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

Optimize repcode predicate, hardcode hasStep == 0 scenario, cosmetic fixes

Browse Source
This commit is contained in:
Elliot Gorokhovsky
2022-04-20 14:40:52 -04:00
parent 2820efe7ec
commit 809f652912
1 changed files with 81 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/compress/zstd_fast.c
+43 -32
View File
@@ -587,7 +587,7 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
U32* const hashTable = ms->hashTable; U32* const hashTable = ms->hashTable;
U32 const hlog = cParams->hashLog; U32 const hlog = cParams->hashLog;
/* support stepSize of 0 */ /* support stepSize of 0 */
U32 const stepSize = cParams->targetLength + !(cParams->targetLength) + 1; size_t const stepSize = hasStep ? (cParams->targetLength + !(cParams->targetLength) + 1) : 2;
const BYTE* const base = ms->window.base; const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase; const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const istart = (const BYTE*)src; const BYTE* const istart = (const BYTE*)src;
@@ -625,14 +625,18 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
const BYTE* nextStep; const BYTE* nextStep;
const size_t kStepIncr = (1 << (kSearchStrength - 1)); const size_t kStepIncr = (1 << (kSearchStrength - 1));
(void)hasStep; /* not currently specialized on whether it's accelerated */
DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1); DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1);
/* switch to "regular" variant if extDict is invalidated due to maxDistance */ /* switch to "regular" variant if extDict is invalidated due to maxDistance */
if (prefixStartIndex == dictStartIndex) if (prefixStartIndex == dictStartIndex)
return ZSTD_compressBlock_fast(ms, seqStore, rep, src, srcSize); return ZSTD_compressBlock_fast(ms, seqStore, rep, src, srcSize);
{ U32 const curr = (U32)(ip0 - base);
U32 const maxRep = curr - dictStartIndex;
if (offset_2 >= maxRep) offset_2 = 0;
if (offset_1 >= maxRep) offset_1 = 0;
}
/* start each op */ /* start each op */
_start: /* Requires: ip0 */ _start: /* Requires: ip0 */
@@ -655,16 +659,13 @@ _start: /* Requires: ip0 */
idxBase = idx < prefixStartIndex ? dictBase : base; idxBase = idx < prefixStartIndex ? dictBase : base;
do { do {
U32 mval; /* src or dict value at match idx */ { /* load repcode match for ip[2] */
U32 const current2 = (U32)(ip2 - base);
/* load repcode match for ip[2] */ U32 const repIndex = current2 - offset_1;
const U32 current2 = (U32)(ip2 - base);
const U32 repIndex = current2 - offset_1;
const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base; const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
U32 rval; U32 rval;
assert(offset_1 > 0); if ( ((U32)(prefixStartIndex - repIndex) >= 4) /* intentional underflow */
if ( ( ((U32)(prefixStartIndex - repIndex) >= 4) /* intentional underflow */ & (offset_1 > 0) ) {
& (offset_1 < current2 - dictStartIndex) ) ) {
rval = MEM_read32(repBase + repIndex); rval = MEM_read32(repBase + repIndex);
} else { } else {
rval = MEM_read32(ip2) ^ 1; /* guaranteed to not match. */ rval = MEM_read32(ip2) ^ 1; /* guaranteed to not match. */
@@ -686,20 +687,16 @@ _start: /* Requires: ip0 */
offcode = REPCODE1_TO_OFFBASE; offcode = REPCODE1_TO_OFFBASE;
mLength += 4; mLength += 4;
goto _match; goto _match;
} } }
/* load match for ip[0] */ { /* load match for ip[0] */
if (idx >= dictStartIndex) { U32 const mval = idx >= dictStartIndex ? MEM_read32(idxBase + idx) : MEM_read32(ip0) ^ 1; /* guaranteed not to match */
mval = MEM_read32(idxBase + idx);
} else {
mval = MEM_read32(ip0) ^ 1; /* guaranteed to not match. */
}
/* check match at ip[0] */ /* check match at ip[0] */
if (MEM_read32(ip0) == mval) { if (MEM_read32(ip0) == mval) {
/* found a match! */ /* found a match! */
goto _offset; goto _offset;
} } }
/* lookup ip[1] */ /* lookup ip[1] */
idx = hashTable[hash1]; idx = hashTable[hash1];
@@ -718,18 +715,14 @@ _start: /* Requires: ip0 */
current0 = (U32)(ip0 - base); current0 = (U32)(ip0 - base);
hashTable[hash0] = current0; hashTable[hash0] = current0;
/* load match for ip[0] */ { /* load match for ip[0] */
if (idx >= dictStartIndex) { U32 const mval = idx >= dictStartIndex ? MEM_read32(idxBase + idx) : MEM_read32(ip0) ^ 1; /* guaranteed not to match */
mval = MEM_read32(idxBase + idx);
} else {
mval = MEM_read32(ip0) ^ 1; /* guaranteed to not match. */
}
/* check match at ip[0] */ /* check match at ip[0] */
if (MEM_read32(ip0) == mval) { if (MEM_read32(ip0) == mval) {
/* found a match! */ /* found a match! */
goto _offset; goto _offset;
} } }
/* lookup ip[1] */ /* lookup ip[1] */
idx = hashTable[hash1]; idx = hashTable[hash1];
@@ -760,8 +753,8 @@ _cleanup:
* them. So let's not. */ * them. So let's not. */
/* save reps for next block */ /* save reps for next block */
rep[0] = offset_1; rep[0] = offset_1 ? offset_1 : rep[0];
rep[1] = offset_2; rep[1] = offset_2 ? offset_2 : rep[1];
/* Return the last literals size */ /* Return the last literals size */
return (size_t)(iend - anchor); return (size_t)(iend - anchor);
@@ -808,11 +801,10 @@ _match: /* Requires: ip0, match0, offcode, matchEnd */
hashTable[ZSTD_hashPtr(base+current0+2, hlog, mls)] = current0+2; /* here because current+2 could be > iend-8 */ 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); hashTable[ZSTD_hashPtr(ip0-2, hlog, mls)] = (U32)(ip0-2-base);
assert(offset_2 > 0);
while (ip0 <= ilimit) { while (ip0 <= ilimit) {
U32 const repIndex2 = (U32)(ip0-base) - offset_2; U32 const repIndex2 = (U32)(ip0-base) - offset_2;
const BYTE* const repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2; const BYTE* const repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) & (offset_2 <= (U32)(ip0-base) - dictStartIndex)) /* intentional overflow */ if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) & (offset_2 > 0)) /* intentional underflow */
&& (MEM_read32(repMatch2) == MEM_read32(ip0)) ) { && (MEM_read32(repMatch2) == MEM_read32(ip0)) ) {
const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend; const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip0+4, repMatch2+4, iend, repEnd2, prefixStart) + 4; size_t const repLength2 = ZSTD_count_2segments(ip0+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
@@ -829,6 +821,11 @@ _match: /* Requires: ip0, match0, offcode, matchEnd */
goto _start; goto _start;
} }
ZSTD_GEN_FAST_FN(extDict, 4, 1)
ZSTD_GEN_FAST_FN(extDict, 5, 1)
ZSTD_GEN_FAST_FN(extDict, 6, 1)
ZSTD_GEN_FAST_FN(extDict, 7, 1)
ZSTD_GEN_FAST_FN(extDict, 4, 0) ZSTD_GEN_FAST_FN(extDict, 4, 0)
ZSTD_GEN_FAST_FN(extDict, 5, 0) ZSTD_GEN_FAST_FN(extDict, 5, 0)
ZSTD_GEN_FAST_FN(extDict, 6, 0) ZSTD_GEN_FAST_FN(extDict, 6, 0)
@@ -839,8 +836,21 @@ size_t ZSTD_compressBlock_fast_extDict(
void const* src, size_t srcSize) void const* src, size_t srcSize)
{ {
U32 const mls = ms->cParams.minMatch; U32 const mls = ms->cParams.minMatch;
switch(mls) assert(ms->dictMatchState == NULL);
{ if (ms->cParams.targetLength > 1) {
switch (mls) {
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_extDict_4_1(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_extDict_5_1(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_extDict_6_1(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_extDict_7_1(ms, seqStore, rep, src, srcSize);
}
} else {
switch (mls) {
default: /* includes case 3 */ default: /* includes case 3 */
case 4 : case 4 :
return ZSTD_compressBlock_fast_extDict_4_0(ms, seqStore, rep, src, srcSize); return ZSTD_compressBlock_fast_extDict_4_0(ms, seqStore, rep, src, srcSize);
@@ -852,3 +862,4 @@ size_t ZSTD_compressBlock_fast_extDict(
return ZSTD_compressBlock_fast_extDict_7_0(ms, seqStore, rep, src, srcSize); return ZSTD_compressBlock_fast_extDict_7_0(ms, seqStore, rep, src, srcSize);
} }
} }
}