facebook/zstd
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Hash Long One Position Ahead (+2.5% Speed)

Browse Source 
Aside from maybe a latency win in the loop, this means that when we find a
short match, we've already done the hash we need to check the next long match.
This commit is contained in:
W. Felix Handte 2021-09-02 12:25:08 -04:00
parent a1ac7205d0
commit db4e1b5479
1 changed files with 12 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
lib/compress/zstd_double_fast.c
View File

@ -118,10 +118,11 @@ _start:
goto _cleanup; goto _cleanup;
} }
hl0 = ZSTD_hashPtr(ip, hBitsL, 8);
/* Main Search Loop */ /* Main Search Loop */
do { do {
curr = (U32)(ip-base); curr = (U32)(ip-base);
hl0 = ZSTD_hashPtr(ip, hBitsL, 8);
hs0 = ZSTD_hashPtr(ip, hBitsS, mls); hs0 = ZSTD_hashPtr(ip, hBitsS, mls);
idxl0 = hashLong[hl0]; idxl0 = hashLong[hl0];
idxs0 = hashSmall[hs0]; idxs0 = hashSmall[hs0];
@ -148,6 +149,8 @@ _start:
} }
} }
hl1 = ZSTD_hashPtr(ip1, hBitsL, 8);
if (idxs0 > prefixLowestIndex) { if (idxs0 > prefixLowestIndex) {
/* check prefix short match */ /* check prefix short match */
if (MEM_read32(matchs0) == MEM_read32(ip)) { if (MEM_read32(matchs0) == MEM_read32(ip)) {
@ -155,18 +158,20 @@ _start:
} }
} }
if (ip >= nextStep) { if (ip1 >= nextStep) {
PREFETCH_L1(ip + 64); PREFETCH_L1(ip1 + 64);
PREFETCH_L1(ip + 128); PREFETCH_L1(ip1 + 128);
step++; step++;
nextStep += kStepIncr; nextStep += kStepIncr;
} }
ip += step; ip = ip1;
ip1 += step;
hl0 = hl1;
#if defined(__aarch64__) #if defined(__aarch64__)
PREFETCH_L1(ip+256); PREFETCH_L1(ip+256);
#endif #endif
} while (ip < ilimit); } while (ip1 < ilimit);
_cleanup: _cleanup:
/* save reps for next block */ /* save reps for next block */
@ -177,8 +182,7 @@ _cleanup:
return (size_t)(iend - anchor); return (size_t)(iend - anchor);
_search_next_long: _search_next_long:
{ hl1 = ZSTD_hashPtr(ip+1, hBitsL, 8); { idxl1 = hashLong[hl1];
idxl1 = hashLong[hl1];
matchl1 = base + idxl1; matchl1 = base + idxl1;
hashLong[hl1] = curr + 1; hashLong[hl1] = curr + 1;