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

Merge pull request #2771 from facebook/opt_investigation

Browse Source 
Improve optimal parser performance on small data
This commit is contained in:
Yann Collet
2021-09-14 10:36:34 -07:00
committed by GitHub
parent d22bbed5db fd94b9d1c9
commit 2e6f5bc0d8
10 changed files with 257 additions and 246 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/compress/zstd_opt.c
+64 -57
View File
@@ -14,7 +14,6 @@
#define ZSTD_LITFREQ_ADD 2 /* scaling factor for litFreq, so that frequencies adapt faster to new stats */
#define ZSTD_FREQ_DIV 4 /* log factor when using previous stats to init next stats */
#define ZSTD_MAX_PRICE (1<<30)
#define ZSTD_PREDEF_THRESHOLD 1024 /* if srcSize < ZSTD_PREDEF_THRESHOLD, symbols' cost is assumed static, directly determined by pre-defined distributions */
@@ -24,11 +23,11 @@
* Price functions for optimal parser
***************************************/
#if 0 /* approximation at bit level */
#if 0 /* approximation at bit level (for tests) */
# define BITCOST_ACCURACY 0
# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
# define WEIGHT(stat) ((void)opt, ZSTD_bitWeight(stat))
#elif 0 /* fractional bit accuracy */
# define WEIGHT(stat, opt) ((void)opt, ZSTD_bitWeight(stat))
#elif 0 /* fractional bit accuracy (for tests) */
# define BITCOST_ACCURACY 8
# define BITCOST_MULTIPLIER (1 << BITCOST_ACCURACY)
# define WEIGHT(stat,opt) ((void)opt, ZSTD_fracWeight(stat))
@@ -79,25 +78,46 @@ static void ZSTD_setBasePrices(optState_t* optPtr, int optLevel)
}
/* ZSTD_downscaleStat() :
* reduce all elements in table by a factor 2^(ZSTD_FREQ_DIV+malus)
* return the resulting sum of elements */
static U32 ZSTD_downscaleStat(unsigned* table, U32 lastEltIndex, int malus)
static U32 sum_u32(const unsigned table[], size_t nbElts)
{
size_t n;
U32 total = 0;
for (n=0; n<nbElts; n++) {
total += table[n];
}
return total;
}
static U32 ZSTD_downscaleStats(unsigned* table, U32 lastEltIndex, U32 shift)
{
U32 s, sum=0;
DEBUGLOG(5, "ZSTD_downscaleStat (nbElts=%u)", (unsigned)lastEltIndex+1);
assert(ZSTD_FREQ_DIV+malus > 0 && ZSTD_FREQ_DIV+malus < 31);
DEBUGLOG(5, "ZSTD_downscaleStats (nbElts=%u, shift=%u)", (unsigned)lastEltIndex+1, (unsigned)shift);
assert(shift < 30);
for (s=0; s<lastEltIndex+1; s++) {
table[s] = 1 + (table[s] >> (ZSTD_FREQ_DIV+malus));
table[s] = 1 + (table[s] >> shift);
sum += table[s];
}
return sum;
}
/* ZSTD_scaleStats() :
* reduce all elements in table is sum too large
* return the resulting sum of elements */
static U32 ZSTD_scaleStats(unsigned* table, U32 lastEltIndex, U32 logTarget)
{
U32 const prevsum = sum_u32(table, lastEltIndex+1);
U32 const factor = prevsum >> logTarget;
DEBUGLOG(5, "ZSTD_scaleStats (nbElts=%u, target=%u)", (unsigned)lastEltIndex+1, (unsigned)logTarget);
assert(logTarget < 30);
if (factor <= 1) return prevsum;
return ZSTD_downscaleStats(table, lastEltIndex, ZSTD_highbit32(factor));
}
/* ZSTD_rescaleFreqs() :
* if first block (detected by optPtr->litLengthSum == 0) : init statistics
* take hints from dictionary if there is one
* or init from zero, using src for literals stats, or flat 1 for match symbols
* and init from zero if there is none,
* using src for literals stats, and baseline stats for sequence symbols
* otherwise downscale existing stats, to be used as seed for next block.
*/
static void
@@ -174,14 +194,18 @@ ZSTD_rescaleFreqs(optState_t* const optPtr,
if (compressedLiterals) {
unsigned lit = MaxLit;
HIST_count_simple(optPtr->litFreq, &lit, src, srcSize); /* use raw first block to init statistics */
optPtr->litSum = ZSTD_downscaleStat(optPtr->litFreq, MaxLit, 1);
optPtr->litSum = ZSTD_downscaleStats(optPtr->litFreq, MaxLit, 8);
}
{ unsigned ll;
for (ll=0; ll<=MaxLL; ll++)
optPtr->litLengthFreq[ll] = 1;
{ unsigned const baseLLfreqs[MaxLL+1] = {
4, 2, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1
};
ZSTD_memcpy(optPtr->litLengthFreq, baseLLfreqs, sizeof(baseLLfreqs)); optPtr->litLengthSum = sum_u32(baseLLfreqs, MaxLL+1);
}
optPtr->litLengthSum = MaxLL+1;
{ unsigned ml;
for (ml=0; ml<=MaxML; ml++)
@@ -189,21 +213,25 @@ ZSTD_rescaleFreqs(optState_t* const optPtr,
}
optPtr->matchLengthSum = MaxML+1;
{ unsigned of;
for (of=0; of<=MaxOff; of++)
optPtr->offCodeFreq[of] = 1;
{ unsigned const baseOFCfreqs[MaxOff+1] = {
6, 2, 1, 1, 2, 3, 4, 4,
4, 3, 2, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1
};
ZSTD_memcpy(optPtr->offCodeFreq, baseOFCfreqs, sizeof(baseOFCfreqs)); optPtr->offCodeSum = sum_u32(baseOFCfreqs, MaxOff+1);
}
optPtr->offCodeSum = MaxOff+1;
}
} else { /* new block : re-use previous statistics, scaled down */
if (compressedLiterals)
optPtr->litSum = ZSTD_downscaleStat(optPtr->litFreq, MaxLit, 1);
optPtr->litLengthSum = ZSTD_downscaleStat(optPtr->litLengthFreq, MaxLL, 0);
optPtr->matchLengthSum = ZSTD_downscaleStat(optPtr->matchLengthFreq, MaxML, 0);
optPtr->offCodeSum = ZSTD_downscaleStat(optPtr->offCodeFreq, MaxOff, 0);
optPtr->litSum = ZSTD_scaleStats(optPtr->litFreq, MaxLit, 12);
optPtr->litLengthSum = ZSTD_scaleStats(optPtr->litLengthFreq, MaxLL, 11);
optPtr->matchLengthSum = ZSTD_scaleStats(optPtr->matchLengthFreq, MaxML, 11);
optPtr->offCodeSum = ZSTD_scaleStats(optPtr->offCodeFreq, MaxOff, 11);
}
ZSTD_setBasePrices(optPtr, optLevel);
@@ -901,11 +929,11 @@ static void ZSTD_optLdm_processMatchCandidate(ZSTD_optLdm_t* optLdm, ZSTD_match_
ZSTD_optLdm_maybeAddMatch(matches, nbMatches, optLdm, currPosInBlock);
}
/*-*******************************
* Optimal parser
*********************************/
static U32 ZSTD_totalLen(ZSTD_optimal_t sol)
{
return sol.litlen + sol.mlen;
@@ -987,7 +1015,7 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
* in every price. We include the literal length to avoid negative
* prices when we subtract the previous literal length.
*/
opt[0].price = ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
opt[0].price = (int)ZSTD_litLengthPrice(litlen, optStatePtr, optLevel);
/* large match -> immediate encoding */
{ U32 const maxML = matches[nbMatches-1].len;
@@ -1007,7 +1035,8 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
} }
/* set prices for first matches starting position == 0 */
{ U32 const literalsPrice = opt[0].price + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
assert(opt[0].price >= 0);
{ U32 const literalsPrice = (U32)opt[0].price + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
U32 pos;
U32 matchNb;
for (pos = 1; pos < minMatch; pos++) {
@@ -1024,7 +1053,7 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
opt[pos].mlen = pos;
opt[pos].off = offset;
opt[pos].litlen = litlen;
opt[pos].price = sequencePrice;
opt[pos].price = (int)sequencePrice;
} }
last_pos = pos-1;
}
@@ -1039,9 +1068,9 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
/* Fix current position with one literal if cheaper */
{ U32 const litlen = (opt[cur-1].mlen == 0) ? opt[cur-1].litlen + 1 : 1;
int const price = opt[cur-1].price
+ ZSTD_rawLiteralsCost(ip+cur-1, 1, optStatePtr, optLevel)
+ ZSTD_litLengthPrice(litlen, optStatePtr, optLevel)
- ZSTD_litLengthPrice(litlen-1, optStatePtr, optLevel);
+ (int)ZSTD_rawLiteralsCost(ip+cur-1, 1, optStatePtr, optLevel)
+ (int)ZSTD_litLengthPrice(litlen, optStatePtr, optLevel)
- (int)ZSTD_litLengthPrice(litlen-1, optStatePtr, optLevel);
assert(price < 1000000000); /* overflow check */
if (price <= opt[cur].price) {
DEBUGLOG(7, "cPos:%zi==rPos:%u : better price (%.2f<=%.2f) using literal (ll==%u) (hist:%u,%u,%u)",
@@ -1084,9 +1113,10 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
continue; /* skip unpromising positions; about ~+6% speed, -0.01 ratio */
}
assert(opt[cur].price >= 0);
{ U32 const ll0 = (opt[cur].mlen != 0);
U32 const litlen = (opt[cur].mlen == 0) ? opt[cur].litlen : 0;
U32 const previousPrice = opt[cur].price;
U32 const previousPrice = (U32)opt[cur].price;
U32 const basePrice = previousPrice + ZSTD_litLengthPrice(0, optStatePtr, optLevel);
U32 nbMatches = ZSTD_BtGetAllMatches(matches, ms, &nextToUpdate3, inr, iend, dictMode, opt[cur].rep, ll0, minMatch);
U32 matchNb;
@@ -1126,7 +1156,7 @@ ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,
for (mlen = lastML; mlen >= startML; mlen--) { /* scan downward */
U32 const pos = cur + mlen;
int const price = basePrice + ZSTD_getMatchPrice(offset, mlen, optStatePtr, optLevel);
int const price = (int)basePrice + (int)ZSTD_getMatchPrice(offset, mlen, optStatePtr, optLevel);
if ((pos > last_pos) || (price < opt[pos].price)) {
DEBUGLOG(7, "rPos:%u (ml=%2u) => new better price (%.2f<%.2f)",
@@ -1222,28 +1252,7 @@ size_t ZSTD_compressBlock_btopt(
}
/* used in 2-pass strategy */
static U32 ZSTD_upscaleStat(unsigned* table, U32 lastEltIndex, int bonus)
{
U32 s, sum=0;
assert(ZSTD_FREQ_DIV+bonus >= 0);
for (s=0; s<lastEltIndex+1; s++) {
table[s] <<= ZSTD_FREQ_DIV+bonus;
table[s]--;
sum += table[s];
}
return sum;
}
/* used in 2-pass strategy */
MEM_STATIC void ZSTD_upscaleStats(optState_t* optPtr)
{
if (ZSTD_compressedLiterals(optPtr))
optPtr->litSum = ZSTD_upscaleStat(optPtr->litFreq, MaxLit, 0);
optPtr->litLengthSum = ZSTD_upscaleStat(optPtr->litLengthFreq, MaxLL, 0);
optPtr->matchLengthSum = ZSTD_upscaleStat(optPtr->matchLengthFreq, MaxML, 0);
optPtr->offCodeSum = ZSTD_upscaleStat(optPtr->offCodeFreq, MaxOff, 0);
}
/* ZSTD_initStats_ultra():
* make a first compression pass, just to seed stats with more accurate starting values.
@@ -1274,8 +1283,6 @@ ZSTD_initStats_ultra(ZSTD_matchState_t* ms,
ms->window.lowLimit = ms->window.dictLimit;
ms->nextToUpdate = ms->window.dictLimit;
/* re-inforce weight of collected statistics */
ZSTD_upscaleStats(&ms->opt);
}
size_t ZSTD_compressBlock_btultra(