[ZSTD_hashPtr(ip, dms->cParams.hashLog, mls)];
2020-12-18 14:46:27 +00:00
for ( ; (matchIndex>=dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
2020-02-08 13:56:59 +00:00
size_t currentMl=0;
const BYTE* const match = dmsBase + matchIndex;
assert(match+4 <= dmsEnd);
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
/* save best solution */
if (currentMl > ml) {
ml = currentMl;
2020-12-18 14:46:27 +00:00
*offsetPtr = curr - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
2020-02-08 13:56:59 +00:00
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
}
if (matchIndex <= dmsMinChain) break;
2020-12-18 14:46:27 +00:00
2020-02-08 13:56:59 +00:00
matchIndex = dmsChainTable[matchIndex & dmsChainMask];
}
}
return ml;
}
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
}
}
static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
}
}
2020-12-18 14:46:27 +00:00
static size_t ZSTD_HcFindBestMatch_dedicatedDictSearch_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dedicatedDictSearch);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dedicatedDictSearch);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dedicatedDictSearch);
}
}
2020-02-08 13:56:59 +00:00
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr)
{
switch(ms->cParams.minMatch)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
}
}
/* *******************************
* Common parser - lazy strategy
*********************************/
typedef enum { search_hashChain, search_binaryTree } searchMethod_e;
FORCE_INLINE_TEMPLATE size_t
ZSTD_compressBlock_lazy_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore,
U32 rep[ZSTD_REP_NUM],
const void* src, size_t srcSize,
const searchMethod_e searchMethod, const U32 depth,
ZSTD_dictMode_e const dictMode)
{
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 prefixLowestIndex = ms->window.dictLimit;
const BYTE* const prefixLowest = base + prefixLowestIndex;
typedef size_t (*searchMax_f)(
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
2020-12-18 14:46:27 +00:00
/**
* This table is indexed first by the four ZSTD_dictMode_e values, and then
* by the two searchMethod_e values. NULLs are placed for configurations
* that should never occur (extDict modes go to the other implementation
* below and there is no DDSS for binary tree search yet).
*/
const searchMax_f searchFuncs[4][2] = {
{
ZSTD_HcFindBestMatch_selectMLS,
ZSTD_BtFindBestMatch_selectMLS
},
{
NULL,
NULL
},
{
ZSTD_HcFindBestMatch_dictMatchState_selectMLS,
ZSTD_BtFindBestMatch_dictMatchState_selectMLS
},
{
ZSTD_HcFindBestMatch_dedicatedDictSearch_selectMLS,
NULL
}
};
searchMax_f const searchMax = searchFuncs[dictMode][searchMethod == search_binaryTree];
2020-02-08 13:56:59 +00:00
U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
2020-12-18 14:46:27 +00:00
const int isDMS = dictMode == ZSTD_dictMatchState;
const int isDDS = dictMode == ZSTD_dedicatedDictSearch;
const int isDxS = isDMS || isDDS;
2020-02-08 13:56:59 +00:00
const ZSTD_matchState_t* const dms = ms->dictMatchState;
2020-12-18 14:46:27 +00:00
const U32 dictLowestIndex = isDxS ? dms->window.dictLimit : 0;
const BYTE* const dictBase = isDxS ? dms->window.base : NULL;
const BYTE* const dictLowest = isDxS ? dictBase + dictLowestIndex : NULL;
const BYTE* const dictEnd = isDxS ? dms->window.nextSrc : NULL;
const U32 dictIndexDelta = isDxS ?
2020-02-08 13:56:59 +00:00
prefixLowestIndex - (U32)(dictEnd - dictBase) :
0;
const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
2020-12-18 14:46:27 +00:00
assert(searchMax != NULL);
2020-06-04 19:03:27 +00:00
DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u)", (U32)dictMode);
2020-02-08 13:56:59 +00:00
/* init */
ip += (dictAndPrefixLength == 0);
if (dictMode == ZSTD_noDict) {
2020-12-18 14:46:27 +00:00
U32 const curr = (U32)(ip - base);
U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, curr, ms->cParams.windowLog);
U32 const maxRep = curr - windowLow;
2020-02-08 13:56:59 +00:00
if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
}
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
/* dictMatchState repCode checks don't currently handle repCode == 0
* disabling. */
assert(offset_1 <= dictAndPrefixLength);
assert(offset_2 <= dictAndPrefixLength);
}
/* Match Loop */
2020-06-04 19:03:27 +00:00
#if defined(__GNUC__) && defined(__x86_64__)
/* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
* code alignment is perturbed. To fix the instability align the loop on 32-bytes.
*/
__asm__(".p2align 5");
#endif
2020-02-08 13:56:59 +00:00
while (ip < ilimit) {
size_t matchLength=0;
size_t offset=0;
const BYTE* start=ip+1;
/* check repCode */
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
2020-12-18 14:46:27 +00:00
const BYTE* repMatch = ((dictMode == ZSTD_dictMatchState || dictMode == ZSTD_dedicatedDictSearch)
2020-02-08 13:56:59 +00:00
&& repIndex < prefixLowestIndex) ?
dictBase + (repIndex - dictIndexDelta) :
base + repIndex;
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
if (depth==0) goto _storeSequence;
}
}
if ( dictMode == ZSTD_noDict
&& ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
if (depth==0) goto _storeSequence;
}
/* first search (depth 0) */
{ size_t offsetFound = 999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
if (ml2 > matchLength)
matchLength = ml2, start = ip, offset=offsetFound;
}
if (matchLength < 4) {
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
continue;
}
/* let's try to find a better solution */
if (depth>=1)
while (ip<ilimit) {
ip ++;
if ( (dictMode == ZSTD_noDict)
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
int const gain2 = (int)(mlRep * 3);
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
if ((mlRep >= 4) && (gain2 > gain1))
matchLength = mlRep, offset = 0, start = ip;
}
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
const U32 repIndex = (U32)(ip - base) - offset_1;
const BYTE* repMatch = repIndex < prefixLowestIndex ?
dictBase + (repIndex - dictIndexDelta) :
base + repIndex;
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
int const gain2 = (int)(mlRep * 3);
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
if ((mlRep >= 4) && (gain2 > gain1))
matchLength = mlRep, offset = 0, start = ip;
}
}
{ size_t offset2=999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
if ((ml2 >= 4) && (gain2 > gain1)) {
matchLength = ml2, offset = offset2, start = ip;
continue; /* search a better one */
} }
/* let's find an even better one */
if ((depth==2) && (ip<ilimit)) {
ip ++;
if ( (dictMode == ZSTD_noDict)
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
int const gain2 = (int)(mlRep * 4);
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
if ((mlRep >= 4) && (gain2 > gain1))
matchLength = mlRep, offset = 0, start = ip;
}
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
const U32 repIndex = (U32)(ip - base) - offset_1;
const BYTE* repMatch = repIndex < prefixLowestIndex ?
dictBase + (repIndex - dictIndexDelta) :
base + repIndex;
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
int const gain2 = (int)(mlRep * 4);
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
if ((mlRep >= 4) && (gain2 > gain1))
matchLength = mlRep, offset = 0, start = ip;
}
}
{ size_t offset2=999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
if ((ml2 >= 4) && (gain2 > gain1)) {
matchLength = ml2, offset = offset2, start = ip;
continue;
} } }
break; /* nothing found : store previous solution */
}
/* NOTE:
* start[-offset+ZSTD_REP_MOVE-1] is undefined behavior.
* (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which
* overflows the pointer, which is undefined behavior.
*/
/* catch up */
if (offset) {
if (dictMode == ZSTD_noDict) {
while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
&& (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */
{ start--; matchLength++; }
}
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
}
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
}
/* store sequence */
_storeSequence:
{ size_t const litLength = start - anchor;
ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
anchor = ip = start + matchLength;
}
/* check immediate repcode */
2020-12-18 14:46:27 +00:00
if (isDxS) {
2020-02-08 13:56:59 +00:00
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex = current2 - offset_2;
2020-12-18 14:46:27 +00:00
const BYTE* repMatch = repIndex < prefixLowestIndex ?
2020-02-08 13:56:59 +00:00
dictBase - dictIndexDelta + repIndex :
base + repIndex;
if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
ip += matchLength;
anchor = ip;
continue;
}
break;
}
}
if (dictMode == ZSTD_noDict) {
while ( ((ip <= ilimit) & (offset_2>0))
&& (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
/* store sequence */
matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
ip += matchLength;
anchor = ip;
continue; /* faster when present ... (?) */
} } }
/* Save reps for next block */
rep[0] = offset_1 ? offset_1 : savedOffset;
rep[1] = offset_2 ? offset_2 : savedOffset;
/* Return the last literals size */
return (size_t)(iend - anchor);
}
size_t ZSTD_compressBlock_btlazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_noDict);
}
size_t ZSTD_compressBlock_lazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_noDict);
}
size_t ZSTD_compressBlock_lazy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_noDict);
}
size_t ZSTD_compressBlock_greedy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_noDict);
}
size_t ZSTD_compressBlock_btlazy2_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_dictMatchState);
}
size_t ZSTD_compressBlock_lazy2_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dictMatchState);
}
size_t ZSTD_compressBlock_lazy_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dictMatchState);
}
size_t ZSTD_compressBlock_greedy_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dictMatchState);
}
2020-12-18 14:46:27 +00:00
size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dedicatedDictSearch);
}
size_t ZSTD_compressBlock_lazy_dedicatedDictSearch(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dedicatedDictSearch);
}
size_t ZSTD_compressBlock_greedy_dedicatedDictSearch(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dedicatedDictSearch);
}
2020-02-08 13:56:59 +00:00
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_lazy_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore,
U32 rep[ZSTD_REP_NUM],
const void* src, size_t srcSize,
const searchMethod_e searchMethod, const U32 depth)
{
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 dictLimit = ms->window.dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const dictEnd = dictBase + dictLimit;
2020-06-04 19:03:27 +00:00
const BYTE* const dictStart = dictBase + ms->window.lowLimit;
const U32 windowLog = ms->cParams.windowLog;
2020-02-08 13:56:59 +00:00
typedef size_t (*searchMax_f)(
ZSTD_matchState_t* ms,
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
searchMax_f searchMax = searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
U32 offset_1 = rep[0], offset_2 = rep[1];
2020-06-04 19:03:27 +00:00
DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic");
2020-02-08 13:56:59 +00:00
/* init */
ip += (ip == prefixStart);
/* Match Loop */
2020-06-04 19:03:27 +00:00
#if defined(__GNUC__) && defined(__x86_64__)
/* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
* code alignment is perturbed. To fix the instability align the loop on 32-bytes.
*/
__asm__(".p2align 5");
#endif
2020-02-08 13:56:59 +00:00
while (ip < ilimit) {
size_t matchLength=0;
size_t offset=0;
const BYTE* start=ip+1;
2020-12-18 14:46:27 +00:00
U32 curr = (U32)(ip-base);
2020-02-08 13:56:59 +00:00
/* check repCode */
2020-12-18 14:46:27 +00:00
{ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr+1, windowLog);
const U32 repIndex = (U32)(curr+1 - offset_1);
2020-02-08 13:56:59 +00:00
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
2020-06-04 19:03:27 +00:00
if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
2020-02-08 13:56:59 +00:00
if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
/* repcode detected we should take it */
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
if (depth==0) goto _storeSequence;
} }
/* first search (depth 0) */
{ size_t offsetFound = 999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
if (ml2 > matchLength)
matchLength = ml2, start = ip, offset=offsetFound;
}
if (matchLength < 4) {
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
continue;
}
/* let's try to find a better solution */
if (depth>=1)
while (ip<ilimit) {
ip ++;
2020-12-18 14:46:27 +00:00
curr++;
2020-02-08 13:56:59 +00:00
/* check repCode */
if (offset) {
2020-12-18 14:46:27 +00:00
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
const U32 repIndex = (U32)(curr - offset_1);
2020-02-08 13:56:59 +00:00
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
2020-06-04 19:03:27 +00:00
if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
2020-02-08 13:56:59 +00:00
if (MEM_read32(ip) == MEM_read32(repMatch)) {
/* repcode detected */
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
int const gain2 = (int)(repLength * 3);
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
if ((repLength >= 4) && (gain2 > gain1))
matchLength = repLength, offset = 0, start = ip;
} }
/* search match, depth 1 */
{ size_t offset2=999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
if ((ml2 >= 4) && (gain2 > gain1)) {
matchLength = ml2, offset = offset2, start = ip;
continue; /* search a better one */
} }
/* let's find an even better one */
if ((depth==2) && (ip<ilimit)) {
ip ++;
2020-12-18 14:46:27 +00:00
curr++;
2020-02-08 13:56:59 +00:00
/* check repCode */
if (offset) {
2020-12-18 14:46:27 +00:00
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
const U32 repIndex = (U32)(curr - offset_1);
2020-02-08 13:56:59 +00:00
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
2020-06-04 19:03:27 +00:00
if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
2020-02-08 13:56:59 +00:00
if (MEM_read32(ip) == MEM_read32(repMatch)) {
/* repcode detected */
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
int const gain2 = (int)(repLength * 4);
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
if ((repLength >= 4) && (gain2 > gain1))
matchLength = repLength, offset = 0, start = ip;
} }
/* search match, depth 2 */
{ size_t offset2=999999999;
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
if ((ml2 >= 4) && (gain2 > gain1)) {
matchLength = ml2, offset = offset2, start = ip;
continue;
} } }
break; /* nothing found : store previous solution */
}
/* catch up */
if (offset) {
U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
}
/* store sequence */
_storeSequence:
{ size_t const litLength = start - anchor;
ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
anchor = ip = start + matchLength;
}
/* check immediate repcode */
while (ip <= ilimit) {
2020-06-04 19:03:27 +00:00
const U32 repCurrent = (U32)(ip-base);
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, repCurrent, windowLog);
const U32 repIndex = repCurrent - offset_2;
2020-02-08 13:56:59 +00:00
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
const BYTE* const repMatch = repBase + repIndex;
2020-06-04 19:03:27 +00:00
if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow)) /* intentional overflow */
2020-02-08 13:56:59 +00:00
if (MEM_read32(ip) == MEM_read32(repMatch)) {
/* repcode detected we should take it */
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
ip += matchLength;
anchor = ip;
continue; /* faster when present ... (?) */
}
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_greedy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0);
}
size_t ZSTD_compressBlock_lazy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1);
}
size_t ZSTD_compressBlock_lazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2);
}
size_t ZSTD_compressBlock_btlazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2);
}