extzstd 0.2 → 0.3
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/HISTORY.ja.md +13 -0
- data/README.md +17 -14
- data/contrib/zstd/{NEWS → CHANGELOG} +115 -2
- data/contrib/zstd/CODE_OF_CONDUCT.md +5 -0
- data/contrib/zstd/Makefile +99 -53
- data/contrib/zstd/README.md +59 -39
- data/contrib/zstd/TESTING.md +1 -1
- data/contrib/zstd/appveyor.yml +17 -6
- data/contrib/zstd/lib/BUCK +29 -2
- data/contrib/zstd/lib/Makefile +118 -21
- data/contrib/zstd/lib/README.md +84 -44
- data/contrib/zstd/lib/common/bitstream.h +17 -33
- data/contrib/zstd/lib/common/compiler.h +62 -8
- data/contrib/zstd/lib/common/cpu.h +215 -0
- data/contrib/zstd/lib/common/debug.c +44 -0
- data/contrib/zstd/lib/common/debug.h +134 -0
- data/contrib/zstd/lib/common/entropy_common.c +16 -1
- data/contrib/zstd/lib/common/error_private.c +7 -0
- data/contrib/zstd/lib/common/fse.h +48 -44
- data/contrib/zstd/lib/common/fse_decompress.c +3 -3
- data/contrib/zstd/lib/common/huf.h +169 -113
- data/contrib/zstd/lib/common/mem.h +20 -2
- data/contrib/zstd/lib/common/pool.c +135 -49
- data/contrib/zstd/lib/common/pool.h +40 -21
- data/contrib/zstd/lib/common/threading.c +2 -2
- data/contrib/zstd/lib/common/threading.h +12 -12
- data/contrib/zstd/lib/common/xxhash.c +3 -2
- data/contrib/zstd/lib/common/zstd_common.c +3 -6
- data/contrib/zstd/lib/common/zstd_errors.h +17 -7
- data/contrib/zstd/lib/common/zstd_internal.h +76 -48
- data/contrib/zstd/lib/compress/fse_compress.c +89 -209
- data/contrib/zstd/lib/compress/hist.c +203 -0
- data/contrib/zstd/lib/compress/hist.h +95 -0
- data/contrib/zstd/lib/compress/huf_compress.c +188 -80
- data/contrib/zstd/lib/compress/zstd_compress.c +2500 -1203
- data/contrib/zstd/lib/compress/zstd_compress_internal.h +463 -62
- data/contrib/zstd/lib/compress/zstd_double_fast.c +321 -131
- data/contrib/zstd/lib/compress/zstd_double_fast.h +13 -4
- data/contrib/zstd/lib/compress/zstd_fast.c +335 -108
- data/contrib/zstd/lib/compress/zstd_fast.h +12 -6
- data/contrib/zstd/lib/compress/zstd_lazy.c +654 -313
- data/contrib/zstd/lib/compress/zstd_lazy.h +44 -16
- data/contrib/zstd/lib/compress/zstd_ldm.c +310 -420
- data/contrib/zstd/lib/compress/zstd_ldm.h +63 -26
- data/contrib/zstd/lib/compress/zstd_opt.c +773 -325
- data/contrib/zstd/lib/compress/zstd_opt.h +31 -5
- data/contrib/zstd/lib/compress/zstdmt_compress.c +1468 -518
- data/contrib/zstd/lib/compress/zstdmt_compress.h +96 -45
- data/contrib/zstd/lib/decompress/huf_decompress.c +518 -282
- data/contrib/zstd/lib/decompress/zstd_ddict.c +240 -0
- data/contrib/zstd/lib/decompress/zstd_ddict.h +44 -0
- data/contrib/zstd/lib/decompress/zstd_decompress.c +613 -1513
- data/contrib/zstd/lib/decompress/zstd_decompress_block.c +1311 -0
- data/contrib/zstd/lib/decompress/zstd_decompress_block.h +59 -0
- data/contrib/zstd/lib/decompress/zstd_decompress_internal.h +175 -0
- data/contrib/zstd/lib/dictBuilder/cover.c +194 -113
- data/contrib/zstd/lib/dictBuilder/cover.h +112 -0
- data/contrib/zstd/lib/dictBuilder/divsufsort.c +3 -3
- data/contrib/zstd/lib/dictBuilder/fastcover.c +740 -0
- data/contrib/zstd/lib/dictBuilder/zdict.c +142 -106
- data/contrib/zstd/lib/dictBuilder/zdict.h +115 -49
- data/contrib/zstd/lib/legacy/zstd_legacy.h +44 -12
- data/contrib/zstd/lib/legacy/zstd_v01.c +41 -10
- data/contrib/zstd/lib/legacy/zstd_v01.h +12 -7
- data/contrib/zstd/lib/legacy/zstd_v02.c +37 -12
- data/contrib/zstd/lib/legacy/zstd_v02.h +12 -7
- data/contrib/zstd/lib/legacy/zstd_v03.c +38 -12
- data/contrib/zstd/lib/legacy/zstd_v03.h +12 -7
- data/contrib/zstd/lib/legacy/zstd_v04.c +55 -174
- data/contrib/zstd/lib/legacy/zstd_v04.h +12 -7
- data/contrib/zstd/lib/legacy/zstd_v05.c +59 -31
- data/contrib/zstd/lib/legacy/zstd_v05.h +12 -7
- data/contrib/zstd/lib/legacy/zstd_v06.c +48 -20
- data/contrib/zstd/lib/legacy/zstd_v06.h +10 -5
- data/contrib/zstd/lib/legacy/zstd_v07.c +62 -29
- data/contrib/zstd/lib/legacy/zstd_v07.h +10 -5
- data/contrib/zstd/lib/zstd.h +1346 -832
- data/ext/extzstd.c +27 -19
- data/ext/extzstd_stream.c +20 -4
- data/ext/zstd_compress.c +1 -0
- data/ext/zstd_decompress.c +4 -0
- data/ext/zstd_dictbuilder.c +4 -0
- data/ext/zstd_dictbuilder_fastcover.c +5 -0
- data/lib/extzstd.rb +52 -220
- data/lib/extzstd/version.rb +1 -1
- metadata +21 -7
- data/contrib/zstd/circle.yml +0 -63
@@ -16,13 +16,19 @@ extern "C" {
|
|
16
16
|
#endif
|
17
17
|
|
18
18
|
#include "mem.h" /* U32 */
|
19
|
-
#include "
|
19
|
+
#include "zstd_compress_internal.h"
|
20
20
|
|
21
|
-
void ZSTD_fillHashTable(
|
22
|
-
|
23
|
-
|
24
|
-
|
25
|
-
|
21
|
+
void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
|
22
|
+
void const* end, ZSTD_dictTableLoadMethod_e dtlm);
|
23
|
+
size_t ZSTD_compressBlock_fast(
|
24
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
25
|
+
void const* src, size_t srcSize);
|
26
|
+
size_t ZSTD_compressBlock_fast_dictMatchState(
|
27
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
28
|
+
void const* src, size_t srcSize);
|
29
|
+
size_t ZSTD_compressBlock_fast_extDict(
|
30
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
31
|
+
void const* src, size_t srcSize);
|
26
32
|
|
27
33
|
#if defined (__cplusplus)
|
28
34
|
}
|
@@ -15,89 +15,108 @@
|
|
15
15
|
/*-*************************************
|
16
16
|
* Binary Tree search
|
17
17
|
***************************************/
|
18
|
-
|
19
|
-
|
20
|
-
|
21
|
-
|
22
|
-
|
23
|
-
U32 nbCompares, U32 const mls, U32 const extDict)
|
18
|
+
|
19
|
+
static void
|
20
|
+
ZSTD_updateDUBT(ZSTD_matchState_t* ms,
|
21
|
+
const BYTE* ip, const BYTE* iend,
|
22
|
+
U32 mls)
|
24
23
|
{
|
25
|
-
|
26
|
-
U32
|
27
|
-
|
28
|
-
|
29
|
-
U32
|
30
|
-
U32
|
31
|
-
U32
|
24
|
+
const ZSTD_compressionParameters* const cParams = &ms->cParams;
|
25
|
+
U32* const hashTable = ms->hashTable;
|
26
|
+
U32 const hashLog = cParams->hashLog;
|
27
|
+
|
28
|
+
U32* const bt = ms->chainTable;
|
29
|
+
U32 const btLog = cParams->chainLog - 1;
|
30
|
+
U32 const btMask = (1 << btLog) - 1;
|
31
|
+
|
32
|
+
const BYTE* const base = ms->window.base;
|
33
|
+
U32 const target = (U32)(ip - base);
|
34
|
+
U32 idx = ms->nextToUpdate;
|
35
|
+
|
36
|
+
if (idx != target)
|
37
|
+
DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
|
38
|
+
idx, target, ms->window.dictLimit);
|
39
|
+
assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */
|
40
|
+
(void)iend;
|
41
|
+
|
42
|
+
assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */
|
43
|
+
for ( ; idx < target ; idx++) {
|
44
|
+
size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */
|
45
|
+
U32 const matchIndex = hashTable[h];
|
46
|
+
|
47
|
+
U32* const nextCandidatePtr = bt + 2*(idx&btMask);
|
48
|
+
U32* const sortMarkPtr = nextCandidatePtr + 1;
|
49
|
+
|
50
|
+
DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
|
51
|
+
hashTable[h] = idx; /* Update Hash Table */
|
52
|
+
*nextCandidatePtr = matchIndex; /* update BT like a chain */
|
53
|
+
*sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
|
54
|
+
}
|
55
|
+
ms->nextToUpdate = target;
|
56
|
+
}
|
57
|
+
|
58
|
+
|
59
|
+
/** ZSTD_insertDUBT1() :
|
60
|
+
* sort one already inserted but unsorted position
|
61
|
+
* assumption : current >= btlow == (current - btmask)
|
62
|
+
* doesn't fail */
|
63
|
+
static void
|
64
|
+
ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
|
65
|
+
U32 current, const BYTE* inputEnd,
|
66
|
+
U32 nbCompares, U32 btLow,
|
67
|
+
const ZSTD_dictMode_e dictMode)
|
68
|
+
{
|
69
|
+
const ZSTD_compressionParameters* const cParams = &ms->cParams;
|
70
|
+
U32* const bt = ms->chainTable;
|
71
|
+
U32 const btLog = cParams->chainLog - 1;
|
72
|
+
U32 const btMask = (1 << btLog) - 1;
|
32
73
|
size_t commonLengthSmaller=0, commonLengthLarger=0;
|
33
|
-
const BYTE* const base =
|
34
|
-
const BYTE* const dictBase =
|
35
|
-
const U32 dictLimit =
|
74
|
+
const BYTE* const base = ms->window.base;
|
75
|
+
const BYTE* const dictBase = ms->window.dictBase;
|
76
|
+
const U32 dictLimit = ms->window.dictLimit;
|
77
|
+
const BYTE* const ip = (current>=dictLimit) ? base + current : dictBase + current;
|
78
|
+
const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit;
|
36
79
|
const BYTE* const dictEnd = dictBase + dictLimit;
|
37
80
|
const BYTE* const prefixStart = base + dictLimit;
|
38
81
|
const BYTE* match;
|
39
|
-
const U32 current = (U32)(ip-base);
|
40
|
-
const U32 btLow = btMask >= current ? 0 : current - btMask;
|
41
82
|
U32* smallerPtr = bt + 2*(current&btMask);
|
42
83
|
U32* largerPtr = smallerPtr + 1;
|
84
|
+
U32 matchIndex = *smallerPtr; /* this candidate is unsorted : next sorted candidate is reached through *smallerPtr, while *largerPtr contains previous unsorted candidate (which is already saved and can be overwritten) */
|
43
85
|
U32 dummy32; /* to be nullified at the end */
|
44
|
-
U32 const windowLow =
|
45
|
-
U32 matchEndIdx = current+8+1;
|
46
|
-
size_t bestLength = 8;
|
47
|
-
#ifdef ZSTD_C_PREDICT
|
48
|
-
U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
|
49
|
-
U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
|
50
|
-
predictedSmall += (predictedSmall>0);
|
51
|
-
predictedLarge += (predictedLarge>0);
|
52
|
-
#endif /* ZSTD_C_PREDICT */
|
53
|
-
|
54
|
-
DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
|
86
|
+
U32 const windowLow = ms->window.lowLimit;
|
55
87
|
|
56
|
-
|
57
|
-
|
88
|
+
DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
|
89
|
+
current, dictLimit, windowLow);
|
90
|
+
assert(current >= btLow);
|
91
|
+
assert(ip < iend); /* condition for ZSTD_count */
|
58
92
|
|
59
93
|
while (nbCompares-- && (matchIndex > windowLow)) {
|
60
94
|
U32* const nextPtr = bt + 2*(matchIndex & btMask);
|
61
95
|
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
|
62
96
|
assert(matchIndex < current);
|
63
|
-
|
64
|
-
|
65
|
-
|
66
|
-
|
67
|
-
|
68
|
-
|
69
|
-
|
70
|
-
|
71
|
-
|
72
|
-
|
73
|
-
|
74
|
-
|
75
|
-
|
76
|
-
*largerPtr = matchIndex;
|
77
|
-
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
|
78
|
-
largerPtr = nextPtr;
|
79
|
-
matchIndex = nextPtr[0];
|
80
|
-
predictedLarge = predictPtr[0] + (predictPtr[0]>0);
|
81
|
-
continue;
|
82
|
-
}
|
83
|
-
#endif
|
84
|
-
|
85
|
-
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
|
86
|
-
assert(matchIndex+matchLength >= dictLimit); /* might be wrong if extDict is incorrectly set to 0 */
|
87
|
-
match = base + matchIndex;
|
97
|
+
/* note : all candidates are now supposed sorted,
|
98
|
+
* but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
|
99
|
+
* when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
|
100
|
+
|
101
|
+
if ( (dictMode != ZSTD_extDict)
|
102
|
+
|| (matchIndex+matchLength >= dictLimit) /* both in current segment*/
|
103
|
+
|| (current < dictLimit) /* both in extDict */) {
|
104
|
+
const BYTE* const mBase = ( (dictMode != ZSTD_extDict)
|
105
|
+
|| (matchIndex+matchLength >= dictLimit)) ?
|
106
|
+
base : dictBase;
|
107
|
+
assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */
|
108
|
+
|| (current < dictLimit) );
|
109
|
+
match = mBase + matchIndex;
|
88
110
|
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
|
89
111
|
} else {
|
90
112
|
match = dictBase + matchIndex;
|
91
113
|
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
|
92
114
|
if (matchIndex+matchLength >= dictLimit)
|
93
|
-
match = base + matchIndex; /*
|
115
|
+
match = base + matchIndex; /* preparation for next read of match[matchLength] */
|
94
116
|
}
|
95
117
|
|
96
|
-
|
97
|
-
|
98
|
-
if (matchLength > matchEndIdx - matchIndex)
|
99
|
-
matchEndIdx = matchIndex + (U32)matchLength;
|
100
|
-
}
|
118
|
+
DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
|
119
|
+
current, matchIndex, (U32)matchLength);
|
101
120
|
|
102
121
|
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
|
103
122
|
break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
|
@@ -108,6 +127,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
|
|
108
127
|
*smallerPtr = matchIndex; /* update smaller idx */
|
109
128
|
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
|
110
129
|
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
|
130
|
+
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
|
131
|
+
matchIndex, btLow, nextPtr[1]);
|
111
132
|
smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
|
112
133
|
matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
|
113
134
|
} else {
|
@@ -115,184 +136,301 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
|
|
115
136
|
*largerPtr = matchIndex;
|
116
137
|
commonLengthLarger = matchLength;
|
117
138
|
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
|
139
|
+
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
|
140
|
+
matchIndex, btLow, nextPtr[0]);
|
118
141
|
largerPtr = nextPtr;
|
119
142
|
matchIndex = nextPtr[0];
|
120
143
|
} }
|
121
144
|
|
122
145
|
*smallerPtr = *largerPtr = 0;
|
123
|
-
if (bestLength > 384) return MIN(192, (U32)(bestLength - 384)); /* speed optimization */
|
124
|
-
assert(matchEndIdx > current + 8);
|
125
|
-
return matchEndIdx - (current + 8);
|
126
146
|
}
|
127
147
|
|
128
|
-
FORCE_INLINE_TEMPLATE
|
129
|
-
void ZSTD_updateTree_internal(ZSTD_CCtx* zc,
|
130
|
-
const BYTE* const ip, const BYTE* const iend,
|
131
|
-
const U32 nbCompares, const U32 mls, const U32 extDict)
|
132
|
-
{
|
133
|
-
const BYTE* const base = zc->base;
|
134
|
-
U32 const target = (U32)(ip - base);
|
135
|
-
U32 idx = zc->nextToUpdate;
|
136
|
-
DEBUGLOG(7, "ZSTD_updateTree_internal, from %u to %u (extDict:%u)",
|
137
|
-
idx, target, extDict);
|
138
148
|
|
139
|
-
|
140
|
-
|
141
|
-
|
142
|
-
|
143
|
-
|
144
|
-
|
145
|
-
|
146
|
-
|
149
|
+
static size_t
|
150
|
+
ZSTD_DUBT_findBetterDictMatch (
|
151
|
+
ZSTD_matchState_t* ms,
|
152
|
+
const BYTE* const ip, const BYTE* const iend,
|
153
|
+
size_t* offsetPtr,
|
154
|
+
size_t bestLength,
|
155
|
+
U32 nbCompares,
|
156
|
+
U32 const mls,
|
157
|
+
const ZSTD_dictMode_e dictMode)
|
147
158
|
{
|
148
|
-
|
149
|
-
|
150
|
-
|
151
|
-
|
152
|
-
|
153
|
-
|
154
|
-
|
155
|
-
|
156
|
-
|
159
|
+
const ZSTD_matchState_t * const dms = ms->dictMatchState;
|
160
|
+
const ZSTD_compressionParameters* const dmsCParams = &dms->cParams;
|
161
|
+
const U32 * const dictHashTable = dms->hashTable;
|
162
|
+
U32 const hashLog = dmsCParams->hashLog;
|
163
|
+
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
|
164
|
+
U32 dictMatchIndex = dictHashTable[h];
|
165
|
+
|
166
|
+
const BYTE* const base = ms->window.base;
|
167
|
+
const BYTE* const prefixStart = base + ms->window.dictLimit;
|
168
|
+
U32 const current = (U32)(ip-base);
|
169
|
+
const BYTE* const dictBase = dms->window.base;
|
170
|
+
const BYTE* const dictEnd = dms->window.nextSrc;
|
171
|
+
U32 const dictHighLimit = (U32)(dms->window.nextSrc - dms->window.base);
|
172
|
+
U32 const dictLowLimit = dms->window.lowLimit;
|
173
|
+
U32 const dictIndexDelta = ms->window.lowLimit - dictHighLimit;
|
174
|
+
|
175
|
+
U32* const dictBt = dms->chainTable;
|
176
|
+
U32 const btLog = dmsCParams->chainLog - 1;
|
177
|
+
U32 const btMask = (1 << btLog) - 1;
|
178
|
+
U32 const btLow = (btMask >= dictHighLimit - dictLowLimit) ? dictLowLimit : dictHighLimit - btMask;
|
157
179
|
|
158
|
-
|
159
|
-
static size_t ZSTD_insertBtAndFindBestMatch (
|
160
|
-
ZSTD_CCtx* zc,
|
161
|
-
const BYTE* const ip, const BYTE* const iend,
|
162
|
-
size_t* offsetPtr,
|
163
|
-
U32 nbCompares, const U32 mls,
|
164
|
-
U32 extDict)
|
165
|
-
{
|
166
|
-
U32* const hashTable = zc->hashTable;
|
167
|
-
U32 const hashLog = zc->appliedParams.cParams.hashLog;
|
168
|
-
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
|
169
|
-
U32* const bt = zc->chainTable;
|
170
|
-
U32 const btLog = zc->appliedParams.cParams.chainLog - 1;
|
171
|
-
U32 const btMask = (1 << btLog) - 1;
|
172
|
-
U32 matchIndex = hashTable[h];
|
173
180
|
size_t commonLengthSmaller=0, commonLengthLarger=0;
|
174
|
-
const BYTE* const base = zc->base;
|
175
|
-
const BYTE* const dictBase = zc->dictBase;
|
176
|
-
const U32 dictLimit = zc->dictLimit;
|
177
|
-
const BYTE* const dictEnd = dictBase + dictLimit;
|
178
|
-
const BYTE* const prefixStart = base + dictLimit;
|
179
|
-
const U32 current = (U32)(ip-base);
|
180
|
-
const U32 btLow = btMask >= current ? 0 : current - btMask;
|
181
|
-
const U32 windowLow = zc->lowLimit;
|
182
|
-
U32* smallerPtr = bt + 2*(current&btMask);
|
183
|
-
U32* largerPtr = bt + 2*(current&btMask) + 1;
|
184
|
-
U32 matchEndIdx = current+8+1;
|
185
|
-
U32 dummy32; /* to be nullified at the end */
|
186
|
-
size_t bestLength = 0;
|
187
181
|
|
188
|
-
|
189
|
-
|
182
|
+
(void)dictMode;
|
183
|
+
assert(dictMode == ZSTD_dictMatchState);
|
190
184
|
|
191
|
-
while (nbCompares-- && (
|
192
|
-
U32* const nextPtr =
|
185
|
+
while (nbCompares-- && (dictMatchIndex > dictLowLimit)) {
|
186
|
+
U32* const nextPtr = dictBt + 2*(dictMatchIndex & btMask);
|
193
187
|
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
|
194
|
-
const BYTE* match;
|
195
|
-
|
196
|
-
if (
|
197
|
-
match = base +
|
198
|
-
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
|
199
|
-
} else {
|
200
|
-
match = dictBase + matchIndex;
|
201
|
-
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
|
202
|
-
if (matchIndex+matchLength >= dictLimit)
|
203
|
-
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
|
204
|
-
}
|
188
|
+
const BYTE* match = dictBase + dictMatchIndex;
|
189
|
+
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
|
190
|
+
if (dictMatchIndex+matchLength >= dictHighLimit)
|
191
|
+
match = base + dictMatchIndex + dictIndexDelta; /* to prepare for next usage of match[matchLength] */
|
205
192
|
|
206
193
|
if (matchLength > bestLength) {
|
207
|
-
|
208
|
-
|
209
|
-
|
194
|
+
U32 matchIndex = dictMatchIndex + dictIndexDelta;
|
195
|
+
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) {
|
196
|
+
DEBUGLOG(9, "ZSTD_DUBT_findBetterDictMatch(%u) : found better match length %u -> %u and offsetCode %u -> %u (dictMatchIndex %u, matchIndex %u)",
|
197
|
+
current, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, ZSTD_REP_MOVE + current - matchIndex, dictMatchIndex, matchIndex);
|
210
198
|
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
|
211
|
-
|
199
|
+
}
|
200
|
+
if (ip+matchLength == iend) { /* reached end of input : ip[matchLength] is not valid, no way to know if it's larger or smaller than match */
|
212
201
|
break; /* drop, to guarantee consistency (miss a little bit of compression) */
|
213
202
|
}
|
214
203
|
}
|
215
204
|
|
216
205
|
if (match[matchLength] < ip[matchLength]) {
|
217
|
-
/*
|
218
|
-
*smallerPtr = matchIndex; /* update smaller idx */
|
206
|
+
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
|
219
207
|
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
|
220
|
-
|
221
|
-
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
|
222
|
-
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
|
208
|
+
dictMatchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
|
223
209
|
} else {
|
224
210
|
/* match is larger than current */
|
225
|
-
|
211
|
+
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
|
226
212
|
commonLengthLarger = matchLength;
|
227
|
-
|
228
|
-
|
229
|
-
|
230
|
-
} }
|
231
|
-
|
232
|
-
*smallerPtr = *largerPtr = 0;
|
213
|
+
dictMatchIndex = nextPtr[0];
|
214
|
+
}
|
215
|
+
}
|
233
216
|
|
234
|
-
|
235
|
-
|
217
|
+
if (bestLength >= MINMATCH) {
|
218
|
+
U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
|
219
|
+
DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
|
220
|
+
current, (U32)bestLength, (U32)*offsetPtr, mIndex);
|
221
|
+
}
|
236
222
|
return bestLength;
|
223
|
+
|
237
224
|
}
|
238
225
|
|
239
226
|
|
240
|
-
|
241
|
-
|
242
|
-
|
243
|
-
const BYTE* const ip, const BYTE* const iLimit,
|
227
|
+
static size_t
|
228
|
+
ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
|
229
|
+
const BYTE* const ip, const BYTE* const iend,
|
244
230
|
size_t* offsetPtr,
|
245
|
-
|
231
|
+
U32 const mls,
|
232
|
+
const ZSTD_dictMode_e dictMode)
|
246
233
|
{
|
247
|
-
|
248
|
-
|
249
|
-
|
234
|
+
const ZSTD_compressionParameters* const cParams = &ms->cParams;
|
235
|
+
U32* const hashTable = ms->hashTable;
|
236
|
+
U32 const hashLog = cParams->hashLog;
|
237
|
+
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
|
238
|
+
U32 matchIndex = hashTable[h];
|
239
|
+
|
240
|
+
const BYTE* const base = ms->window.base;
|
241
|
+
U32 const current = (U32)(ip-base);
|
242
|
+
U32 const windowLow = ms->window.lowLimit;
|
243
|
+
|
244
|
+
U32* const bt = ms->chainTable;
|
245
|
+
U32 const btLog = cParams->chainLog - 1;
|
246
|
+
U32 const btMask = (1 << btLog) - 1;
|
247
|
+
U32 const btLow = (btMask >= current) ? 0 : current - btMask;
|
248
|
+
U32 const unsortLimit = MAX(btLow, windowLow);
|
249
|
+
|
250
|
+
U32* nextCandidate = bt + 2*(matchIndex&btMask);
|
251
|
+
U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1;
|
252
|
+
U32 nbCompares = 1U << cParams->searchLog;
|
253
|
+
U32 nbCandidates = nbCompares;
|
254
|
+
U32 previousCandidate = 0;
|
255
|
+
|
256
|
+
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", current);
|
257
|
+
assert(ip <= iend-8); /* required for h calculation */
|
258
|
+
|
259
|
+
/* reach end of unsorted candidates list */
|
260
|
+
while ( (matchIndex > unsortLimit)
|
261
|
+
&& (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
|
262
|
+
&& (nbCandidates > 1) ) {
|
263
|
+
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
|
264
|
+
matchIndex);
|
265
|
+
*unsortedMark = previousCandidate; /* the unsortedMark becomes a reversed chain, to move up back to original position */
|
266
|
+
previousCandidate = matchIndex;
|
267
|
+
matchIndex = *nextCandidate;
|
268
|
+
nextCandidate = bt + 2*(matchIndex&btMask);
|
269
|
+
unsortedMark = bt + 2*(matchIndex&btMask) + 1;
|
270
|
+
nbCandidates --;
|
271
|
+
}
|
272
|
+
|
273
|
+
/* nullify last candidate if it's still unsorted
|
274
|
+
* simplification, detrimental to compression ratio, beneficial for speed */
|
275
|
+
if ( (matchIndex > unsortLimit)
|
276
|
+
&& (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
|
277
|
+
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
|
278
|
+
matchIndex);
|
279
|
+
*nextCandidate = *unsortedMark = 0;
|
280
|
+
}
|
281
|
+
|
282
|
+
/* batch sort stacked candidates */
|
283
|
+
matchIndex = previousCandidate;
|
284
|
+
while (matchIndex) { /* will end on matchIndex == 0 */
|
285
|
+
U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
|
286
|
+
U32 const nextCandidateIdx = *nextCandidateIdxPtr;
|
287
|
+
ZSTD_insertDUBT1(ms, matchIndex, iend,
|
288
|
+
nbCandidates, unsortLimit, dictMode);
|
289
|
+
matchIndex = nextCandidateIdx;
|
290
|
+
nbCandidates++;
|
291
|
+
}
|
292
|
+
|
293
|
+
/* find longest match */
|
294
|
+
{ size_t commonLengthSmaller = 0, commonLengthLarger = 0;
|
295
|
+
const BYTE* const dictBase = ms->window.dictBase;
|
296
|
+
const U32 dictLimit = ms->window.dictLimit;
|
297
|
+
const BYTE* const dictEnd = dictBase + dictLimit;
|
298
|
+
const BYTE* const prefixStart = base + dictLimit;
|
299
|
+
U32* smallerPtr = bt + 2*(current&btMask);
|
300
|
+
U32* largerPtr = bt + 2*(current&btMask) + 1;
|
301
|
+
U32 matchEndIdx = current + 8 + 1;
|
302
|
+
U32 dummy32; /* to be nullified at the end */
|
303
|
+
size_t bestLength = 0;
|
304
|
+
|
305
|
+
matchIndex = hashTable[h];
|
306
|
+
hashTable[h] = current; /* Update Hash Table */
|
307
|
+
|
308
|
+
while (nbCompares-- && (matchIndex > windowLow)) {
|
309
|
+
U32* const nextPtr = bt + 2*(matchIndex & btMask);
|
310
|
+
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
|
311
|
+
const BYTE* match;
|
312
|
+
|
313
|
+
if ((dictMode != ZSTD_extDict) || (matchIndex+matchLength >= dictLimit)) {
|
314
|
+
match = base + matchIndex;
|
315
|
+
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
|
316
|
+
} else {
|
317
|
+
match = dictBase + matchIndex;
|
318
|
+
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
|
319
|
+
if (matchIndex+matchLength >= dictLimit)
|
320
|
+
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
|
321
|
+
}
|
322
|
+
|
323
|
+
if (matchLength > bestLength) {
|
324
|
+
if (matchLength > matchEndIdx - matchIndex)
|
325
|
+
matchEndIdx = matchIndex + (U32)matchLength;
|
326
|
+
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
|
327
|
+
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
|
328
|
+
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
|
329
|
+
if (dictMode == ZSTD_dictMatchState) {
|
330
|
+
nbCompares = 0; /* in addition to avoiding checking any
|
331
|
+
* further in this loop, make sure we
|
332
|
+
* skip checking in the dictionary. */
|
333
|
+
}
|
334
|
+
break; /* drop, to guarantee consistency (miss a little bit of compression) */
|
335
|
+
}
|
336
|
+
}
|
337
|
+
|
338
|
+
if (match[matchLength] < ip[matchLength]) {
|
339
|
+
/* match is smaller than current */
|
340
|
+
*smallerPtr = matchIndex; /* update smaller idx */
|
341
|
+
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
|
342
|
+
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
|
343
|
+
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
|
344
|
+
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
|
345
|
+
} else {
|
346
|
+
/* match is larger than current */
|
347
|
+
*largerPtr = matchIndex;
|
348
|
+
commonLengthLarger = matchLength;
|
349
|
+
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
|
350
|
+
largerPtr = nextPtr;
|
351
|
+
matchIndex = nextPtr[0];
|
352
|
+
} }
|
353
|
+
|
354
|
+
*smallerPtr = *largerPtr = 0;
|
355
|
+
|
356
|
+
if (dictMode == ZSTD_dictMatchState && nbCompares) {
|
357
|
+
bestLength = ZSTD_DUBT_findBetterDictMatch(
|
358
|
+
ms, ip, iend,
|
359
|
+
offsetPtr, bestLength, nbCompares,
|
360
|
+
mls, dictMode);
|
361
|
+
}
|
362
|
+
|
363
|
+
assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */
|
364
|
+
ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
|
365
|
+
if (bestLength >= MINMATCH) {
|
366
|
+
U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
|
367
|
+
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
|
368
|
+
current, (U32)bestLength, (U32)*offsetPtr, mIndex);
|
369
|
+
}
|
370
|
+
return bestLength;
|
371
|
+
}
|
250
372
|
}
|
251
373
|
|
252
374
|
|
253
|
-
|
254
|
-
|
255
|
-
|
256
|
-
|
257
|
-
|
375
|
+
/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
|
376
|
+
FORCE_INLINE_TEMPLATE size_t
|
377
|
+
ZSTD_BtFindBestMatch( ZSTD_matchState_t* ms,
|
378
|
+
const BYTE* const ip, const BYTE* const iLimit,
|
379
|
+
size_t* offsetPtr,
|
380
|
+
const U32 mls /* template */,
|
381
|
+
const ZSTD_dictMode_e dictMode)
|
382
|
+
{
|
383
|
+
DEBUGLOG(7, "ZSTD_BtFindBestMatch");
|
384
|
+
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
|
385
|
+
ZSTD_updateDUBT(ms, ip, iLimit, mls);
|
386
|
+
return ZSTD_DUBT_findBestMatch(ms, ip, iLimit, offsetPtr, mls, dictMode);
|
387
|
+
}
|
388
|
+
|
389
|
+
|
390
|
+
static size_t
|
391
|
+
ZSTD_BtFindBestMatch_selectMLS ( ZSTD_matchState_t* ms,
|
392
|
+
const BYTE* ip, const BYTE* const iLimit,
|
393
|
+
size_t* offsetPtr)
|
258
394
|
{
|
259
|
-
switch(
|
395
|
+
switch(ms->cParams.minMatch)
|
260
396
|
{
|
261
397
|
default : /* includes case 3 */
|
262
|
-
case 4 : return ZSTD_BtFindBestMatch(
|
263
|
-
case 5 : return ZSTD_BtFindBestMatch(
|
398
|
+
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
|
399
|
+
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
|
264
400
|
case 7 :
|
265
|
-
case 6 : return ZSTD_BtFindBestMatch(
|
401
|
+
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
|
266
402
|
}
|
267
403
|
}
|
268
404
|
|
269
405
|
|
270
|
-
|
271
|
-
|
272
|
-
|
273
|
-
|
274
|
-
size_t* offsetPtr,
|
275
|
-
const U32 maxNbAttempts, const U32 mls)
|
406
|
+
static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
|
407
|
+
ZSTD_matchState_t* ms,
|
408
|
+
const BYTE* ip, const BYTE* const iLimit,
|
409
|
+
size_t* offsetPtr)
|
276
410
|
{
|
277
|
-
|
278
|
-
|
279
|
-
|
411
|
+
switch(ms->cParams.minMatch)
|
412
|
+
{
|
413
|
+
default : /* includes case 3 */
|
414
|
+
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
|
415
|
+
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
|
416
|
+
case 7 :
|
417
|
+
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
|
418
|
+
}
|
280
419
|
}
|
281
420
|
|
282
421
|
|
283
|
-
static size_t
|
284
|
-
|
422
|
+
static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
|
423
|
+
ZSTD_matchState_t* ms,
|
285
424
|
const BYTE* ip, const BYTE* const iLimit,
|
286
|
-
size_t* offsetPtr
|
287
|
-
const U32 maxNbAttempts, const U32 matchLengthSearch)
|
425
|
+
size_t* offsetPtr)
|
288
426
|
{
|
289
|
-
switch(
|
427
|
+
switch(ms->cParams.minMatch)
|
290
428
|
{
|
291
429
|
default : /* includes case 3 */
|
292
|
-
case 4 : return
|
293
|
-
case 5 : return
|
430
|
+
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
|
431
|
+
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
|
294
432
|
case 7 :
|
295
|
-
case 6 : return
|
433
|
+
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
|
296
434
|
}
|
297
435
|
}
|
298
436
|
|
@@ -301,19 +439,22 @@ static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
|
|
301
439
|
/* *********************************
|
302
440
|
* Hash Chain
|
303
441
|
***********************************/
|
304
|
-
#define NEXT_IN_CHAIN(d, mask) chainTable[(d) & mask]
|
442
|
+
#define NEXT_IN_CHAIN(d, mask) chainTable[(d) & (mask)]
|
305
443
|
|
306
444
|
/* Update chains up to ip (excluded)
|
307
445
|
Assumption : always within prefix (i.e. not within extDict) */
|
308
|
-
U32
|
446
|
+
static U32 ZSTD_insertAndFindFirstIndex_internal(
|
447
|
+
ZSTD_matchState_t* ms,
|
448
|
+
const ZSTD_compressionParameters* const cParams,
|
449
|
+
const BYTE* ip, U32 const mls)
|
309
450
|
{
|
310
|
-
U32* const hashTable =
|
311
|
-
const U32 hashLog =
|
312
|
-
U32* const chainTable =
|
313
|
-
const U32 chainMask = (1 <<
|
314
|
-
const BYTE* const base =
|
451
|
+
U32* const hashTable = ms->hashTable;
|
452
|
+
const U32 hashLog = cParams->hashLog;
|
453
|
+
U32* const chainTable = ms->chainTable;
|
454
|
+
const U32 chainMask = (1 << cParams->chainLog) - 1;
|
455
|
+
const BYTE* const base = ms->window.base;
|
315
456
|
const U32 target = (U32)(ip - base);
|
316
|
-
U32 idx =
|
457
|
+
U32 idx = ms->nextToUpdate;
|
317
458
|
|
318
459
|
while(idx < target) { /* catch up */
|
319
460
|
size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
|
@@ -322,40 +463,47 @@ U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
|
|
322
463
|
idx++;
|
323
464
|
}
|
324
465
|
|
325
|
-
|
466
|
+
ms->nextToUpdate = target;
|
326
467
|
return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
|
327
468
|
}
|
328
469
|
|
470
|
+
U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
|
471
|
+
const ZSTD_compressionParameters* const cParams = &ms->cParams;
|
472
|
+
return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
|
473
|
+
}
|
474
|
+
|
329
475
|
|
330
476
|
/* inlining is important to hardwire a hot branch (template emulation) */
|
331
477
|
FORCE_INLINE_TEMPLATE
|
332
478
|
size_t ZSTD_HcFindBestMatch_generic (
|
333
|
-
|
479
|
+
ZSTD_matchState_t* ms,
|
334
480
|
const BYTE* const ip, const BYTE* const iLimit,
|
335
481
|
size_t* offsetPtr,
|
336
|
-
const U32
|
482
|
+
const U32 mls, const ZSTD_dictMode_e dictMode)
|
337
483
|
{
|
338
|
-
|
339
|
-
const
|
484
|
+
const ZSTD_compressionParameters* const cParams = &ms->cParams;
|
485
|
+
U32* const chainTable = ms->chainTable;
|
486
|
+
const U32 chainSize = (1 << cParams->chainLog);
|
340
487
|
const U32 chainMask = chainSize-1;
|
341
|
-
const BYTE* const base =
|
342
|
-
const BYTE* const dictBase =
|
343
|
-
const U32 dictLimit =
|
488
|
+
const BYTE* const base = ms->window.base;
|
489
|
+
const BYTE* const dictBase = ms->window.dictBase;
|
490
|
+
const U32 dictLimit = ms->window.dictLimit;
|
344
491
|
const BYTE* const prefixStart = base + dictLimit;
|
345
492
|
const BYTE* const dictEnd = dictBase + dictLimit;
|
346
|
-
const U32 lowLimit =
|
493
|
+
const U32 lowLimit = ms->window.lowLimit;
|
347
494
|
const U32 current = (U32)(ip-base);
|
348
495
|
const U32 minChain = current > chainSize ? current - chainSize : 0;
|
349
|
-
|
496
|
+
U32 nbAttempts = 1U << cParams->searchLog;
|
350
497
|
size_t ml=4-1;
|
351
498
|
|
352
499
|
/* HC4 match finder */
|
353
|
-
U32 matchIndex =
|
500
|
+
U32 matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
|
354
501
|
|
355
502
|
for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
|
356
503
|
size_t currentMl=0;
|
357
|
-
if ((
|
504
|
+
if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
|
358
505
|
const BYTE* const match = base + matchIndex;
|
506
|
+
assert(matchIndex >= dictLimit); /* ensures this is true if dictMode != ZSTD_extDict */
|
359
507
|
if (match[ml] == ip[ml]) /* potentially better */
|
360
508
|
currentMl = ZSTD_count(ip, match, iLimit);
|
361
509
|
} else {
|
@@ -376,40 +524,87 @@ size_t ZSTD_HcFindBestMatch_generic (
|
|
376
524
|
matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
|
377
525
|
}
|
378
526
|
|
527
|
+
if (dictMode == ZSTD_dictMatchState) {
|
528
|
+
const ZSTD_matchState_t* const dms = ms->dictMatchState;
|
529
|
+
const U32* const dmsChainTable = dms->chainTable;
|
530
|
+
const U32 dmsChainSize = (1 << dms->cParams.chainLog);
|
531
|
+
const U32 dmsChainMask = dmsChainSize - 1;
|
532
|
+
const U32 dmsLowestIndex = dms->window.dictLimit;
|
533
|
+
const BYTE* const dmsBase = dms->window.base;
|
534
|
+
const BYTE* const dmsEnd = dms->window.nextSrc;
|
535
|
+
const U32 dmsSize = (U32)(dmsEnd - dmsBase);
|
536
|
+
const U32 dmsIndexDelta = dictLimit - dmsSize;
|
537
|
+
const U32 dmsMinChain = dmsSize > dmsChainSize ? dmsSize - dmsChainSize : 0;
|
538
|
+
|
539
|
+
matchIndex = dms->hashTable[ZSTD_hashPtr(ip, dms->cParams.hashLog, mls)];
|
540
|
+
|
541
|
+
for ( ; (matchIndex>dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
|
542
|
+
size_t currentMl=0;
|
543
|
+
const BYTE* const match = dmsBase + matchIndex;
|
544
|
+
assert(match+4 <= dmsEnd);
|
545
|
+
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
|
546
|
+
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
|
547
|
+
|
548
|
+
/* save best solution */
|
549
|
+
if (currentMl > ml) {
|
550
|
+
ml = currentMl;
|
551
|
+
*offsetPtr = current - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
|
552
|
+
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
|
553
|
+
}
|
554
|
+
|
555
|
+
if (matchIndex <= dmsMinChain) break;
|
556
|
+
matchIndex = dmsChainTable[matchIndex & dmsChainMask];
|
557
|
+
}
|
558
|
+
}
|
559
|
+
|
379
560
|
return ml;
|
380
561
|
}
|
381
562
|
|
382
563
|
|
383
564
|
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
|
384
|
-
|
565
|
+
ZSTD_matchState_t* ms,
|
385
566
|
const BYTE* ip, const BYTE* const iLimit,
|
386
|
-
size_t* offsetPtr
|
387
|
-
|
567
|
+
size_t* offsetPtr)
|
568
|
+
{
|
569
|
+
switch(ms->cParams.minMatch)
|
570
|
+
{
|
571
|
+
default : /* includes case 3 */
|
572
|
+
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
|
573
|
+
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
|
574
|
+
case 7 :
|
575
|
+
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
|
576
|
+
}
|
577
|
+
}
|
578
|
+
|
579
|
+
|
580
|
+
static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
|
581
|
+
ZSTD_matchState_t* ms,
|
582
|
+
const BYTE* ip, const BYTE* const iLimit,
|
583
|
+
size_t* offsetPtr)
|
388
584
|
{
|
389
|
-
switch(
|
585
|
+
switch(ms->cParams.minMatch)
|
390
586
|
{
|
391
587
|
default : /* includes case 3 */
|
392
|
-
case 4 : return ZSTD_HcFindBestMatch_generic(
|
393
|
-
case 5 : return ZSTD_HcFindBestMatch_generic(
|
588
|
+
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
|
589
|
+
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
|
394
590
|
case 7 :
|
395
|
-
case 6 : return ZSTD_HcFindBestMatch_generic(
|
591
|
+
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
|
396
592
|
}
|
397
593
|
}
|
398
594
|
|
399
595
|
|
400
596
|
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
|
401
|
-
|
597
|
+
ZSTD_matchState_t* ms,
|
402
598
|
const BYTE* ip, const BYTE* const iLimit,
|
403
|
-
size_t*
|
404
|
-
U32 const maxNbAttempts, U32 const matchLengthSearch)
|
599
|
+
size_t* offsetPtr)
|
405
600
|
{
|
406
|
-
switch(
|
601
|
+
switch(ms->cParams.minMatch)
|
407
602
|
{
|
408
603
|
default : /* includes case 3 */
|
409
|
-
case 4 : return ZSTD_HcFindBestMatch_generic(
|
410
|
-
case 5 : return ZSTD_HcFindBestMatch_generic(
|
604
|
+
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
|
605
|
+
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
|
411
606
|
case 7 :
|
412
|
-
case 6 : return ZSTD_HcFindBestMatch_generic(
|
607
|
+
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
|
413
608
|
}
|
414
609
|
}
|
415
610
|
|
@@ -418,34 +613,58 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
|
|
418
613
|
* Common parser - lazy strategy
|
419
614
|
*********************************/
|
420
615
|
FORCE_INLINE_TEMPLATE
|
421
|
-
size_t ZSTD_compressBlock_lazy_generic(
|
422
|
-
|
423
|
-
|
616
|
+
size_t ZSTD_compressBlock_lazy_generic(
|
617
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore,
|
618
|
+
U32 rep[ZSTD_REP_NUM],
|
619
|
+
const void* src, size_t srcSize,
|
620
|
+
const U32 searchMethod, const U32 depth,
|
621
|
+
ZSTD_dictMode_e const dictMode)
|
424
622
|
{
|
425
|
-
seqStore_t* seqStorePtr = &(ctx->seqStore);
|
426
623
|
const BYTE* const istart = (const BYTE*)src;
|
427
624
|
const BYTE* ip = istart;
|
428
625
|
const BYTE* anchor = istart;
|
429
626
|
const BYTE* const iend = istart + srcSize;
|
430
627
|
const BYTE* const ilimit = iend - 8;
|
431
|
-
const BYTE* const base =
|
432
|
-
|
433
|
-
|
434
|
-
|
435
|
-
|
436
|
-
|
437
|
-
size_t* offsetPtr
|
438
|
-
|
439
|
-
|
440
|
-
|
628
|
+
const BYTE* const base = ms->window.base;
|
629
|
+
const U32 prefixLowestIndex = ms->window.dictLimit;
|
630
|
+
const BYTE* const prefixLowest = base + prefixLowestIndex;
|
631
|
+
|
632
|
+
typedef size_t (*searchMax_f)(
|
633
|
+
ZSTD_matchState_t* ms,
|
634
|
+
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
|
635
|
+
searchMax_f const searchMax = dictMode == ZSTD_dictMatchState ?
|
636
|
+
(searchMethod ? ZSTD_BtFindBestMatch_dictMatchState_selectMLS : ZSTD_HcFindBestMatch_dictMatchState_selectMLS) :
|
637
|
+
(searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS);
|
638
|
+
U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
|
639
|
+
|
640
|
+
const ZSTD_matchState_t* const dms = ms->dictMatchState;
|
641
|
+
const U32 dictLowestIndex = dictMode == ZSTD_dictMatchState ?
|
642
|
+
dms->window.dictLimit : 0;
|
643
|
+
const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
|
644
|
+
dms->window.base : NULL;
|
645
|
+
const BYTE* const dictLowest = dictMode == ZSTD_dictMatchState ?
|
646
|
+
dictBase + dictLowestIndex : NULL;
|
647
|
+
const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
|
648
|
+
dms->window.nextSrc : NULL;
|
649
|
+
const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
|
650
|
+
prefixLowestIndex - (U32)(dictEnd - dictBase) :
|
651
|
+
0;
|
652
|
+
const U32 dictAndPrefixLength = (U32)(ip - prefixLowest + dictEnd - dictLowest);
|
441
653
|
|
442
654
|
/* init */
|
443
|
-
ip += (
|
444
|
-
|
445
|
-
|
655
|
+
ip += (dictAndPrefixLength == 0);
|
656
|
+
ms->nextToUpdate3 = ms->nextToUpdate;
|
657
|
+
if (dictMode == ZSTD_noDict) {
|
658
|
+
U32 const maxRep = (U32)(ip - prefixLowest);
|
446
659
|
if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
|
447
660
|
if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
|
448
661
|
}
|
662
|
+
if (dictMode == ZSTD_dictMatchState) {
|
663
|
+
/* dictMatchState repCode checks don't currently handle repCode == 0
|
664
|
+
* disabling. */
|
665
|
+
assert(offset_1 <= dictAndPrefixLength);
|
666
|
+
assert(offset_2 <= dictAndPrefixLength);
|
667
|
+
}
|
449
668
|
|
450
669
|
/* Match Loop */
|
451
670
|
while (ip < ilimit) {
|
@@ -454,21 +673,34 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
|
|
454
673
|
const BYTE* start=ip+1;
|
455
674
|
|
456
675
|
/* check repCode */
|
457
|
-
if (
|
458
|
-
|
676
|
+
if (dictMode == ZSTD_dictMatchState) {
|
677
|
+
const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
|
678
|
+
const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
|
679
|
+
&& repIndex < prefixLowestIndex) ?
|
680
|
+
dictBase + (repIndex - dictIndexDelta) :
|
681
|
+
base + repIndex;
|
682
|
+
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
|
683
|
+
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
|
684
|
+
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
|
685
|
+
matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
|
686
|
+
if (depth==0) goto _storeSequence;
|
687
|
+
}
|
688
|
+
}
|
689
|
+
if ( dictMode == ZSTD_noDict
|
690
|
+
&& ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
|
459
691
|
matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
|
460
692
|
if (depth==0) goto _storeSequence;
|
461
693
|
}
|
462
694
|
|
463
695
|
/* first search (depth 0) */
|
464
|
-
{ size_t offsetFound =
|
465
|
-
size_t const ml2 = searchMax(
|
696
|
+
{ size_t offsetFound = 999999999;
|
697
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
|
466
698
|
if (ml2 > matchLength)
|
467
699
|
matchLength = ml2, start = ip, offset=offsetFound;
|
468
700
|
}
|
469
701
|
|
470
702
|
if (matchLength < 4) {
|
471
|
-
ip += ((ip-anchor) >>
|
703
|
+
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
|
472
704
|
continue;
|
473
705
|
}
|
474
706
|
|
@@ -476,15 +708,31 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
|
|
476
708
|
if (depth>=1)
|
477
709
|
while (ip<ilimit) {
|
478
710
|
ip ++;
|
479
|
-
if (
|
711
|
+
if ( (dictMode == ZSTD_noDict)
|
712
|
+
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
|
480
713
|
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
|
481
714
|
int const gain2 = (int)(mlRep * 3);
|
482
715
|
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
|
483
716
|
if ((mlRep >= 4) && (gain2 > gain1))
|
484
717
|
matchLength = mlRep, offset = 0, start = ip;
|
485
718
|
}
|
486
|
-
{
|
487
|
-
|
719
|
+
if (dictMode == ZSTD_dictMatchState) {
|
720
|
+
const U32 repIndex = (U32)(ip - base) - offset_1;
|
721
|
+
const BYTE* repMatch = repIndex < prefixLowestIndex ?
|
722
|
+
dictBase + (repIndex - dictIndexDelta) :
|
723
|
+
base + repIndex;
|
724
|
+
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
|
725
|
+
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
|
726
|
+
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
|
727
|
+
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
|
728
|
+
int const gain2 = (int)(mlRep * 3);
|
729
|
+
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
|
730
|
+
if ((mlRep >= 4) && (gain2 > gain1))
|
731
|
+
matchLength = mlRep, offset = 0, start = ip;
|
732
|
+
}
|
733
|
+
}
|
734
|
+
{ size_t offset2=999999999;
|
735
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
|
488
736
|
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
|
489
737
|
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
|
490
738
|
if ((ml2 >= 4) && (gain2 > gain1)) {
|
@@ -495,15 +743,31 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
|
|
495
743
|
/* let's find an even better one */
|
496
744
|
if ((depth==2) && (ip<ilimit)) {
|
497
745
|
ip ++;
|
498
|
-
if (
|
499
|
-
|
500
|
-
|
746
|
+
if ( (dictMode == ZSTD_noDict)
|
747
|
+
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
|
748
|
+
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
|
749
|
+
int const gain2 = (int)(mlRep * 4);
|
501
750
|
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
|
502
|
-
if ((
|
503
|
-
matchLength =
|
751
|
+
if ((mlRep >= 4) && (gain2 > gain1))
|
752
|
+
matchLength = mlRep, offset = 0, start = ip;
|
504
753
|
}
|
505
|
-
{
|
506
|
-
|
754
|
+
if (dictMode == ZSTD_dictMatchState) {
|
755
|
+
const U32 repIndex = (U32)(ip - base) - offset_1;
|
756
|
+
const BYTE* repMatch = repIndex < prefixLowestIndex ?
|
757
|
+
dictBase + (repIndex - dictIndexDelta) :
|
758
|
+
base + repIndex;
|
759
|
+
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
|
760
|
+
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
|
761
|
+
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
|
762
|
+
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
|
763
|
+
int const gain2 = (int)(mlRep * 4);
|
764
|
+
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
|
765
|
+
if ((mlRep >= 4) && (gain2 > gain1))
|
766
|
+
matchLength = mlRep, offset = 0, start = ip;
|
767
|
+
}
|
768
|
+
}
|
769
|
+
{ size_t offset2=999999999;
|
770
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
|
507
771
|
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
|
508
772
|
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
|
509
773
|
if ((ml2 >= 4) && (gain2 > gain1)) {
|
@@ -520,91 +784,156 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
|
|
520
784
|
*/
|
521
785
|
/* catch up */
|
522
786
|
if (offset) {
|
523
|
-
|
524
|
-
|
525
|
-
|
787
|
+
if (dictMode == ZSTD_noDict) {
|
788
|
+
while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
|
789
|
+
&& (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */
|
790
|
+
{ start--; matchLength++; }
|
791
|
+
}
|
792
|
+
if (dictMode == ZSTD_dictMatchState) {
|
793
|
+
U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
|
794
|
+
const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
|
795
|
+
const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
|
796
|
+
while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
|
797
|
+
}
|
526
798
|
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
|
527
799
|
}
|
528
800
|
/* store sequence */
|
529
801
|
_storeSequence:
|
530
802
|
{ size_t const litLength = start - anchor;
|
531
|
-
ZSTD_storeSeq(
|
803
|
+
ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
|
532
804
|
anchor = ip = start + matchLength;
|
533
805
|
}
|
534
806
|
|
535
807
|
/* check immediate repcode */
|
536
|
-
|
537
|
-
|
538
|
-
|
539
|
-
|
540
|
-
|
541
|
-
|
542
|
-
|
543
|
-
|
544
|
-
|
545
|
-
|
808
|
+
if (dictMode == ZSTD_dictMatchState) {
|
809
|
+
while (ip <= ilimit) {
|
810
|
+
U32 const current2 = (U32)(ip-base);
|
811
|
+
U32 const repIndex = current2 - offset_2;
|
812
|
+
const BYTE* repMatch = dictMode == ZSTD_dictMatchState
|
813
|
+
&& repIndex < prefixLowestIndex ?
|
814
|
+
dictBase - dictIndexDelta + repIndex :
|
815
|
+
base + repIndex;
|
816
|
+
if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
|
817
|
+
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
|
818
|
+
const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
|
819
|
+
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
|
820
|
+
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset_2 <=> offset_1 */
|
821
|
+
ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
|
822
|
+
ip += matchLength;
|
823
|
+
anchor = ip;
|
824
|
+
continue;
|
825
|
+
}
|
826
|
+
break;
|
827
|
+
}
|
828
|
+
}
|
829
|
+
|
830
|
+
if (dictMode == ZSTD_noDict) {
|
831
|
+
while ( ((ip <= ilimit) & (offset_2>0))
|
832
|
+
&& (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
|
833
|
+
/* store sequence */
|
834
|
+
matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
|
835
|
+
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
|
836
|
+
ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
|
837
|
+
ip += matchLength;
|
838
|
+
anchor = ip;
|
839
|
+
continue; /* faster when present ... (?) */
|
840
|
+
} } }
|
546
841
|
|
547
842
|
/* Save reps for next block */
|
548
|
-
|
549
|
-
|
843
|
+
rep[0] = offset_1 ? offset_1 : savedOffset;
|
844
|
+
rep[1] = offset_2 ? offset_2 : savedOffset;
|
550
845
|
|
551
846
|
/* Return the last literals size */
|
552
847
|
return iend - anchor;
|
553
848
|
}
|
554
849
|
|
555
850
|
|
556
|
-
size_t ZSTD_compressBlock_btlazy2(
|
851
|
+
size_t ZSTD_compressBlock_btlazy2(
|
852
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
853
|
+
void const* src, size_t srcSize)
|
557
854
|
{
|
558
|
-
return ZSTD_compressBlock_lazy_generic(
|
855
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 1, 2, ZSTD_noDict);
|
559
856
|
}
|
560
857
|
|
561
|
-
size_t ZSTD_compressBlock_lazy2(
|
858
|
+
size_t ZSTD_compressBlock_lazy2(
|
859
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
860
|
+
void const* src, size_t srcSize)
|
562
861
|
{
|
563
|
-
return ZSTD_compressBlock_lazy_generic(
|
862
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 2, ZSTD_noDict);
|
564
863
|
}
|
565
864
|
|
566
|
-
size_t ZSTD_compressBlock_lazy(
|
865
|
+
size_t ZSTD_compressBlock_lazy(
|
866
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
867
|
+
void const* src, size_t srcSize)
|
567
868
|
{
|
568
|
-
return ZSTD_compressBlock_lazy_generic(
|
869
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 1, ZSTD_noDict);
|
569
870
|
}
|
570
871
|
|
571
|
-
size_t ZSTD_compressBlock_greedy(
|
872
|
+
size_t ZSTD_compressBlock_greedy(
|
873
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
874
|
+
void const* src, size_t srcSize)
|
572
875
|
{
|
573
|
-
return ZSTD_compressBlock_lazy_generic(
|
876
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 0, ZSTD_noDict);
|
877
|
+
}
|
878
|
+
|
879
|
+
size_t ZSTD_compressBlock_btlazy2_dictMatchState(
|
880
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
881
|
+
void const* src, size_t srcSize)
|
882
|
+
{
|
883
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 1, 2, ZSTD_dictMatchState);
|
884
|
+
}
|
885
|
+
|
886
|
+
size_t ZSTD_compressBlock_lazy2_dictMatchState(
|
887
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
888
|
+
void const* src, size_t srcSize)
|
889
|
+
{
|
890
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 2, ZSTD_dictMatchState);
|
891
|
+
}
|
892
|
+
|
893
|
+
size_t ZSTD_compressBlock_lazy_dictMatchState(
|
894
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
895
|
+
void const* src, size_t srcSize)
|
896
|
+
{
|
897
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 1, ZSTD_dictMatchState);
|
898
|
+
}
|
899
|
+
|
900
|
+
size_t ZSTD_compressBlock_greedy_dictMatchState(
|
901
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
902
|
+
void const* src, size_t srcSize)
|
903
|
+
{
|
904
|
+
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, 0, 0, ZSTD_dictMatchState);
|
574
905
|
}
|
575
906
|
|
576
907
|
|
577
908
|
FORCE_INLINE_TEMPLATE
|
578
|
-
size_t ZSTD_compressBlock_lazy_extDict_generic(
|
579
|
-
|
580
|
-
|
909
|
+
size_t ZSTD_compressBlock_lazy_extDict_generic(
|
910
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore,
|
911
|
+
U32 rep[ZSTD_REP_NUM],
|
912
|
+
const void* src, size_t srcSize,
|
913
|
+
const U32 searchMethod, const U32 depth)
|
581
914
|
{
|
582
|
-
seqStore_t* seqStorePtr = &(ctx->seqStore);
|
583
915
|
const BYTE* const istart = (const BYTE*)src;
|
584
916
|
const BYTE* ip = istart;
|
585
917
|
const BYTE* anchor = istart;
|
586
918
|
const BYTE* const iend = istart + srcSize;
|
587
919
|
const BYTE* const ilimit = iend - 8;
|
588
|
-
const BYTE* const base =
|
589
|
-
const U32 dictLimit =
|
590
|
-
const U32 lowestIndex =
|
920
|
+
const BYTE* const base = ms->window.base;
|
921
|
+
const U32 dictLimit = ms->window.dictLimit;
|
922
|
+
const U32 lowestIndex = ms->window.lowLimit;
|
591
923
|
const BYTE* const prefixStart = base + dictLimit;
|
592
|
-
const BYTE* const dictBase =
|
924
|
+
const BYTE* const dictBase = ms->window.dictBase;
|
593
925
|
const BYTE* const dictEnd = dictBase + dictLimit;
|
594
|
-
const BYTE* const dictStart = dictBase +
|
926
|
+
const BYTE* const dictStart = dictBase + lowestIndex;
|
595
927
|
|
596
|
-
|
597
|
-
|
928
|
+
typedef size_t (*searchMax_f)(
|
929
|
+
ZSTD_matchState_t* ms,
|
930
|
+
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
|
931
|
+
searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
|
598
932
|
|
599
|
-
|
600
|
-
size_t* offsetPtr,
|
601
|
-
U32 maxNbAttempts, U32 matchLengthSearch);
|
602
|
-
searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
|
603
|
-
|
604
|
-
U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1];
|
933
|
+
U32 offset_1 = rep[0], offset_2 = rep[1];
|
605
934
|
|
606
935
|
/* init */
|
607
|
-
|
936
|
+
ms->nextToUpdate3 = ms->nextToUpdate;
|
608
937
|
ip += (ip == prefixStart);
|
609
938
|
|
610
939
|
/* Match Loop */
|
@@ -627,14 +956,14 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
|
|
627
956
|
} }
|
628
957
|
|
629
958
|
/* first search (depth 0) */
|
630
|
-
{ size_t offsetFound =
|
631
|
-
size_t const ml2 = searchMax(
|
959
|
+
{ size_t offsetFound = 999999999;
|
960
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
|
632
961
|
if (ml2 > matchLength)
|
633
962
|
matchLength = ml2, start = ip, offset=offsetFound;
|
634
963
|
}
|
635
964
|
|
636
965
|
if (matchLength < 4) {
|
637
|
-
ip += ((ip-anchor) >>
|
966
|
+
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
|
638
967
|
continue;
|
639
968
|
}
|
640
969
|
|
@@ -660,8 +989,8 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
|
|
660
989
|
} }
|
661
990
|
|
662
991
|
/* search match, depth 1 */
|
663
|
-
{ size_t offset2=
|
664
|
-
size_t const ml2 = searchMax(
|
992
|
+
{ size_t offset2=999999999;
|
993
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
|
665
994
|
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
|
666
995
|
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
|
667
996
|
if ((ml2 >= 4) && (gain2 > gain1)) {
|
@@ -690,8 +1019,8 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
|
|
690
1019
|
} }
|
691
1020
|
|
692
1021
|
/* search match, depth 2 */
|
693
|
-
{ size_t offset2=
|
694
|
-
size_t const ml2 = searchMax(
|
1022
|
+
{ size_t offset2=999999999;
|
1023
|
+
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
|
695
1024
|
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
|
696
1025
|
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
|
697
1026
|
if ((ml2 >= 4) && (gain2 > gain1)) {
|
@@ -713,7 +1042,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
|
|
713
1042
|
/* store sequence */
|
714
1043
|
_storeSequence:
|
715
1044
|
{ size_t const litLength = start - anchor;
|
716
|
-
ZSTD_storeSeq(
|
1045
|
+
ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
|
717
1046
|
anchor = ip = start + matchLength;
|
718
1047
|
}
|
719
1048
|
|
@@ -728,7 +1057,7 @@ _storeSequence:
|
|
728
1057
|
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
|
729
1058
|
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
|
730
1059
|
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */
|
731
|
-
ZSTD_storeSeq(
|
1060
|
+
ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
|
732
1061
|
ip += matchLength;
|
733
1062
|
anchor = ip;
|
734
1063
|
continue; /* faster when present ... (?) */
|
@@ -737,29 +1066,41 @@ _storeSequence:
|
|
737
1066
|
} }
|
738
1067
|
|
739
1068
|
/* Save reps for next block */
|
740
|
-
|
1069
|
+
rep[0] = offset_1;
|
1070
|
+
rep[1] = offset_2;
|
741
1071
|
|
742
1072
|
/* Return the last literals size */
|
743
1073
|
return iend - anchor;
|
744
1074
|
}
|
745
1075
|
|
746
1076
|
|
747
|
-
size_t ZSTD_compressBlock_greedy_extDict(
|
1077
|
+
size_t ZSTD_compressBlock_greedy_extDict(
|
1078
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
1079
|
+
void const* src, size_t srcSize)
|
748
1080
|
{
|
749
|
-
return ZSTD_compressBlock_lazy_extDict_generic(
|
1081
|
+
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, 0, 0);
|
750
1082
|
}
|
751
1083
|
|
752
|
-
size_t ZSTD_compressBlock_lazy_extDict(
|
1084
|
+
size_t ZSTD_compressBlock_lazy_extDict(
|
1085
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
1086
|
+
void const* src, size_t srcSize)
|
1087
|
+
|
753
1088
|
{
|
754
|
-
return ZSTD_compressBlock_lazy_extDict_generic(
|
1089
|
+
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, 0, 1);
|
755
1090
|
}
|
756
1091
|
|
757
|
-
size_t ZSTD_compressBlock_lazy2_extDict(
|
1092
|
+
size_t ZSTD_compressBlock_lazy2_extDict(
|
1093
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
1094
|
+
void const* src, size_t srcSize)
|
1095
|
+
|
758
1096
|
{
|
759
|
-
return ZSTD_compressBlock_lazy_extDict_generic(
|
1097
|
+
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, 0, 2);
|
760
1098
|
}
|
761
1099
|
|
762
|
-
size_t ZSTD_compressBlock_btlazy2_extDict(
|
1100
|
+
size_t ZSTD_compressBlock_btlazy2_extDict(
|
1101
|
+
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
|
1102
|
+
void const* src, size_t srcSize)
|
1103
|
+
|
763
1104
|
{
|
764
|
-
return ZSTD_compressBlock_lazy_extDict_generic(
|
1105
|
+
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, 1, 2);
|
765
1106
|
}
|