lib/zstd
1
0
Fork 0
mirror of https://github.com/facebook/zstd.git synced 2025-08-07 06:23:00 +03:00
Code Activity

Merge pull request #4170 from facebook/dict_cSpeed

Browse Source 
Improve dictionary compression speed
This commit is contained in:
Yann Collet
2024-10-16 17:36:49 -07:00
committed by GitHub
parent def3ee9548 730d2dce41
commit 18a42190c2
3 changed files with 23 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
lib/compress/zstd_double_fast.c
View File

@@ -140,7 +140,6 @@ size_t ZSTD_compressBlock_doubleFast_noDict_generic(
U32 idxl1; /* the long match index for ip1 */ U32 idxl1; /* the long match index for ip1 */
const BYTE* matchl0; /* the long match for ip */ const BYTE* matchl0; /* the long match for ip */
const BYTE* matchl0_safe; /* matchl0 or safe address */
const BYTE* matchs0; /* the short match for ip */ const BYTE* matchs0; /* the short match for ip */
const BYTE* matchl1; /* the long match for ip1 */ const BYTE* matchl1; /* the long match for ip1 */
const BYTE* matchs0_safe; /* matchs0 or safe address */ const BYTE* matchs0_safe; /* matchs0 or safe address */
@@ -201,7 +200,7 @@ size_t ZSTD_compressBlock_doubleFast_noDict_generic(
* However expression below complies into conditional move. Since * However expression below complies into conditional move. Since
* match is unlikely and we only *branch* on idxl0 > prefixLowestIndex * match is unlikely and we only *branch* on idxl0 > prefixLowestIndex
* if there is a match, all branches become predictable. */ * if there is a match, all branches become predictable. */
matchl0_safe = ZSTD_selectAddr(idxl0, prefixLowestIndex, matchl0, &dummy[0]); { const BYTE* const matchl0_safe = ZSTD_selectAddr(idxl0, prefixLowestIndex, matchl0, &dummy[0]);
/* check prefix long match */ /* check prefix long match */
if (MEM_read64(matchl0_safe) == MEM_read64(ip) && matchl0_safe == matchl0) { if (MEM_read64(matchl0_safe) == MEM_read64(ip) && matchl0_safe == matchl0) {
@@ -209,7 +208,7 @@ size_t ZSTD_compressBlock_doubleFast_noDict_generic(
offset = (U32)(ip-matchl0); offset = (U32)(ip-matchl0);
while (((ip>anchor) & (matchl0>prefixLowest)) && (ip[-1] == matchl0[-1])) { ip--; matchl0--; mLength++; } /* catch up */ while (((ip>anchor) & (matchl0>prefixLowest)) && (ip[-1] == matchl0[-1])) { ip--; matchl0--; mLength++; } /* catch up */
goto _match_found; goto _match_found;
} } }
idxl1 = hashLong[hl1]; idxl1 = hashLong[hl1];
matchl1 = base + idxl1; matchl1 = base + idxl1;
@@ -412,14 +411,12 @@ size_t ZSTD_compressBlock_doubleFast_dictMatchState_generic(
goto _match_stored; goto _match_stored;
} }
if (matchIndexL > prefixLowestIndex) { if ((matchIndexL >= prefixLowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
/* check prefix long match */ /* check prefix long match */
if (MEM_read64(matchLong) == MEM_read64(ip)) {
mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8; mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
offset = (U32)(ip-matchLong); offset = (U32)(ip-matchLong);
while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */ while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
goto _match_found; goto _match_found;
}
} else if (dictTagsMatchL) { } else if (dictTagsMatchL) {
/* check dictMatchState long match */ /* check dictMatchState long match */
U32 const dictMatchIndexL = dictMatchIndexAndTagL >> ZSTD_SHORT_CACHE_TAG_BITS; U32 const dictMatchIndexL = dictMatchIndexAndTagL >> ZSTD_SHORT_CACHE_TAG_BITS;
@@ -434,7 +431,7 @@ size_t ZSTD_compressBlock_doubleFast_dictMatchState_generic(
} } } }
if (matchIndexS > prefixLowestIndex) { if (matchIndexS > prefixLowestIndex) {
/* check prefix short match */ /* short match candidate */
if (MEM_read32(match) == MEM_read32(ip)) { if (MEM_read32(match) == MEM_read32(ip)) {
goto _search_next_long; goto _search_next_long;
} }
@@ -464,14 +461,12 @@ _search_next_long:
hashLong[hl3] = curr + 1; hashLong[hl3] = curr + 1;
/* check prefix long +1 match */ /* check prefix long +1 match */
if (matchIndexL3 > prefixLowestIndex) { if ((matchIndexL3 >= prefixLowestIndex) && (MEM_read64(matchL3) == MEM_read64(ip+1))) {
if (MEM_read64(matchL3) == MEM_read64(ip+1)) {
mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8; mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
ip++; ip++;
offset = (U32)(ip-matchL3); offset = (U32)(ip-matchL3);
while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */ while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
goto _match_found; goto _match_found;
}
} else if (dictTagsMatchL3) { } else if (dictTagsMatchL3) {
/* check dict long +1 match */ /* check dict long +1 match */
U32 const dictMatchIndexL3 = dictMatchIndexAndTagL3 >> ZSTD_SHORT_CACHE_TAG_BITS; U32 const dictMatchIndexL3 = dictMatchIndexAndTagL3 >> ZSTD_SHORT_CACHE_TAG_BITS;

4
lib/compress/zstd_fast.c
View File

@@ -597,8 +597,8 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
} }
} }
if (matchIndex > prefixStartIndex && MEM_read32(match) == MEM_read32(ip0)) { if (ZSTD_match4Found_cmov(ip0, match, matchIndex, prefixStartIndex)) {
/* found a regular match */ /* found a regular match of size >= 4 */
U32 const offset = (U32) (ip0 - match); U32 const offset = (U32) (ip0 - match);
mLength = ZSTD_count(ip0 + 4, match + 4, iend) + 4; mLength = ZSTD_count(ip0 + 4, match + 4, iend) + 4;
while (((ip0 > anchor) & (match > prefixStart)) while (((ip0 > anchor) & (match > prefixStart))

2
tests/fuzzer.c
View File

@@ -2450,7 +2450,7 @@ static int basicUnitTests(U32 const seed, double compressibility)
3742, 3675, 3674, 3665, 3664, 3742, 3675, 3674, 3665, 3664,
3663, 3662, 3661, 3660, 3660, 3663, 3662, 3661, 3660, 3660,
3660, 3660, 3660 }; 3660, 3660, 3660 };
size_t const target_wdict_cSize[22+1] = { 2830, 2896, 2893, 2820, 2940, size_t const target_wdict_cSize[22+1] = { 2830, 2896, 2893, 2840, 2950,
2950, 2950, 2925, 2900, 2892, 2950, 2950, 2925, 2900, 2892,
2910, 2910, 2910, 2780, 2775, 2910, 2910, 2910, 2780, 2775,
2765, 2760, 2755, 2754, 2753, 2765, 2760, 2755, 2754, 2753,