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

slightly improved compression speed

Browse Source
This commit is contained in:
Yann Collet
2017-11-19 14:40:21 -08:00
parent 42c1e64270
commit 3f457264d1
3 changed files with 27 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/compress/zstd_compress_internal.h
+11 -7
View File
@@ -347,13 +347,17 @@ MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* co
const BYTE* const pStart = pIn; const BYTE* const pStart = pIn;
const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1); const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
while (pIn < pInLoopLimit) { if (pIn < pInLoopLimit) {
size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn); { size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; } if (diff) return ZSTD_NbCommonBytes(diff); }
pIn += ZSTD_NbCommonBytes(diff); pIn+=sizeof(size_t); pMatch+=sizeof(size_t);
return (size_t)(pIn - pStart); while (pIn < pInLoopLimit) {
} size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
if (MEM_64bits()) if ((pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; } if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
pIn += ZSTD_NbCommonBytes(diff);
return (size_t)(pIn - pStart);
} }
if (MEM_64bits() && (pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; } if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++; if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
return (size_t)(pIn - pStart); return (size_t)(pIn - pStart);
lib/compress/zstd_lazy.c
+14 -16
View File
@@ -85,8 +85,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) { if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
assert(matchIndex+matchLength >= dictLimit); /* might be wrong if extDict is incorrectly set to 0 */ assert(matchIndex+matchLength >= dictLimit); /* might be wrong if extDict is incorrectly set to 0 */
match = base + matchIndex; match = base + matchIndex;
if (match[matchLength] == ip[matchLength]) //if (match[matchLength] == ip[matchLength]) matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1; matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else { } else {
match = dictBase + matchIndex; match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart); matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
@@ -196,8 +196,8 @@ static size_t ZSTD_insertBtAndFindBestMatch (
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) { if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
match = base + matchIndex; match = base + matchIndex;
if (match[matchLength] == ip[matchLength]) //if (match[matchLength] == ip[matchLength]) matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1; matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else { } else {
match = dictBase + matchIndex; match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart); matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
@@ -355,14 +355,14 @@ size_t ZSTD_HcFindBestMatch_generic (
U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls); U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) { for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
const BYTE* match;
size_t currentMl=0; size_t currentMl=0;
if ((!extDict) || matchIndex >= dictLimit) { if ((!extDict) || matchIndex >= dictLimit) {
match = base + matchIndex; const BYTE* const match = base + matchIndex;
if (match[ml] == ip[ml]) /* potentially better */ if (match[ml] == ip[ml]) /* potentially better */
currentMl = ZSTD_count(ip, match, iLimit); currentMl = ZSTD_count(ip, match, iLimit);
} else { } else {
match = dictBase + matchIndex; const BYTE* const match = dictBase + matchIndex;
assert(match+4 <= dictEnd);
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */ 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; currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
} }
@@ -400,10 +400,10 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS ( FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
ZSTD_CCtx* zc, ZSTD_CCtx* const zc,
const BYTE* ip, const BYTE* const iLimit, const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr, size_t* const offsetPtr,
const U32 maxNbAttempts, const U32 matchLengthSearch) U32 const maxNbAttempts, U32 const matchLengthSearch)
{ {
switch(matchLengthSearch) switch(matchLengthSearch)
{ {
@@ -522,9 +522,8 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
*/ */
/* catch up */ /* catch up */
if (offset) { if (offset) {
while ( (start > anchor) while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > base))
&& (start > base+offset-ZSTD_REP_MOVE) && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */
&& (start[-1] == (start-offset+ZSTD_REP_MOVE)[-1]) ) /* only search for offset within prefix */
{ start--; matchLength++; } { start--; matchLength++; }
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE); offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
} }
@@ -536,9 +535,8 @@ _storeSequence:
} }
/* check immediate repcode */ /* check immediate repcode */
while ( (ip <= ilimit) while ( ((ip <= ilimit) & (offset_2>0))
&& ((offset_2>0) && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
& (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
/* store sequence */ /* store sequence */
matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4; matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */ offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
lib/compress/zstd_opt.c
+2 -3
View File
@@ -345,9 +345,8 @@ U32 ZSTD_insertBtAndGetAllMatches (
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) { if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
assert(matchIndex+matchLength >= dictLimit); /* ensure the condition is correct when !extDict */ assert(matchIndex+matchLength >= dictLimit); /* ensure the condition is correct when !extDict */
match = base + matchIndex; match = base + matchIndex;
if (match[matchLength] == ip[matchLength]) { //if (match[matchLength] == ip[matchLength]) { matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1; }
matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1; matchLength += ZSTD_count(ip+matchLength, match+matchLength, iLimit);
}
} else { } else {
match = dictBase + matchIndex; match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart); matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);