Book a Demo!
CoCalc Logo Icon
StoreFeaturesDocsShareSupportNewsAboutPoliciesSign UpSign In
freebsd
GitHub Repository: freebsd/freebsd-src
Path: blob/main/sys/contrib/openzfs/module/zstd/lib/compress/zstd_lazy.c
48774 views
1
// SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0-only
2
/*
3
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
4
* All rights reserved.
5
*
6
* This source code is licensed under both the BSD-style license (found in the
7
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
8
* in the COPYING file in the root directory of this source tree).
9
* You may select, at your option, one of the above-listed licenses.
10
*/
11
12
#include "zstd_compress_internal.h"
13
#include "zstd_lazy.h"
14
15
16
/*-*************************************
17
* Binary Tree search
18
***************************************/
19
20
static void
21
ZSTD_updateDUBT(ZSTD_matchState_t* ms,
22
const BYTE* ip, const BYTE* iend,
23
U32 mls)
24
{
25
const ZSTD_compressionParameters* const cParams = &ms->cParams;
26
U32* const hashTable = ms->hashTable;
27
U32 const hashLog = cParams->hashLog;
28
29
U32* const bt = ms->chainTable;
30
U32 const btLog = cParams->chainLog - 1;
31
U32 const btMask = (1 << btLog) - 1;
32
33
const BYTE* const base = ms->window.base;
34
U32 const target = (U32)(ip - base);
35
U32 idx = ms->nextToUpdate;
36
37
if (idx != target)
38
DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
39
idx, target, ms->window.dictLimit);
40
assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */
41
(void)iend;
42
43
assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */
44
for ( ; idx < target ; idx++) {
45
size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */
46
U32 const matchIndex = hashTable[h];
47
48
U32* const nextCandidatePtr = bt + 2*(idx&btMask);
49
U32* const sortMarkPtr = nextCandidatePtr + 1;
50
51
DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
52
hashTable[h] = idx; /* Update Hash Table */
53
*nextCandidatePtr = matchIndex; /* update BT like a chain */
54
*sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
55
}
56
ms->nextToUpdate = target;
57
}
58
59
60
/** ZSTD_insertDUBT1() :
61
* sort one already inserted but unsorted position
62
* assumption : current >= btlow == (current - btmask)
63
* doesn't fail */
64
static void
65
ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
66
U32 current, const BYTE* inputEnd,
67
U32 nbCompares, U32 btLow,
68
const ZSTD_dictMode_e dictMode)
69
{
70
const ZSTD_compressionParameters* const cParams = &ms->cParams;
71
U32* const bt = ms->chainTable;
72
U32 const btLog = cParams->chainLog - 1;
73
U32 const btMask = (1 << btLog) - 1;
74
size_t commonLengthSmaller=0, commonLengthLarger=0;
75
const BYTE* const base = ms->window.base;
76
const BYTE* const dictBase = ms->window.dictBase;
77
const U32 dictLimit = ms->window.dictLimit;
78
const BYTE* const ip = (current>=dictLimit) ? base + current : dictBase + current;
79
const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit;
80
const BYTE* const dictEnd = dictBase + dictLimit;
81
const BYTE* const prefixStart = base + dictLimit;
82
const BYTE* match;
83
U32* smallerPtr = bt + 2*(current&btMask);
84
U32* largerPtr = smallerPtr + 1;
85
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) */
86
U32 dummy32; /* to be nullified at the end */
87
U32 const windowValid = ms->window.lowLimit;
88
U32 const maxDistance = 1U << cParams->windowLog;
89
U32 const windowLow = (current - windowValid > maxDistance) ? current - maxDistance : windowValid;
90
91
92
DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
93
current, dictLimit, windowLow);
94
assert(current >= btLow);
95
assert(ip < iend); /* condition for ZSTD_count */
96
97
while (nbCompares-- && (matchIndex > windowLow)) {
98
U32* const nextPtr = bt + 2*(matchIndex & btMask);
99
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
100
assert(matchIndex < current);
101
/* note : all candidates are now supposed sorted,
102
* but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
103
* when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
104
105
if ( (dictMode != ZSTD_extDict)
106
|| (matchIndex+matchLength >= dictLimit) /* both in current segment*/
107
|| (current < dictLimit) /* both in extDict */) {
108
const BYTE* const mBase = ( (dictMode != ZSTD_extDict)
109
|| (matchIndex+matchLength >= dictLimit)) ?
110
base : dictBase;
111
assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */
112
|| (current < dictLimit) );
113
match = mBase + matchIndex;
114
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
115
} else {
116
match = dictBase + matchIndex;
117
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
118
if (matchIndex+matchLength >= dictLimit)
119
match = base + matchIndex; /* preparation for next read of match[matchLength] */
120
}
121
122
DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
123
current, matchIndex, (U32)matchLength);
124
125
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
126
break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
127
}
128
129
if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */
130
/* match is smaller than current */
131
*smallerPtr = matchIndex; /* update smaller idx */
132
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
133
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
134
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
135
matchIndex, btLow, nextPtr[1]);
136
smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
137
matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
138
} else {
139
/* match is larger than current */
140
*largerPtr = matchIndex;
141
commonLengthLarger = matchLength;
142
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
143
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
144
matchIndex, btLow, nextPtr[0]);
145
largerPtr = nextPtr;
146
matchIndex = nextPtr[0];
147
} }
148
149
*smallerPtr = *largerPtr = 0;
150
}
151
152
153
static size_t
154
ZSTD_DUBT_findBetterDictMatch (
155
ZSTD_matchState_t* ms,
156
const BYTE* const ip, const BYTE* const iend,
157
size_t* offsetPtr,
158
size_t bestLength,
159
U32 nbCompares,
160
U32 const mls,
161
const ZSTD_dictMode_e dictMode)
162
{
163
const ZSTD_matchState_t * const dms = ms->dictMatchState;
164
const ZSTD_compressionParameters* const dmsCParams = &dms->cParams;
165
const U32 * const dictHashTable = dms->hashTable;
166
U32 const hashLog = dmsCParams->hashLog;
167
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
168
U32 dictMatchIndex = dictHashTable[h];
169
170
const BYTE* const base = ms->window.base;
171
const BYTE* const prefixStart = base + ms->window.dictLimit;
172
U32 const current = (U32)(ip-base);
173
const BYTE* const dictBase = dms->window.base;
174
const BYTE* const dictEnd = dms->window.nextSrc;
175
U32 const dictHighLimit = (U32)(dms->window.nextSrc - dms->window.base);
176
U32 const dictLowLimit = dms->window.lowLimit;
177
U32 const dictIndexDelta = ms->window.lowLimit - dictHighLimit;
178
179
U32* const dictBt = dms->chainTable;
180
U32 const btLog = dmsCParams->chainLog - 1;
181
U32 const btMask = (1 << btLog) - 1;
182
U32 const btLow = (btMask >= dictHighLimit - dictLowLimit) ? dictLowLimit : dictHighLimit - btMask;
183
184
size_t commonLengthSmaller=0, commonLengthLarger=0;
185
186
(void)dictMode;
187
assert(dictMode == ZSTD_dictMatchState);
188
189
while (nbCompares-- && (dictMatchIndex > dictLowLimit)) {
190
U32* const nextPtr = dictBt + 2*(dictMatchIndex & btMask);
191
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
192
const BYTE* match = dictBase + dictMatchIndex;
193
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
194
if (dictMatchIndex+matchLength >= dictHighLimit)
195
match = base + dictMatchIndex + dictIndexDelta; /* to prepare for next usage of match[matchLength] */
196
197
if (matchLength > bestLength) {
198
U32 matchIndex = dictMatchIndex + dictIndexDelta;
199
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) {
200
DEBUGLOG(9, "ZSTD_DUBT_findBetterDictMatch(%u) : found better match length %u -> %u and offsetCode %u -> %u (dictMatchIndex %u, matchIndex %u)",
201
current, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, ZSTD_REP_MOVE + current - matchIndex, dictMatchIndex, matchIndex);
202
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
203
}
204
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 */
205
break; /* drop, to guarantee consistency (miss a little bit of compression) */
206
}
207
}
208
209
if (match[matchLength] < ip[matchLength]) {
210
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
211
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
212
dictMatchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
213
} else {
214
/* match is larger than current */
215
if (dictMatchIndex <= btLow) { break; } /* beyond tree size, stop the search */
216
commonLengthLarger = matchLength;
217
dictMatchIndex = nextPtr[0];
218
}
219
}
220
221
if (bestLength >= MINMATCH) {
222
U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
223
DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
224
current, (U32)bestLength, (U32)*offsetPtr, mIndex);
225
}
226
return bestLength;
227
228
}
229
230
231
static size_t
232
ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
233
const BYTE* const ip, const BYTE* const iend,
234
size_t* offsetPtr,
235
U32 const mls,
236
const ZSTD_dictMode_e dictMode)
237
{
238
const ZSTD_compressionParameters* const cParams = &ms->cParams;
239
U32* const hashTable = ms->hashTable;
240
U32 const hashLog = cParams->hashLog;
241
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
242
U32 matchIndex = hashTable[h];
243
244
const BYTE* const base = ms->window.base;
245
U32 const current = (U32)(ip-base);
246
U32 const windowLow = ZSTD_getLowestMatchIndex(ms, current, cParams->windowLog);
247
248
U32* const bt = ms->chainTable;
249
U32 const btLog = cParams->chainLog - 1;
250
U32 const btMask = (1 << btLog) - 1;
251
U32 const btLow = (btMask >= current) ? 0 : current - btMask;
252
U32 const unsortLimit = MAX(btLow, windowLow);
253
254
U32* nextCandidate = bt + 2*(matchIndex&btMask);
255
U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1;
256
U32 nbCompares = 1U << cParams->searchLog;
257
U32 nbCandidates = nbCompares;
258
U32 previousCandidate = 0;
259
260
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", current);
261
assert(ip <= iend-8); /* required for h calculation */
262
263
/* reach end of unsorted candidates list */
264
while ( (matchIndex > unsortLimit)
265
&& (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
266
&& (nbCandidates > 1) ) {
267
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
268
matchIndex);
269
*unsortedMark = previousCandidate; /* the unsortedMark becomes a reversed chain, to move up back to original position */
270
previousCandidate = matchIndex;
271
matchIndex = *nextCandidate;
272
nextCandidate = bt + 2*(matchIndex&btMask);
273
unsortedMark = bt + 2*(matchIndex&btMask) + 1;
274
nbCandidates --;
275
}
276
277
/* nullify last candidate if it's still unsorted
278
* simplification, detrimental to compression ratio, beneficial for speed */
279
if ( (matchIndex > unsortLimit)
280
&& (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
281
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
282
matchIndex);
283
*nextCandidate = *unsortedMark = 0;
284
}
285
286
/* batch sort stacked candidates */
287
matchIndex = previousCandidate;
288
while (matchIndex) { /* will end on matchIndex == 0 */
289
U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
290
U32 const nextCandidateIdx = *nextCandidateIdxPtr;
291
ZSTD_insertDUBT1(ms, matchIndex, iend,
292
nbCandidates, unsortLimit, dictMode);
293
matchIndex = nextCandidateIdx;
294
nbCandidates++;
295
}
296
297
/* find longest match */
298
{ size_t commonLengthSmaller = 0, commonLengthLarger = 0;
299
const BYTE* const dictBase = ms->window.dictBase;
300
const U32 dictLimit = ms->window.dictLimit;
301
const BYTE* const dictEnd = dictBase + dictLimit;
302
const BYTE* const prefixStart = base + dictLimit;
303
U32* smallerPtr = bt + 2*(current&btMask);
304
U32* largerPtr = bt + 2*(current&btMask) + 1;
305
U32 matchEndIdx = current + 8 + 1;
306
U32 dummy32; /* to be nullified at the end */
307
size_t bestLength = 0;
308
309
matchIndex = hashTable[h];
310
hashTable[h] = current; /* Update Hash Table */
311
312
while (nbCompares-- && (matchIndex > windowLow)) {
313
U32* const nextPtr = bt + 2*(matchIndex & btMask);
314
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
315
const BYTE* match;
316
317
if ((dictMode != ZSTD_extDict) || (matchIndex+matchLength >= dictLimit)) {
318
match = base + matchIndex;
319
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
320
} else {
321
match = dictBase + matchIndex;
322
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
323
if (matchIndex+matchLength >= dictLimit)
324
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
325
}
326
327
if (matchLength > bestLength) {
328
if (matchLength > matchEndIdx - matchIndex)
329
matchEndIdx = matchIndex + (U32)matchLength;
330
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
331
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
332
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
333
if (dictMode == ZSTD_dictMatchState) {
334
nbCompares = 0; /* in addition to avoiding checking any
335
* further in this loop, make sure we
336
* skip checking in the dictionary. */
337
}
338
break; /* drop, to guarantee consistency (miss a little bit of compression) */
339
}
340
}
341
342
if (match[matchLength] < ip[matchLength]) {
343
/* match is smaller than current */
344
*smallerPtr = matchIndex; /* update smaller idx */
345
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
346
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
347
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
348
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
349
} else {
350
/* match is larger than current */
351
*largerPtr = matchIndex;
352
commonLengthLarger = matchLength;
353
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
354
largerPtr = nextPtr;
355
matchIndex = nextPtr[0];
356
} }
357
358
*smallerPtr = *largerPtr = 0;
359
360
if (dictMode == ZSTD_dictMatchState && nbCompares) {
361
bestLength = ZSTD_DUBT_findBetterDictMatch(
362
ms, ip, iend,
363
offsetPtr, bestLength, nbCompares,
364
mls, dictMode);
365
}
366
367
assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */
368
ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
369
if (bestLength >= MINMATCH) {
370
U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
371
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
372
current, (U32)bestLength, (U32)*offsetPtr, mIndex);
373
}
374
return bestLength;
375
}
376
}
377
378
379
/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
380
FORCE_INLINE_TEMPLATE size_t
381
ZSTD_BtFindBestMatch( ZSTD_matchState_t* ms,
382
const BYTE* const ip, const BYTE* const iLimit,
383
size_t* offsetPtr,
384
const U32 mls /* template */,
385
const ZSTD_dictMode_e dictMode)
386
{
387
DEBUGLOG(7, "ZSTD_BtFindBestMatch");
388
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
389
ZSTD_updateDUBT(ms, ip, iLimit, mls);
390
return ZSTD_DUBT_findBestMatch(ms, ip, iLimit, offsetPtr, mls, dictMode);
391
}
392
393
394
static size_t
395
ZSTD_BtFindBestMatch_selectMLS ( ZSTD_matchState_t* ms,
396
const BYTE* ip, const BYTE* const iLimit,
397
size_t* offsetPtr)
398
{
399
switch(ms->cParams.minMatch)
400
{
401
default : /* includes case 3 */
402
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
403
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
404
case 7 :
405
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
406
}
407
}
408
409
410
static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
411
ZSTD_matchState_t* ms,
412
const BYTE* ip, const BYTE* const iLimit,
413
size_t* offsetPtr)
414
{
415
switch(ms->cParams.minMatch)
416
{
417
default : /* includes case 3 */
418
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
419
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
420
case 7 :
421
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
422
}
423
}
424
425
426
static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
427
ZSTD_matchState_t* ms,
428
const BYTE* ip, const BYTE* const iLimit,
429
size_t* offsetPtr)
430
{
431
switch(ms->cParams.minMatch)
432
{
433
default : /* includes case 3 */
434
case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
435
case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
436
case 7 :
437
case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
438
}
439
}
440
441
442
443
/* *********************************
444
* Hash Chain
445
***********************************/
446
#define NEXT_IN_CHAIN(d, mask) chainTable[(d) & (mask)]
447
448
/* Update chains up to ip (excluded)
449
Assumption : always within prefix (i.e. not within extDict) */
450
static U32 ZSTD_insertAndFindFirstIndex_internal(
451
ZSTD_matchState_t* ms,
452
const ZSTD_compressionParameters* const cParams,
453
const BYTE* ip, U32 const mls)
454
{
455
U32* const hashTable = ms->hashTable;
456
const U32 hashLog = cParams->hashLog;
457
U32* const chainTable = ms->chainTable;
458
const U32 chainMask = (1 << cParams->chainLog) - 1;
459
const BYTE* const base = ms->window.base;
460
const U32 target = (U32)(ip - base);
461
U32 idx = ms->nextToUpdate;
462
463
while(idx < target) { /* catch up */
464
size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
465
NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
466
hashTable[h] = idx;
467
idx++;
468
}
469
470
ms->nextToUpdate = target;
471
return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
472
}
473
474
U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
475
const ZSTD_compressionParameters* const cParams = &ms->cParams;
476
return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
477
}
478
479
480
/* inlining is important to hardwire a hot branch (template emulation) */
481
FORCE_INLINE_TEMPLATE
482
size_t ZSTD_HcFindBestMatch_generic (
483
ZSTD_matchState_t* ms,
484
const BYTE* const ip, const BYTE* const iLimit,
485
size_t* offsetPtr,
486
const U32 mls, const ZSTD_dictMode_e dictMode)
487
{
488
const ZSTD_compressionParameters* const cParams = &ms->cParams;
489
U32* const chainTable = ms->chainTable;
490
const U32 chainSize = (1 << cParams->chainLog);
491
const U32 chainMask = chainSize-1;
492
const BYTE* const base = ms->window.base;
493
const BYTE* const dictBase = ms->window.dictBase;
494
const U32 dictLimit = ms->window.dictLimit;
495
const BYTE* const prefixStart = base + dictLimit;
496
const BYTE* const dictEnd = dictBase + dictLimit;
497
const U32 current = (U32)(ip-base);
498
const U32 maxDistance = 1U << cParams->windowLog;
499
const U32 lowestValid = ms->window.lowLimit;
500
const U32 withinMaxDistance = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
501
const U32 isDictionary = (ms->loadedDictEnd != 0);
502
const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
503
const U32 minChain = current > chainSize ? current - chainSize : 0;
504
U32 nbAttempts = 1U << cParams->searchLog;
505
size_t ml=4-1;
506
507
/* HC4 match finder */
508
U32 matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
509
510
for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
511
size_t currentMl=0;
512
if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
513
const BYTE* const match = base + matchIndex;
514
assert(matchIndex >= dictLimit); /* ensures this is true if dictMode != ZSTD_extDict */
515
if (match[ml] == ip[ml]) /* potentially better */
516
currentMl = ZSTD_count(ip, match, iLimit);
517
} else {
518
const BYTE* const match = dictBase + matchIndex;
519
assert(match+4 <= dictEnd);
520
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
521
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
522
}
523
524
/* save best solution */
525
if (currentMl > ml) {
526
ml = currentMl;
527
*offsetPtr = current - matchIndex + ZSTD_REP_MOVE;
528
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
529
}
530
531
if (matchIndex <= minChain) break;
532
matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
533
}
534
535
if (dictMode == ZSTD_dictMatchState) {
536
const ZSTD_matchState_t* const dms = ms->dictMatchState;
537
const U32* const dmsChainTable = dms->chainTable;
538
const U32 dmsChainSize = (1 << dms->cParams.chainLog);
539
const U32 dmsChainMask = dmsChainSize - 1;
540
const U32 dmsLowestIndex = dms->window.dictLimit;
541
const BYTE* const dmsBase = dms->window.base;
542
const BYTE* const dmsEnd = dms->window.nextSrc;
543
const U32 dmsSize = (U32)(dmsEnd - dmsBase);
544
const U32 dmsIndexDelta = dictLimit - dmsSize;
545
const U32 dmsMinChain = dmsSize > dmsChainSize ? dmsSize - dmsChainSize : 0;
546
547
matchIndex = dms->hashTable[ZSTD_hashPtr(ip, dms->cParams.hashLog, mls)];
548
549
for ( ; (matchIndex>dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
550
size_t currentMl=0;
551
const BYTE* const match = dmsBase + matchIndex;
552
assert(match+4 <= dmsEnd);
553
if (MEM_read32(match) == MEM_read32(ip)) /* assumption : matchIndex <= dictLimit-4 (by table construction) */
554
currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
555
556
/* save best solution */
557
if (currentMl > ml) {
558
ml = currentMl;
559
*offsetPtr = current - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
560
if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
561
}
562
563
if (matchIndex <= dmsMinChain) break;
564
matchIndex = dmsChainTable[matchIndex & dmsChainMask];
565
}
566
}
567
568
return ml;
569
}
570
571
572
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
573
ZSTD_matchState_t* ms,
574
const BYTE* ip, const BYTE* const iLimit,
575
size_t* offsetPtr)
576
{
577
switch(ms->cParams.minMatch)
578
{
579
default : /* includes case 3 */
580
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
581
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
582
case 7 :
583
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
584
}
585
}
586
587
588
static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
589
ZSTD_matchState_t* ms,
590
const BYTE* ip, const BYTE* const iLimit,
591
size_t* offsetPtr)
592
{
593
switch(ms->cParams.minMatch)
594
{
595
default : /* includes case 3 */
596
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
597
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
598
case 7 :
599
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
600
}
601
}
602
603
604
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
605
ZSTD_matchState_t* ms,
606
const BYTE* ip, const BYTE* const iLimit,
607
size_t* offsetPtr)
608
{
609
switch(ms->cParams.minMatch)
610
{
611
default : /* includes case 3 */
612
case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
613
case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
614
case 7 :
615
case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
616
}
617
}
618
619
620
/* *******************************
621
* Common parser - lazy strategy
622
*********************************/
623
typedef enum { search_hashChain, search_binaryTree } searchMethod_e;
624
625
FORCE_INLINE_TEMPLATE size_t
626
ZSTD_compressBlock_lazy_generic(
627
ZSTD_matchState_t* ms, seqStore_t* seqStore,
628
U32 rep[ZSTD_REP_NUM],
629
const void* src, size_t srcSize,
630
const searchMethod_e searchMethod, const U32 depth,
631
ZSTD_dictMode_e const dictMode)
632
{
633
const BYTE* const istart = (const BYTE*)src;
634
const BYTE* ip = istart;
635
const BYTE* anchor = istart;
636
const BYTE* const iend = istart + srcSize;
637
const BYTE* const ilimit = iend - 8;
638
const BYTE* const base = ms->window.base;
639
const U32 prefixLowestIndex = ms->window.dictLimit;
640
const BYTE* const prefixLowest = base + prefixLowestIndex;
641
642
typedef size_t (*searchMax_f)(
643
ZSTD_matchState_t* ms,
644
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
645
searchMax_f const searchMax = dictMode == ZSTD_dictMatchState ?
646
(searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_dictMatchState_selectMLS
647
: ZSTD_HcFindBestMatch_dictMatchState_selectMLS) :
648
(searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_selectMLS
649
: ZSTD_HcFindBestMatch_selectMLS);
650
U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
651
652
const ZSTD_matchState_t* const dms = ms->dictMatchState;
653
const U32 dictLowestIndex = dictMode == ZSTD_dictMatchState ?
654
dms->window.dictLimit : 0;
655
const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
656
dms->window.base : NULL;
657
const BYTE* const dictLowest = dictMode == ZSTD_dictMatchState ?
658
dictBase + dictLowestIndex : NULL;
659
const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
660
dms->window.nextSrc : NULL;
661
const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
662
prefixLowestIndex - (U32)(dictEnd - dictBase) :
663
0;
664
const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
665
666
DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u)", (U32)dictMode);
667
668
/* init */
669
ip += (dictAndPrefixLength == 0);
670
if (dictMode == ZSTD_noDict) {
671
U32 const current = (U32)(ip - base);
672
U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, ms->cParams.windowLog);
673
U32 const maxRep = current - windowLow;
674
if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
675
if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
676
}
677
if (dictMode == ZSTD_dictMatchState) {
678
/* dictMatchState repCode checks don't currently handle repCode == 0
679
* disabling. */
680
assert(offset_1 <= dictAndPrefixLength);
681
assert(offset_2 <= dictAndPrefixLength);
682
}
683
684
/* Match Loop */
685
#if defined(__GNUC__) && defined(__x86_64__)
686
/* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
687
* code alignment is perturbed. To fix the instability align the loop on 32-bytes.
688
*/
689
__asm__(".p2align 5");
690
#endif
691
while (ip < ilimit) {
692
size_t matchLength=0;
693
size_t offset=0;
694
const BYTE* start=ip+1;
695
696
/* check repCode */
697
if (dictMode == ZSTD_dictMatchState) {
698
const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
699
const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
700
&& repIndex < prefixLowestIndex) ?
701
dictBase + (repIndex - dictIndexDelta) :
702
base + repIndex;
703
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
704
&& (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
705
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
706
matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
707
if (depth==0) goto _storeSequence;
708
}
709
}
710
if ( dictMode == ZSTD_noDict
711
&& ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
712
matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
713
if (depth==0) goto _storeSequence;
714
}
715
716
/* first search (depth 0) */
717
{ size_t offsetFound = 999999999;
718
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
719
if (ml2 > matchLength)
720
matchLength = ml2, start = ip, offset=offsetFound;
721
}
722
723
if (matchLength < 4) {
724
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
725
continue;
726
}
727
728
/* let's try to find a better solution */
729
if (depth>=1)
730
while (ip<ilimit) {
731
ip ++;
732
if ( (dictMode == ZSTD_noDict)
733
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
734
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
735
int const gain2 = (int)(mlRep * 3);
736
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
737
if ((mlRep >= 4) && (gain2 > gain1))
738
matchLength = mlRep, offset = 0, start = ip;
739
}
740
if (dictMode == ZSTD_dictMatchState) {
741
const U32 repIndex = (U32)(ip - base) - offset_1;
742
const BYTE* repMatch = repIndex < prefixLowestIndex ?
743
dictBase + (repIndex - dictIndexDelta) :
744
base + repIndex;
745
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
746
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
747
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
748
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
749
int const gain2 = (int)(mlRep * 3);
750
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
751
if ((mlRep >= 4) && (gain2 > gain1))
752
matchLength = mlRep, offset = 0, start = ip;
753
}
754
}
755
{ size_t offset2=999999999;
756
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
757
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
758
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
759
if ((ml2 >= 4) && (gain2 > gain1)) {
760
matchLength = ml2, offset = offset2, start = ip;
761
continue; /* search a better one */
762
} }
763
764
/* let's find an even better one */
765
if ((depth==2) && (ip<ilimit)) {
766
ip ++;
767
if ( (dictMode == ZSTD_noDict)
768
&& (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
769
size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
770
int const gain2 = (int)(mlRep * 4);
771
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
772
if ((mlRep >= 4) && (gain2 > gain1))
773
matchLength = mlRep, offset = 0, start = ip;
774
}
775
if (dictMode == ZSTD_dictMatchState) {
776
const U32 repIndex = (U32)(ip - base) - offset_1;
777
const BYTE* repMatch = repIndex < prefixLowestIndex ?
778
dictBase + (repIndex - dictIndexDelta) :
779
base + repIndex;
780
if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
781
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
782
const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
783
size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
784
int const gain2 = (int)(mlRep * 4);
785
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
786
if ((mlRep >= 4) && (gain2 > gain1))
787
matchLength = mlRep, offset = 0, start = ip;
788
}
789
}
790
{ size_t offset2=999999999;
791
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
792
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
793
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
794
if ((ml2 >= 4) && (gain2 > gain1)) {
795
matchLength = ml2, offset = offset2, start = ip;
796
continue;
797
} } }
798
break; /* nothing found : store previous solution */
799
}
800
801
/* NOTE:
802
* start[-offset+ZSTD_REP_MOVE-1] is undefined behavior.
803
* (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which
804
* overflows the pointer, which is undefined behavior.
805
*/
806
/* catch up */
807
if (offset) {
808
if (dictMode == ZSTD_noDict) {
809
while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
810
&& (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) ) /* only search for offset within prefix */
811
{ start--; matchLength++; }
812
}
813
if (dictMode == ZSTD_dictMatchState) {
814
U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
815
const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
816
const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
817
while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
818
}
819
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
820
}
821
/* store sequence */
822
_storeSequence:
823
{ size_t const litLength = start - anchor;
824
ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
825
anchor = ip = start + matchLength;
826
}
827
828
/* check immediate repcode */
829
if (dictMode == ZSTD_dictMatchState) {
830
while (ip <= ilimit) {
831
U32 const current2 = (U32)(ip-base);
832
U32 const repIndex = current2 - offset_2;
833
const BYTE* repMatch = dictMode == ZSTD_dictMatchState
834
&& repIndex < prefixLowestIndex ?
835
dictBase - dictIndexDelta + repIndex :
836
base + repIndex;
837
if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
838
&& (MEM_read32(repMatch) == MEM_read32(ip)) ) {
839
const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
840
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
841
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset_2 <=> offset_1 */
842
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
843
ip += matchLength;
844
anchor = ip;
845
continue;
846
}
847
break;
848
}
849
}
850
851
if (dictMode == ZSTD_noDict) {
852
while ( ((ip <= ilimit) & (offset_2>0))
853
&& (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
854
/* store sequence */
855
matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
856
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
857
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
858
ip += matchLength;
859
anchor = ip;
860
continue; /* faster when present ... (?) */
861
} } }
862
863
/* Save reps for next block */
864
rep[0] = offset_1 ? offset_1 : savedOffset;
865
rep[1] = offset_2 ? offset_2 : savedOffset;
866
867
/* Return the last literals size */
868
return (size_t)(iend - anchor);
869
}
870
871
872
size_t ZSTD_compressBlock_btlazy2(
873
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
874
void const* src, size_t srcSize)
875
{
876
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_noDict);
877
}
878
879
size_t ZSTD_compressBlock_lazy2(
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, search_hashChain, 2, ZSTD_noDict);
884
}
885
886
size_t ZSTD_compressBlock_lazy(
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, search_hashChain, 1, ZSTD_noDict);
891
}
892
893
size_t ZSTD_compressBlock_greedy(
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, search_hashChain, 0, ZSTD_noDict);
898
}
899
900
size_t ZSTD_compressBlock_btlazy2_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, search_binaryTree, 2, ZSTD_dictMatchState);
905
}
906
907
size_t ZSTD_compressBlock_lazy2_dictMatchState(
908
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
909
void const* src, size_t srcSize)
910
{
911
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dictMatchState);
912
}
913
914
size_t ZSTD_compressBlock_lazy_dictMatchState(
915
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
916
void const* src, size_t srcSize)
917
{
918
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dictMatchState);
919
}
920
921
size_t ZSTD_compressBlock_greedy_dictMatchState(
922
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
923
void const* src, size_t srcSize)
924
{
925
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dictMatchState);
926
}
927
928
929
FORCE_INLINE_TEMPLATE
930
size_t ZSTD_compressBlock_lazy_extDict_generic(
931
ZSTD_matchState_t* ms, seqStore_t* seqStore,
932
U32 rep[ZSTD_REP_NUM],
933
const void* src, size_t srcSize,
934
const searchMethod_e searchMethod, const U32 depth)
935
{
936
const BYTE* const istart = (const BYTE*)src;
937
const BYTE* ip = istart;
938
const BYTE* anchor = istart;
939
const BYTE* const iend = istart + srcSize;
940
const BYTE* const ilimit = iend - 8;
941
const BYTE* const base = ms->window.base;
942
const U32 dictLimit = ms->window.dictLimit;
943
const BYTE* const prefixStart = base + dictLimit;
944
const BYTE* const dictBase = ms->window.dictBase;
945
const BYTE* const dictEnd = dictBase + dictLimit;
946
const BYTE* const dictStart = dictBase + ms->window.lowLimit;
947
const U32 windowLog = ms->cParams.windowLog;
948
949
typedef size_t (*searchMax_f)(
950
ZSTD_matchState_t* ms,
951
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
952
searchMax_f searchMax = searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
953
954
U32 offset_1 = rep[0], offset_2 = rep[1];
955
956
DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic");
957
958
/* init */
959
ip += (ip == prefixStart);
960
961
/* Match Loop */
962
#if defined(__GNUC__) && defined(__x86_64__)
963
/* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
964
* code alignment is perturbed. To fix the instability align the loop on 32-bytes.
965
*/
966
__asm__(".p2align 5");
967
#endif
968
while (ip < ilimit) {
969
size_t matchLength=0;
970
size_t offset=0;
971
const BYTE* start=ip+1;
972
U32 current = (U32)(ip-base);
973
974
/* check repCode */
975
{ const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current+1, windowLog);
976
const U32 repIndex = (U32)(current+1 - offset_1);
977
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
978
const BYTE* const repMatch = repBase + repIndex;
979
if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow */
980
& (offset_1 < current+1 - windowLow) ) /* note: we are searching at current+1 */
981
if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
982
/* repcode detected we should take it */
983
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
984
matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
985
if (depth==0) goto _storeSequence;
986
} }
987
988
/* first search (depth 0) */
989
{ size_t offsetFound = 999999999;
990
size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
991
if (ml2 > matchLength)
992
matchLength = ml2, start = ip, offset=offsetFound;
993
}
994
995
if (matchLength < 4) {
996
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
997
continue;
998
}
999
1000
/* let's try to find a better solution */
1001
if (depth>=1)
1002
while (ip<ilimit) {
1003
ip ++;
1004
current++;
1005
/* check repCode */
1006
if (offset) {
1007
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
1008
const U32 repIndex = (U32)(current - offset_1);
1009
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1010
const BYTE* const repMatch = repBase + repIndex;
1011
if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
1012
& (offset_1 < current - windowLow) ) /* equivalent to `current > repIndex >= windowLow` */
1013
if (MEM_read32(ip) == MEM_read32(repMatch)) {
1014
/* repcode detected */
1015
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1016
size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1017
int const gain2 = (int)(repLength * 3);
1018
int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
1019
if ((repLength >= 4) && (gain2 > gain1))
1020
matchLength = repLength, offset = 0, start = ip;
1021
} }
1022
1023
/* search match, depth 1 */
1024
{ size_t offset2=999999999;
1025
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1026
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
1027
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
1028
if ((ml2 >= 4) && (gain2 > gain1)) {
1029
matchLength = ml2, offset = offset2, start = ip;
1030
continue; /* search a better one */
1031
} }
1032
1033
/* let's find an even better one */
1034
if ((depth==2) && (ip<ilimit)) {
1035
ip ++;
1036
current++;
1037
/* check repCode */
1038
if (offset) {
1039
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, current, windowLog);
1040
const U32 repIndex = (U32)(current - offset_1);
1041
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1042
const BYTE* const repMatch = repBase + repIndex;
1043
if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
1044
& (offset_1 < current - windowLow) ) /* equivalent to `current > repIndex >= windowLow` */
1045
if (MEM_read32(ip) == MEM_read32(repMatch)) {
1046
/* repcode detected */
1047
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1048
size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1049
int const gain2 = (int)(repLength * 4);
1050
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
1051
if ((repLength >= 4) && (gain2 > gain1))
1052
matchLength = repLength, offset = 0, start = ip;
1053
} }
1054
1055
/* search match, depth 2 */
1056
{ size_t offset2=999999999;
1057
size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1058
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
1059
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
1060
if ((ml2 >= 4) && (gain2 > gain1)) {
1061
matchLength = ml2, offset = offset2, start = ip;
1062
continue;
1063
} } }
1064
break; /* nothing found : store previous solution */
1065
}
1066
1067
/* catch up */
1068
if (offset) {
1069
U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
1070
const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
1071
const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
1072
while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; } /* catch up */
1073
offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
1074
}
1075
1076
/* store sequence */
1077
_storeSequence:
1078
{ size_t const litLength = start - anchor;
1079
ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
1080
anchor = ip = start + matchLength;
1081
}
1082
1083
/* check immediate repcode */
1084
while (ip <= ilimit) {
1085
const U32 repCurrent = (U32)(ip-base);
1086
const U32 windowLow = ZSTD_getLowestMatchIndex(ms, repCurrent, windowLog);
1087
const U32 repIndex = repCurrent - offset_2;
1088
const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1089
const BYTE* const repMatch = repBase + repIndex;
1090
if ( ((U32)((dictLimit-1) - repIndex) >= 3) /* intentional overflow : do not test positions overlapping 2 memory segments */
1091
& (offset_2 < repCurrent - windowLow) ) /* equivalent to `curr > repIndex >= windowLow` */
1092
if (MEM_read32(ip) == MEM_read32(repMatch)) {
1093
/* repcode detected we should take it */
1094
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1095
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1096
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */
1097
ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
1098
ip += matchLength;
1099
anchor = ip;
1100
continue; /* faster when present ... (?) */
1101
}
1102
break;
1103
} }
1104
1105
/* Save reps for next block */
1106
rep[0] = offset_1;
1107
rep[1] = offset_2;
1108
1109
/* Return the last literals size */
1110
return (size_t)(iend - anchor);
1111
}
1112
1113
1114
size_t ZSTD_compressBlock_greedy_extDict(
1115
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1116
void const* src, size_t srcSize)
1117
{
1118
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0);
1119
}
1120
1121
size_t ZSTD_compressBlock_lazy_extDict(
1122
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1123
void const* src, size_t srcSize)
1124
1125
{
1126
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1);
1127
}
1128
1129
size_t ZSTD_compressBlock_lazy2_extDict(
1130
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1131
void const* src, size_t srcSize)
1132
1133
{
1134
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2);
1135
}
1136
1137
size_t ZSTD_compressBlock_btlazy2_extDict(
1138
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1139
void const* src, size_t srcSize)
1140
1141
{
1142
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2);
1143
}
1144
1145