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

Merge pull request #2921 from felixhandte/neg-lvl-stagger-step

Browse Source 
Stagger Stepping in Negative Levels
This commit is contained in:
Felix Handte
2021-12-14 14:13:57 -05:00
committed by GitHub
parent 3e2a70b6fb 450fca9704
commit 5e2fede604
3 changed files with 193 additions and 164 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/compress/zstd_fast.c
+80 -51
View File
@@ -93,13 +93,13 @@ FORCE_INLINE_TEMPLATE size_t
ZSTD_compressBlock_fast_noDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const mls)
U32 const mls, U32 const hasStep)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
U32 const hlog = cParams->hashLog;
/* support stepSize of 0 */
size_t const stepSize = cParams->targetLength + !(cParams->targetLength);
size_t const stepSize = hasStep ? (cParams->targetLength + !(cParams->targetLength) + 1) : 2;
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
@@ -128,6 +128,10 @@ ZSTD_compressBlock_fast_noDict_generic(
const BYTE* match0;
size_t mLength;
/* ip0 and ip1 are always adjacent. The targetLength skipping and
* uncompressibility acceleration is applied to every other position,
* matching the behavior of #1562. step therefore represents the gap
* between pairs of positions, from ip0 to ip2 or ip1 to ip3. */
size_t step;
const BYTE* nextStep;
const size_t kStepIncr = (1 << (kSearchStrength - 1));
@@ -148,9 +152,9 @@ _start: /* Requires: ip0 */
nextStep = ip0 + kStepIncr;
/* calculate positions, ip0 - anchor == 0, so we skip step calc */
ip1 = ip0 + stepSize;
ip2 = ip1 + stepSize;
ip3 = ip2 + stepSize;
ip1 = ip0 + 1;
ip2 = ip0 + step;
ip3 = ip2 + 1;
if (ip3 >= ilimit) {
goto _cleanup;
@@ -230,19 +234,19 @@ _start: /* Requires: ip0 */
hash0 = hash1;
hash1 = ZSTD_hashPtr(ip2, hlog, mls);
/* calculate step */
if (ip2 >= nextStep) {
PREFETCH_L1(ip1 + 64);
PREFETCH_L1(ip1 + 128);
step++;
nextStep += kStepIncr;
}
/* advance to next positions */
ip0 = ip1;
ip1 = ip2;
ip2 = ip2 + step;
ip3 = ip2 + step;
ip2 = ip0 + step;
ip3 = ip1 + step;
/* calculate step */
if (ip2 >= nextStep) {
step++;
PREFETCH_L1(ip1 + 64);
PREFETCH_L1(ip1 + 128);
nextStep += kStepIncr;
}
} while (ip3 < ilimit);
_cleanup:
@@ -310,18 +314,23 @@ _match: /* Requires: ip0, match0, offcode */
goto _start;
}
#define ZSTD_GEN_FAST_FN(dictMode, mls) \
static size_t ZSTD_compressBlock_fast_##dictMode##_##mls( \
#define ZSTD_GEN_FAST_FN(dictMode, mls, step) \
static size_t ZSTD_compressBlock_fast_##dictMode##_##mls##_##step( \
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], \
void const* src, size_t srcSize) \
{ \
return ZSTD_compressBlock_fast_##dictMode##_generic(ms, seqStore, rep, src, srcSize, mls); \
return ZSTD_compressBlock_fast_##dictMode##_generic(ms, seqStore, rep, src, srcSize, mls, step); \
}
ZSTD_GEN_FAST_FN(noDict, 4)
ZSTD_GEN_FAST_FN(noDict, 5)
ZSTD_GEN_FAST_FN(noDict, 6)
ZSTD_GEN_FAST_FN(noDict, 7)
ZSTD_GEN_FAST_FN(noDict, 4, 1)
ZSTD_GEN_FAST_FN(noDict, 5, 1)
ZSTD_GEN_FAST_FN(noDict, 6, 1)
ZSTD_GEN_FAST_FN(noDict, 7, 1)
ZSTD_GEN_FAST_FN(noDict, 4, 0)
ZSTD_GEN_FAST_FN(noDict, 5, 0)
ZSTD_GEN_FAST_FN(noDict, 6, 0)
ZSTD_GEN_FAST_FN(noDict, 7, 0)
size_t ZSTD_compressBlock_fast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -329,24 +338,40 @@ size_t ZSTD_compressBlock_fast(
{
U32 const mls = ms->cParams.minMatch;
assert(ms->dictMatchState == NULL);
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_noDict_4(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_noDict_5(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_noDict_6(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_noDict_7(ms, seqStore, rep, src, srcSize);
if (ms->cParams.targetLength > 1) {
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_noDict_4_1(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_noDict_5_1(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_noDict_6_1(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_noDict_7_1(ms, seqStore, rep, src, srcSize);
}
} else {
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_noDict_4_0(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_noDict_5_0(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_noDict_6_0(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_noDict_7_0(ms, seqStore, rep, src, srcSize);
}
}
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_fast_dictMatchState_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize, U32 const mls)
void const* src, size_t srcSize, U32 const mls, U32 const hasStep)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
@@ -382,6 +407,8 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
assert(endIndex - prefixStartIndex <= maxDistance);
(void)maxDistance; (void)endIndex; /* these variables are not used when assert() is disabled */
(void)hasStep; /* not currently specialized on whether it's accelerated */
/* ensure there will be no underflow
* when translating a dict index into a local index */
assert(prefixStartIndex >= (U32)(dictEnd - dictBase));
@@ -492,10 +519,10 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
}
ZSTD_GEN_FAST_FN(dictMatchState, 4)
ZSTD_GEN_FAST_FN(dictMatchState, 5)
ZSTD_GEN_FAST_FN(dictMatchState, 6)
ZSTD_GEN_FAST_FN(dictMatchState, 7)
ZSTD_GEN_FAST_FN(dictMatchState, 4, 0)
ZSTD_GEN_FAST_FN(dictMatchState, 5, 0)
ZSTD_GEN_FAST_FN(dictMatchState, 6, 0)
ZSTD_GEN_FAST_FN(dictMatchState, 7, 0)
size_t ZSTD_compressBlock_fast_dictMatchState(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -507,20 +534,20 @@ size_t ZSTD_compressBlock_fast_dictMatchState(
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_dictMatchState_4(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_dictMatchState_4_0(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_dictMatchState_5(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_dictMatchState_5_0(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_dictMatchState_6(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_dictMatchState_6_0(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_dictMatchState_7(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_dictMatchState_7_0(ms, seqStore, rep, src, srcSize);
}
}
static size_t ZSTD_compressBlock_fast_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize, U32 const mls)
void const* src, size_t srcSize, U32 const mls, U32 const hasStep)
{
const ZSTD_compressionParameters* const cParams = &ms->cParams;
U32* const hashTable = ms->hashTable;
@@ -544,6 +571,8 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
const BYTE* const ilimit = iend - 8;
U32 offset_1=rep[0], offset_2=rep[1];
(void)hasStep; /* not currently specialized on whether it's accelerated */
DEBUGLOG(5, "ZSTD_compressBlock_fast_extDict_generic (offset_1=%u)", offset_1);
/* switch to "regular" variant if extDict is invalidated due to maxDistance */
@@ -621,10 +650,10 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
return (size_t)(iend - anchor);
}
ZSTD_GEN_FAST_FN(extDict, 4)
ZSTD_GEN_FAST_FN(extDict, 5)
ZSTD_GEN_FAST_FN(extDict, 6)
ZSTD_GEN_FAST_FN(extDict, 7)
ZSTD_GEN_FAST_FN(extDict, 4, 0)
ZSTD_GEN_FAST_FN(extDict, 5, 0)
ZSTD_GEN_FAST_FN(extDict, 6, 0)
ZSTD_GEN_FAST_FN(extDict, 7, 0)
size_t ZSTD_compressBlock_fast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -635,12 +664,12 @@ size_t ZSTD_compressBlock_fast_extDict(
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_extDict_4(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_extDict_4_0(ms, seqStore, rep, src, srcSize);
case 5 :
return ZSTD_compressBlock_fast_extDict_5(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_extDict_5_0(ms, seqStore, rep, src, srcSize);
case 6 :
return ZSTD_compressBlock_fast_extDict_6(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_extDict_6_0(ms, seqStore, rep, src, srcSize);
case 7 :
return ZSTD_compressBlock_fast_extDict_7(ms, seqStore, rep, src, srcSize);
return ZSTD_compressBlock_fast_extDict_7_0(ms, seqStore, rep, src, srcSize);
}
}