xref: /openbmc/linux/lib/zstd/compress/zstd_lazy.c (revision 647d41d3)
1 /*
2  * Copyright (c) Yann Collet, Facebook, Inc.
3  * All rights reserved.
4  *
5  * This source code is licensed under both the BSD-style license (found in the
6  * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7  * in the COPYING file in the root directory of this source tree).
8  * You may select, at your option, one of the above-listed licenses.
9  */
10 
11 #include "zstd_compress_internal.h"
12 #include "zstd_lazy.h"
13 
14 
15 /*-*************************************
16 *  Binary Tree search
17 ***************************************/
18 
19 static void
20 ZSTD_updateDUBT(ZSTD_matchState_t* ms,
21                 const BYTE* ip, const BYTE* iend,
22                 U32 mls)
23 {
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 : curr >= btlow == (curr - btmask)
62  *  doesn't fail */
63 static void
64 ZSTD_insertDUBT1(ZSTD_matchState_t* ms,
65                  U32 curr, 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;
73     size_t commonLengthSmaller=0, commonLengthLarger=0;
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 = (curr>=dictLimit) ? base + curr : dictBase + curr;
78     const BYTE* const iend = (curr>=dictLimit) ? inputEnd : dictBase + dictLimit;
79     const BYTE* const dictEnd = dictBase + dictLimit;
80     const BYTE* const prefixStart = base + dictLimit;
81     const BYTE* match;
82     U32* smallerPtr = bt + 2*(curr&btMask);
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) */
85     U32 dummy32;   /* to be nullified at the end */
86     U32 const windowValid = ms->window.lowLimit;
87     U32 const maxDistance = 1U << cParams->windowLog;
88     U32 const windowLow = (curr - windowValid > maxDistance) ? curr - maxDistance : windowValid;
89 
90 
91     DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
92                 curr, dictLimit, windowLow);
93     assert(curr >= btLow);
94     assert(ip < iend);   /* condition for ZSTD_count */
95 
96     for (; nbCompares && (matchIndex > windowLow); --nbCompares) {
97         U32* const nextPtr = bt + 2*(matchIndex & btMask);
98         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
99         assert(matchIndex < curr);
100         /* note : all candidates are now supposed sorted,
101          * but it's still possible to have nextPtr[1] == ZSTD_DUBT_UNSORTED_MARK
102          * when a real index has the same value as ZSTD_DUBT_UNSORTED_MARK */
103 
104         if ( (dictMode != ZSTD_extDict)
105           || (matchIndex+matchLength >= dictLimit)  /* both in current segment*/
106           || (curr < dictLimit) /* both in extDict */) {
107             const BYTE* const mBase = ( (dictMode != ZSTD_extDict)
108                                      || (matchIndex+matchLength >= dictLimit)) ?
109                                         base : dictBase;
110             assert( (matchIndex+matchLength >= dictLimit)   /* might be wrong if extDict is incorrectly set to 0 */
111                  || (curr < dictLimit) );
112             match = mBase + matchIndex;
113             matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
114         } else {
115             match = dictBase + matchIndex;
116             matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
117             if (matchIndex+matchLength >= dictLimit)
118                 match = base + matchIndex;   /* preparation for next read of match[matchLength] */
119         }
120 
121         DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
122                     curr, matchIndex, (U32)matchLength);
123 
124         if (ip+matchLength == iend) {   /* equal : no way to know if inf or sup */
125             break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
126         }
127 
128         if (match[matchLength] < ip[matchLength]) {  /* necessarily within buffer */
129             /* match is smaller than current */
130             *smallerPtr = matchIndex;             /* update smaller idx */
131             commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
132             if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
133             DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
134                         matchIndex, btLow, nextPtr[1]);
135             smallerPtr = nextPtr+1;               /* new "candidate" => larger than match, which was smaller than target */
136             matchIndex = nextPtr[1];              /* new matchIndex, larger than previous and closer to current */
137         } else {
138             /* match is larger than current */
139             *largerPtr = matchIndex;
140             commonLengthLarger = matchLength;
141             if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
142             DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
143                         matchIndex, btLow, nextPtr[0]);
144             largerPtr = nextPtr;
145             matchIndex = nextPtr[0];
146     }   }
147 
148     *smallerPtr = *largerPtr = 0;
149 }
150 
151 
152 static size_t
153 ZSTD_DUBT_findBetterDictMatch (
154         ZSTD_matchState_t* ms,
155         const BYTE* const ip, const BYTE* const iend,
156         size_t* offsetPtr,
157         size_t bestLength,
158         U32 nbCompares,
159         U32 const mls,
160         const ZSTD_dictMode_e dictMode)
161 {
162     const ZSTD_matchState_t * const dms = ms->dictMatchState;
163     const ZSTD_compressionParameters* const dmsCParams = &dms->cParams;
164     const U32 * const dictHashTable = dms->hashTable;
165     U32         const hashLog = dmsCParams->hashLog;
166     size_t      const h  = ZSTD_hashPtr(ip, hashLog, mls);
167     U32               dictMatchIndex = dictHashTable[h];
168 
169     const BYTE* const base = ms->window.base;
170     const BYTE* const prefixStart = base + ms->window.dictLimit;
171     U32         const curr = (U32)(ip-base);
172     const BYTE* const dictBase = dms->window.base;
173     const BYTE* const dictEnd = dms->window.nextSrc;
174     U32         const dictHighLimit = (U32)(dms->window.nextSrc - dms->window.base);
175     U32         const dictLowLimit = dms->window.lowLimit;
176     U32         const dictIndexDelta = ms->window.lowLimit - dictHighLimit;
177 
178     U32*        const dictBt = dms->chainTable;
179     U32         const btLog  = dmsCParams->chainLog - 1;
180     U32         const btMask = (1 << btLog) - 1;
181     U32         const btLow = (btMask >= dictHighLimit - dictLowLimit) ? dictLowLimit : dictHighLimit - btMask;
182 
183     size_t commonLengthSmaller=0, commonLengthLarger=0;
184 
185     (void)dictMode;
186     assert(dictMode == ZSTD_dictMatchState);
187 
188     for (; nbCompares && (dictMatchIndex > dictLowLimit); --nbCompares) {
189         U32* const nextPtr = dictBt + 2*(dictMatchIndex & btMask);
190         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
191         const BYTE* match = dictBase + dictMatchIndex;
192         matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
193         if (dictMatchIndex+matchLength >= dictHighLimit)
194             match = base + dictMatchIndex + dictIndexDelta;   /* to prepare for next usage of match[matchLength] */
195 
196         if (matchLength > bestLength) {
197             U32 matchIndex = dictMatchIndex + dictIndexDelta;
198             if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) ) {
199                 DEBUGLOG(9, "ZSTD_DUBT_findBetterDictMatch(%u) : found better match length %u -> %u and offsetCode %u -> %u (dictMatchIndex %u, matchIndex %u)",
200                     curr, (U32)bestLength, (U32)matchLength, (U32)*offsetPtr, ZSTD_REP_MOVE + curr - matchIndex, dictMatchIndex, matchIndex);
201                 bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + curr - matchIndex;
202             }
203             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 */
204                 break;   /* drop, to guarantee consistency (miss a little bit of compression) */
205             }
206         }
207 
208         if (match[matchLength] < ip[matchLength]) {
209             if (dictMatchIndex <= btLow) { break; }   /* beyond tree size, stop the search */
210             commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
211             dictMatchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
212         } else {
213             /* match is larger than current */
214             if (dictMatchIndex <= btLow) { break; }   /* beyond tree size, stop the search */
215             commonLengthLarger = matchLength;
216             dictMatchIndex = nextPtr[0];
217         }
218     }
219 
220     if (bestLength >= MINMATCH) {
221         U32 const mIndex = curr - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
222         DEBUGLOG(8, "ZSTD_DUBT_findBetterDictMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
223                     curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
224     }
225     return bestLength;
226 
227 }
228 
229 
230 static size_t
231 ZSTD_DUBT_findBestMatch(ZSTD_matchState_t* ms,
232                         const BYTE* const ip, const BYTE* const iend,
233                         size_t* offsetPtr,
234                         U32 const mls,
235                         const ZSTD_dictMode_e dictMode)
236 {
237     const ZSTD_compressionParameters* const cParams = &ms->cParams;
238     U32*   const hashTable = ms->hashTable;
239     U32    const hashLog = cParams->hashLog;
240     size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
241     U32          matchIndex  = hashTable[h];
242 
243     const BYTE* const base = ms->window.base;
244     U32    const curr = (U32)(ip-base);
245     U32    const windowLow = ZSTD_getLowestMatchIndex(ms, curr, cParams->windowLog);
246 
247     U32*   const bt = ms->chainTable;
248     U32    const btLog  = cParams->chainLog - 1;
249     U32    const btMask = (1 << btLog) - 1;
250     U32    const btLow = (btMask >= curr) ? 0 : curr - btMask;
251     U32    const unsortLimit = MAX(btLow, windowLow);
252 
253     U32*         nextCandidate = bt + 2*(matchIndex&btMask);
254     U32*         unsortedMark = bt + 2*(matchIndex&btMask) + 1;
255     U32          nbCompares = 1U << cParams->searchLog;
256     U32          nbCandidates = nbCompares;
257     U32          previousCandidate = 0;
258 
259     DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", curr);
260     assert(ip <= iend-8);   /* required for h calculation */
261     assert(dictMode != ZSTD_dedicatedDictSearch);
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*(curr&btMask);
304         U32* largerPtr  = bt + 2*(curr&btMask) + 1;
305         U32 matchEndIdx = curr + 8 + 1;
306         U32 dummy32;   /* to be nullified at the end */
307         size_t bestLength = 0;
308 
309         matchIndex  = hashTable[h];
310         hashTable[h] = curr;   /* Update Hash Table */
311 
312         for (; nbCompares && (matchIndex > windowLow); --nbCompares) {
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(curr-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
331                     bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + curr - 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         assert(nbCompares <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
361         if (dictMode == ZSTD_dictMatchState && nbCompares) {
362             bestLength = ZSTD_DUBT_findBetterDictMatch(
363                     ms, ip, iend,
364                     offsetPtr, bestLength, nbCompares,
365                     mls, dictMode);
366         }
367 
368         assert(matchEndIdx > curr+8); /* ensure nextToUpdate is increased */
369         ms->nextToUpdate = matchEndIdx - 8;   /* skip repetitive patterns */
370         if (bestLength >= MINMATCH) {
371             U32 const mIndex = curr - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
372             DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
373                         curr, (U32)bestLength, (U32)*offsetPtr, mIndex);
374         }
375         return bestLength;
376     }
377 }
378 
379 
380 /* ZSTD_BtFindBestMatch() : Tree updater, providing best match */
381 FORCE_INLINE_TEMPLATE size_t
382 ZSTD_BtFindBestMatch( ZSTD_matchState_t* ms,
383                 const BYTE* const ip, const BYTE* const iLimit,
384                       size_t* offsetPtr,
385                 const U32 mls /* template */,
386                 const ZSTD_dictMode_e dictMode)
387 {
388     DEBUGLOG(7, "ZSTD_BtFindBestMatch");
389     if (ip < ms->window.base + ms->nextToUpdate) return 0;   /* skipped area */
390     ZSTD_updateDUBT(ms, ip, iLimit, mls);
391     return ZSTD_DUBT_findBestMatch(ms, ip, iLimit, offsetPtr, mls, dictMode);
392 }
393 
394 
395 static size_t
396 ZSTD_BtFindBestMatch_selectMLS (  ZSTD_matchState_t* ms,
397                             const BYTE* ip, const BYTE* const iLimit,
398                                   size_t* offsetPtr)
399 {
400     switch(ms->cParams.minMatch)
401     {
402     default : /* includes case 3 */
403     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
404     case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
405     case 7 :
406     case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
407     }
408 }
409 
410 
411 static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
412                         ZSTD_matchState_t* ms,
413                         const BYTE* ip, const BYTE* const iLimit,
414                         size_t* offsetPtr)
415 {
416     switch(ms->cParams.minMatch)
417     {
418     default : /* includes case 3 */
419     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
420     case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
421     case 7 :
422     case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
423     }
424 }
425 
426 
427 static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
428                         ZSTD_matchState_t* ms,
429                         const BYTE* ip, const BYTE* const iLimit,
430                         size_t* offsetPtr)
431 {
432     switch(ms->cParams.minMatch)
433     {
434     default : /* includes case 3 */
435     case 4 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
436     case 5 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
437     case 7 :
438     case 6 : return ZSTD_BtFindBestMatch(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
439     }
440 }
441 
442 
443 
444 /* *********************************
445 *  Hash Chain
446 ***********************************/
447 #define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & (mask)]
448 
449 /* Update chains up to ip (excluded)
450    Assumption : always within prefix (i.e. not within extDict) */
451 FORCE_INLINE_TEMPLATE U32 ZSTD_insertAndFindFirstIndex_internal(
452                         ZSTD_matchState_t* ms,
453                         const ZSTD_compressionParameters* const cParams,
454                         const BYTE* ip, U32 const mls)
455 {
456     U32* const hashTable  = ms->hashTable;
457     const U32 hashLog = cParams->hashLog;
458     U32* const chainTable = ms->chainTable;
459     const U32 chainMask = (1 << cParams->chainLog) - 1;
460     const BYTE* const base = ms->window.base;
461     const U32 target = (U32)(ip - base);
462     U32 idx = ms->nextToUpdate;
463 
464     while(idx < target) { /* catch up */
465         size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
466         NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
467         hashTable[h] = idx;
468         idx++;
469     }
470 
471     ms->nextToUpdate = target;
472     return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
473 }
474 
475 U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip) {
476     const ZSTD_compressionParameters* const cParams = &ms->cParams;
477     return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, ms->cParams.minMatch);
478 }
479 
480 void ZSTD_dedicatedDictSearch_lazy_loadDictionary(ZSTD_matchState_t* ms, const BYTE* const ip)
481 {
482     const BYTE* const base = ms->window.base;
483     U32 const target = (U32)(ip - base);
484     U32* const hashTable = ms->hashTable;
485     U32* const chainTable = ms->chainTable;
486     U32 const chainSize = 1 << ms->cParams.chainLog;
487     U32 idx = ms->nextToUpdate;
488     U32 const minChain = chainSize < target ? target - chainSize : idx;
489     U32 const bucketSize = 1 << ZSTD_LAZY_DDSS_BUCKET_LOG;
490     U32 const cacheSize = bucketSize - 1;
491     U32 const chainAttempts = (1 << ms->cParams.searchLog) - cacheSize;
492     U32 const chainLimit = chainAttempts > 255 ? 255 : chainAttempts;
493 
494     /* We know the hashtable is oversized by a factor of `bucketSize`.
495      * We are going to temporarily pretend `bucketSize == 1`, keeping only a
496      * single entry. We will use the rest of the space to construct a temporary
497      * chaintable.
498      */
499     U32 const hashLog = ms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG;
500     U32* const tmpHashTable = hashTable;
501     U32* const tmpChainTable = hashTable + ((size_t)1 << hashLog);
502     U32 const tmpChainSize = ((1 << ZSTD_LAZY_DDSS_BUCKET_LOG) - 1) << hashLog;
503     U32 const tmpMinChain = tmpChainSize < target ? target - tmpChainSize : idx;
504 
505     U32 hashIdx;
506 
507     assert(ms->cParams.chainLog <= 24);
508     assert(ms->cParams.hashLog >= ms->cParams.chainLog);
509     assert(idx != 0);
510     assert(tmpMinChain <= minChain);
511 
512     /* fill conventional hash table and conventional chain table */
513     for ( ; idx < target; idx++) {
514         U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch);
515         if (idx >= tmpMinChain) {
516             tmpChainTable[idx - tmpMinChain] = hashTable[h];
517         }
518         tmpHashTable[h] = idx;
519     }
520 
521     /* sort chains into ddss chain table */
522     {
523         U32 chainPos = 0;
524         for (hashIdx = 0; hashIdx < (1U << hashLog); hashIdx++) {
525             U32 count;
526             U32 countBeyondMinChain = 0;
527             U32 i = tmpHashTable[hashIdx];
528             for (count = 0; i >= tmpMinChain && count < cacheSize; count++) {
529                 /* skip through the chain to the first position that won't be
530                  * in the hash cache bucket */
531                 if (i < minChain) {
532                     countBeyondMinChain++;
533                 }
534                 i = tmpChainTable[i - tmpMinChain];
535             }
536             if (count == cacheSize) {
537                 for (count = 0; count < chainLimit;) {
538                     if (i < minChain) {
539                         if (!i || countBeyondMinChain++ > cacheSize) {
540                             /* only allow pulling `cacheSize` number of entries
541                              * into the cache or chainTable beyond `minChain`,
542                              * to replace the entries pulled out of the
543                              * chainTable into the cache. This lets us reach
544                              * back further without increasing the total number
545                              * of entries in the chainTable, guaranteeing the
546                              * DDSS chain table will fit into the space
547                              * allocated for the regular one. */
548                             break;
549                         }
550                     }
551                     chainTable[chainPos++] = i;
552                     count++;
553                     if (i < tmpMinChain) {
554                         break;
555                     }
556                     i = tmpChainTable[i - tmpMinChain];
557                 }
558             } else {
559                 count = 0;
560             }
561             if (count) {
562                 tmpHashTable[hashIdx] = ((chainPos - count) << 8) + count;
563             } else {
564                 tmpHashTable[hashIdx] = 0;
565             }
566         }
567         assert(chainPos <= chainSize); /* I believe this is guaranteed... */
568     }
569 
570     /* move chain pointers into the last entry of each hash bucket */
571     for (hashIdx = (1 << hashLog); hashIdx; ) {
572         U32 const bucketIdx = --hashIdx << ZSTD_LAZY_DDSS_BUCKET_LOG;
573         U32 const chainPackedPointer = tmpHashTable[hashIdx];
574         U32 i;
575         for (i = 0; i < cacheSize; i++) {
576             hashTable[bucketIdx + i] = 0;
577         }
578         hashTable[bucketIdx + bucketSize - 1] = chainPackedPointer;
579     }
580 
581     /* fill the buckets of the hash table */
582     for (idx = ms->nextToUpdate; idx < target; idx++) {
583         U32 const h = (U32)ZSTD_hashPtr(base + idx, hashLog, ms->cParams.minMatch)
584                    << ZSTD_LAZY_DDSS_BUCKET_LOG;
585         U32 i;
586         /* Shift hash cache down 1. */
587         for (i = cacheSize - 1; i; i--)
588             hashTable[h + i] = hashTable[h + i - 1];
589         hashTable[h] = idx;
590     }
591 
592     ms->nextToUpdate = target;
593 }
594 
595 
596 /* inlining is important to hardwire a hot branch (template emulation) */
597 FORCE_INLINE_TEMPLATE
598 size_t ZSTD_HcFindBestMatch_generic (
599                         ZSTD_matchState_t* ms,
600                         const BYTE* const ip, const BYTE* const iLimit,
601                         size_t* offsetPtr,
602                         const U32 mls, const ZSTD_dictMode_e dictMode)
603 {
604     const ZSTD_compressionParameters* const cParams = &ms->cParams;
605     U32* const chainTable = ms->chainTable;
606     const U32 chainSize = (1 << cParams->chainLog);
607     const U32 chainMask = chainSize-1;
608     const BYTE* const base = ms->window.base;
609     const BYTE* const dictBase = ms->window.dictBase;
610     const U32 dictLimit = ms->window.dictLimit;
611     const BYTE* const prefixStart = base + dictLimit;
612     const BYTE* const dictEnd = dictBase + dictLimit;
613     const U32 curr = (U32)(ip-base);
614     const U32 maxDistance = 1U << cParams->windowLog;
615     const U32 lowestValid = ms->window.lowLimit;
616     const U32 withinMaxDistance = (curr - lowestValid > maxDistance) ? curr - maxDistance : lowestValid;
617     const U32 isDictionary = (ms->loadedDictEnd != 0);
618     const U32 lowLimit = isDictionary ? lowestValid : withinMaxDistance;
619     const U32 minChain = curr > chainSize ? curr - chainSize : 0;
620     U32 nbAttempts = 1U << cParams->searchLog;
621     size_t ml=4-1;
622 
623     const ZSTD_matchState_t* const dms = ms->dictMatchState;
624     const U32 ddsHashLog = dictMode == ZSTD_dedicatedDictSearch
625                          ? dms->cParams.hashLog - ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
626     const size_t ddsIdx = dictMode == ZSTD_dedicatedDictSearch
627                         ? ZSTD_hashPtr(ip, ddsHashLog, mls) << ZSTD_LAZY_DDSS_BUCKET_LOG : 0;
628 
629     U32 matchIndex;
630 
631     if (dictMode == ZSTD_dedicatedDictSearch) {
632         const U32* entry = &dms->hashTable[ddsIdx];
633         PREFETCH_L1(entry);
634     }
635 
636     /* HC4 match finder */
637     matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
638 
639     for ( ; (matchIndex>=lowLimit) & (nbAttempts>0) ; nbAttempts--) {
640         size_t currentMl=0;
641         if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
642             const BYTE* const match = base + matchIndex;
643             assert(matchIndex >= dictLimit);   /* ensures this is true if dictMode != ZSTD_extDict */
644             if (match[ml] == ip[ml])   /* potentially better */
645                 currentMl = ZSTD_count(ip, match, iLimit);
646         } else {
647             const BYTE* const match = dictBase + matchIndex;
648             assert(match+4 <= dictEnd);
649             if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
650                 currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
651         }
652 
653         /* save best solution */
654         if (currentMl > ml) {
655             ml = currentMl;
656             *offsetPtr = curr - matchIndex + ZSTD_REP_MOVE;
657             if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
658         }
659 
660         if (matchIndex <= minChain) break;
661         matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
662     }
663 
664     assert(nbAttempts <= (1U << ZSTD_SEARCHLOG_MAX)); /* Check we haven't underflowed. */
665     if (dictMode == ZSTD_dedicatedDictSearch) {
666         const U32 ddsLowestIndex  = dms->window.dictLimit;
667         const BYTE* const ddsBase = dms->window.base;
668         const BYTE* const ddsEnd  = dms->window.nextSrc;
669         const U32 ddsSize         = (U32)(ddsEnd - ddsBase);
670         const U32 ddsIndexDelta   = dictLimit - ddsSize;
671         const U32 bucketSize      = (1 << ZSTD_LAZY_DDSS_BUCKET_LOG);
672         const U32 bucketLimit     = nbAttempts < bucketSize - 1 ? nbAttempts : bucketSize - 1;
673         U32 ddsAttempt;
674 
675         for (ddsAttempt = 0; ddsAttempt < bucketSize - 1; ddsAttempt++) {
676             PREFETCH_L1(ddsBase + dms->hashTable[ddsIdx + ddsAttempt]);
677         }
678 
679         {
680             U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
681             U32 const chainIndex = chainPackedPointer >> 8;
682 
683             PREFETCH_L1(&dms->chainTable[chainIndex]);
684         }
685 
686         for (ddsAttempt = 0; ddsAttempt < bucketLimit; ddsAttempt++) {
687             size_t currentMl=0;
688             const BYTE* match;
689             matchIndex = dms->hashTable[ddsIdx + ddsAttempt];
690             match = ddsBase + matchIndex;
691 
692             if (!matchIndex) {
693                 return ml;
694             }
695 
696             /* guaranteed by table construction */
697             (void)ddsLowestIndex;
698             assert(matchIndex >= ddsLowestIndex);
699             assert(match+4 <= ddsEnd);
700             if (MEM_read32(match) == MEM_read32(ip)) {
701                 /* assumption : matchIndex <= dictLimit-4 (by table construction) */
702                 currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
703             }
704 
705             /* save best solution */
706             if (currentMl > ml) {
707                 ml = currentMl;
708                 *offsetPtr = curr - (matchIndex + ddsIndexDelta) + ZSTD_REP_MOVE;
709                 if (ip+currentMl == iLimit) {
710                     /* best possible, avoids read overflow on next attempt */
711                     return ml;
712                 }
713             }
714         }
715 
716         {
717             U32 const chainPackedPointer = dms->hashTable[ddsIdx + bucketSize - 1];
718             U32 chainIndex = chainPackedPointer >> 8;
719             U32 const chainLength = chainPackedPointer & 0xFF;
720             U32 const chainAttempts = nbAttempts - ddsAttempt;
721             U32 const chainLimit = chainAttempts > chainLength ? chainLength : chainAttempts;
722             U32 chainAttempt;
723 
724             for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++) {
725                 PREFETCH_L1(ddsBase + dms->chainTable[chainIndex + chainAttempt]);
726             }
727 
728             for (chainAttempt = 0 ; chainAttempt < chainLimit; chainAttempt++, chainIndex++) {
729                 size_t currentMl=0;
730                 const BYTE* match;
731                 matchIndex = dms->chainTable[chainIndex];
732                 match = ddsBase + matchIndex;
733 
734                 /* guaranteed by table construction */
735                 assert(matchIndex >= ddsLowestIndex);
736                 assert(match+4 <= ddsEnd);
737                 if (MEM_read32(match) == MEM_read32(ip)) {
738                     /* assumption : matchIndex <= dictLimit-4 (by table construction) */
739                     currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, ddsEnd, prefixStart) + 4;
740                 }
741 
742                 /* save best solution */
743                 if (currentMl > ml) {
744                     ml = currentMl;
745                     *offsetPtr = curr - (matchIndex + ddsIndexDelta) + ZSTD_REP_MOVE;
746                     if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
747                 }
748             }
749         }
750     } else if (dictMode == ZSTD_dictMatchState) {
751         const U32* const dmsChainTable = dms->chainTable;
752         const U32 dmsChainSize         = (1 << dms->cParams.chainLog);
753         const U32 dmsChainMask         = dmsChainSize - 1;
754         const U32 dmsLowestIndex       = dms->window.dictLimit;
755         const BYTE* const dmsBase      = dms->window.base;
756         const BYTE* const dmsEnd       = dms->window.nextSrc;
757         const U32 dmsSize              = (U32)(dmsEnd - dmsBase);
758         const U32 dmsIndexDelta        = dictLimit - dmsSize;
759         const U32 dmsMinChain = dmsSize > dmsChainSize ? dmsSize - dmsChainSize : 0;
760 
761         matchIndex = dms->hashTable[ZSTD_hashPtr(ip, dms->cParams.hashLog, mls)];
762 
763         for ( ; (matchIndex>=dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
764             size_t currentMl=0;
765             const BYTE* const match = dmsBase + matchIndex;
766             assert(match+4 <= dmsEnd);
767             if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
768                 currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
769 
770             /* save best solution */
771             if (currentMl > ml) {
772                 ml = currentMl;
773                 *offsetPtr = curr - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
774                 if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
775             }
776 
777             if (matchIndex <= dmsMinChain) break;
778 
779             matchIndex = dmsChainTable[matchIndex & dmsChainMask];
780         }
781     }
782 
783     return ml;
784 }
785 
786 
787 FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
788                         ZSTD_matchState_t* ms,
789                         const BYTE* ip, const BYTE* const iLimit,
790                         size_t* offsetPtr)
791 {
792     switch(ms->cParams.minMatch)
793     {
794     default : /* includes case 3 */
795     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
796     case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
797     case 7 :
798     case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
799     }
800 }
801 
802 
803 static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
804                         ZSTD_matchState_t* ms,
805                         const BYTE* ip, const BYTE* const iLimit,
806                         size_t* offsetPtr)
807 {
808     switch(ms->cParams.minMatch)
809     {
810     default : /* includes case 3 */
811     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
812     case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
813     case 7 :
814     case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
815     }
816 }
817 
818 
819 static size_t ZSTD_HcFindBestMatch_dedicatedDictSearch_selectMLS (
820                         ZSTD_matchState_t* ms,
821                         const BYTE* ip, const BYTE* const iLimit,
822                         size_t* offsetPtr)
823 {
824     switch(ms->cParams.minMatch)
825     {
826     default : /* includes case 3 */
827     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_dedicatedDictSearch);
828     case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_dedicatedDictSearch);
829     case 7 :
830     case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_dedicatedDictSearch);
831     }
832 }
833 
834 
835 FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
836                         ZSTD_matchState_t* ms,
837                         const BYTE* ip, const BYTE* const iLimit,
838                         size_t* offsetPtr)
839 {
840     switch(ms->cParams.minMatch)
841     {
842     default : /* includes case 3 */
843     case 4 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
844     case 5 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
845     case 7 :
846     case 6 : return ZSTD_HcFindBestMatch_generic(ms, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
847     }
848 }
849 
850 
851 /* *******************************
852 *  Common parser - lazy strategy
853 *********************************/
854 typedef enum { search_hashChain, search_binaryTree } searchMethod_e;
855 
856 FORCE_INLINE_TEMPLATE size_t
857 ZSTD_compressBlock_lazy_generic(
858                         ZSTD_matchState_t* ms, seqStore_t* seqStore,
859                         U32 rep[ZSTD_REP_NUM],
860                         const void* src, size_t srcSize,
861                         const searchMethod_e searchMethod, const U32 depth,
862                         ZSTD_dictMode_e const dictMode)
863 {
864     const BYTE* const istart = (const BYTE*)src;
865     const BYTE* ip = istart;
866     const BYTE* anchor = istart;
867     const BYTE* const iend = istart + srcSize;
868     const BYTE* const ilimit = iend - 8;
869     const BYTE* const base = ms->window.base;
870     const U32 prefixLowestIndex = ms->window.dictLimit;
871     const BYTE* const prefixLowest = base + prefixLowestIndex;
872 
873     typedef size_t (*searchMax_f)(
874                         ZSTD_matchState_t* ms,
875                         const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
876 
877     /*
878      * This table is indexed first by the four ZSTD_dictMode_e values, and then
879      * by the two searchMethod_e values. NULLs are placed for configurations
880      * that should never occur (extDict modes go to the other implementation
881      * below and there is no DDSS for binary tree search yet).
882      */
883     const searchMax_f searchFuncs[4][2] = {
884         {
885             ZSTD_HcFindBestMatch_selectMLS,
886             ZSTD_BtFindBestMatch_selectMLS
887         },
888         {
889             NULL,
890             NULL
891         },
892         {
893             ZSTD_HcFindBestMatch_dictMatchState_selectMLS,
894             ZSTD_BtFindBestMatch_dictMatchState_selectMLS
895         },
896         {
897             ZSTD_HcFindBestMatch_dedicatedDictSearch_selectMLS,
898             NULL
899         }
900     };
901 
902     searchMax_f const searchMax = searchFuncs[dictMode][searchMethod == search_binaryTree];
903     U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
904 
905     const int isDMS = dictMode == ZSTD_dictMatchState;
906     const int isDDS = dictMode == ZSTD_dedicatedDictSearch;
907     const int isDxS = isDMS || isDDS;
908     const ZSTD_matchState_t* const dms = ms->dictMatchState;
909     const U32 dictLowestIndex      = isDxS ? dms->window.dictLimit : 0;
910     const BYTE* const dictBase     = isDxS ? dms->window.base : NULL;
911     const BYTE* const dictLowest   = isDxS ? dictBase + dictLowestIndex : NULL;
912     const BYTE* const dictEnd      = isDxS ? dms->window.nextSrc : NULL;
913     const U32 dictIndexDelta       = isDxS ?
914                                      prefixLowestIndex - (U32)(dictEnd - dictBase) :
915                                      0;
916     const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictLowest));
917 
918     assert(searchMax != NULL);
919 
920     DEBUGLOG(5, "ZSTD_compressBlock_lazy_generic (dictMode=%u)", (U32)dictMode);
921 
922     /* init */
923     ip += (dictAndPrefixLength == 0);
924     if (dictMode == ZSTD_noDict) {
925         U32 const curr = (U32)(ip - base);
926         U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, curr, ms->cParams.windowLog);
927         U32 const maxRep = curr - windowLow;
928         if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
929         if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
930     }
931     if (isDxS) {
932         /* dictMatchState repCode checks don't currently handle repCode == 0
933          * disabling. */
934         assert(offset_1 <= dictAndPrefixLength);
935         assert(offset_2 <= dictAndPrefixLength);
936     }
937 
938     /* Match Loop */
939 #if defined(__x86_64__)
940     /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
941      * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
942      */
943     __asm__(".p2align 5");
944 #endif
945     while (ip < ilimit) {
946         size_t matchLength=0;
947         size_t offset=0;
948         const BYTE* start=ip+1;
949 
950         /* check repCode */
951         if (isDxS) {
952             const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
953             const BYTE* repMatch = ((dictMode == ZSTD_dictMatchState || dictMode == ZSTD_dedicatedDictSearch)
954                                 && repIndex < prefixLowestIndex) ?
955                                    dictBase + (repIndex - dictIndexDelta) :
956                                    base + repIndex;
957             if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
958                 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
959                 const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
960                 matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
961                 if (depth==0) goto _storeSequence;
962             }
963         }
964         if ( dictMode == ZSTD_noDict
965           && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
966             matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
967             if (depth==0) goto _storeSequence;
968         }
969 
970         /* first search (depth 0) */
971         {   size_t offsetFound = 999999999;
972             size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
973             if (ml2 > matchLength)
974                 matchLength = ml2, start = ip, offset=offsetFound;
975         }
976 
977         if (matchLength < 4) {
978             ip += ((ip-anchor) >> kSearchStrength) + 1;   /* jump faster over incompressible sections */
979             continue;
980         }
981 
982         /* let's try to find a better solution */
983         if (depth>=1)
984         while (ip<ilimit) {
985             ip ++;
986             if ( (dictMode == ZSTD_noDict)
987               && (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
988                 size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
989                 int const gain2 = (int)(mlRep * 3);
990                 int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
991                 if ((mlRep >= 4) && (gain2 > gain1))
992                     matchLength = mlRep, offset = 0, start = ip;
993             }
994             if (isDxS) {
995                 const U32 repIndex = (U32)(ip - base) - offset_1;
996                 const BYTE* repMatch = repIndex < prefixLowestIndex ?
997                                dictBase + (repIndex - dictIndexDelta) :
998                                base + repIndex;
999                 if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
1000                     && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
1001                     const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
1002                     size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
1003                     int const gain2 = (int)(mlRep * 3);
1004                     int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
1005                     if ((mlRep >= 4) && (gain2 > gain1))
1006                         matchLength = mlRep, offset = 0, start = ip;
1007                 }
1008             }
1009             {   size_t offset2=999999999;
1010                 size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1011                 int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
1012                 int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
1013                 if ((ml2 >= 4) && (gain2 > gain1)) {
1014                     matchLength = ml2, offset = offset2, start = ip;
1015                     continue;   /* search a better one */
1016             }   }
1017 
1018             /* let's find an even better one */
1019             if ((depth==2) && (ip<ilimit)) {
1020                 ip ++;
1021                 if ( (dictMode == ZSTD_noDict)
1022                   && (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
1023                     size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
1024                     int const gain2 = (int)(mlRep * 4);
1025                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
1026                     if ((mlRep >= 4) && (gain2 > gain1))
1027                         matchLength = mlRep, offset = 0, start = ip;
1028                 }
1029                 if (isDxS) {
1030                     const U32 repIndex = (U32)(ip - base) - offset_1;
1031                     const BYTE* repMatch = repIndex < prefixLowestIndex ?
1032                                    dictBase + (repIndex - dictIndexDelta) :
1033                                    base + repIndex;
1034                     if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
1035                         && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
1036                         const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
1037                         size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
1038                         int const gain2 = (int)(mlRep * 4);
1039                         int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
1040                         if ((mlRep >= 4) && (gain2 > gain1))
1041                             matchLength = mlRep, offset = 0, start = ip;
1042                     }
1043                 }
1044                 {   size_t offset2=999999999;
1045                     size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1046                     int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
1047                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
1048                     if ((ml2 >= 4) && (gain2 > gain1)) {
1049                         matchLength = ml2, offset = offset2, start = ip;
1050                         continue;
1051             }   }   }
1052             break;  /* nothing found : store previous solution */
1053         }
1054 
1055         /* NOTE:
1056          * start[-offset+ZSTD_REP_MOVE-1] is undefined behavior.
1057          * (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which
1058          * overflows the pointer, which is undefined behavior.
1059          */
1060         /* catch up */
1061         if (offset) {
1062             if (dictMode == ZSTD_noDict) {
1063                 while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
1064                      && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) )  /* only search for offset within prefix */
1065                     { start--; matchLength++; }
1066             }
1067             if (isDxS) {
1068                 U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
1069                 const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
1070                 const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
1071                 while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
1072             }
1073             offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
1074         }
1075         /* store sequence */
1076 _storeSequence:
1077         {   size_t const litLength = start - anchor;
1078             ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
1079             anchor = ip = start + matchLength;
1080         }
1081 
1082         /* check immediate repcode */
1083         if (isDxS) {
1084             while (ip <= ilimit) {
1085                 U32 const current2 = (U32)(ip-base);
1086                 U32 const repIndex = current2 - offset_2;
1087                 const BYTE* repMatch = repIndex < prefixLowestIndex ?
1088                         dictBase - dictIndexDelta + repIndex :
1089                         base + repIndex;
1090                 if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
1091                    && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
1092                     const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
1093                     matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
1094                     offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset_2 <=> offset_1 */
1095                     ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
1096                     ip += matchLength;
1097                     anchor = ip;
1098                     continue;
1099                 }
1100                 break;
1101             }
1102         }
1103 
1104         if (dictMode == ZSTD_noDict) {
1105             while ( ((ip <= ilimit) & (offset_2>0))
1106                  && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
1107                 /* store sequence */
1108                 matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
1109                 offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
1110                 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
1111                 ip += matchLength;
1112                 anchor = ip;
1113                 continue;   /* faster when present ... (?) */
1114     }   }   }
1115 
1116     /* Save reps for next block */
1117     rep[0] = offset_1 ? offset_1 : savedOffset;
1118     rep[1] = offset_2 ? offset_2 : savedOffset;
1119 
1120     /* Return the last literals size */
1121     return (size_t)(iend - anchor);
1122 }
1123 
1124 
1125 size_t ZSTD_compressBlock_btlazy2(
1126         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1127         void const* src, size_t srcSize)
1128 {
1129     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_noDict);
1130 }
1131 
1132 size_t ZSTD_compressBlock_lazy2(
1133         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1134         void const* src, size_t srcSize)
1135 {
1136     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_noDict);
1137 }
1138 
1139 size_t ZSTD_compressBlock_lazy(
1140         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1141         void const* src, size_t srcSize)
1142 {
1143     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_noDict);
1144 }
1145 
1146 size_t ZSTD_compressBlock_greedy(
1147         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1148         void const* src, size_t srcSize)
1149 {
1150     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_noDict);
1151 }
1152 
1153 size_t ZSTD_compressBlock_btlazy2_dictMatchState(
1154         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1155         void const* src, size_t srcSize)
1156 {
1157     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2, ZSTD_dictMatchState);
1158 }
1159 
1160 size_t ZSTD_compressBlock_lazy2_dictMatchState(
1161         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1162         void const* src, size_t srcSize)
1163 {
1164     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dictMatchState);
1165 }
1166 
1167 size_t ZSTD_compressBlock_lazy_dictMatchState(
1168         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1169         void const* src, size_t srcSize)
1170 {
1171     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dictMatchState);
1172 }
1173 
1174 size_t ZSTD_compressBlock_greedy_dictMatchState(
1175         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1176         void const* src, size_t srcSize)
1177 {
1178     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dictMatchState);
1179 }
1180 
1181 
1182 size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch(
1183         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1184         void const* src, size_t srcSize)
1185 {
1186     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2, ZSTD_dedicatedDictSearch);
1187 }
1188 
1189 size_t ZSTD_compressBlock_lazy_dedicatedDictSearch(
1190         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1191         void const* src, size_t srcSize)
1192 {
1193     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1, ZSTD_dedicatedDictSearch);
1194 }
1195 
1196 size_t ZSTD_compressBlock_greedy_dedicatedDictSearch(
1197         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1198         void const* src, size_t srcSize)
1199 {
1200     return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0, ZSTD_dedicatedDictSearch);
1201 }
1202 
1203 
1204 FORCE_INLINE_TEMPLATE
1205 size_t ZSTD_compressBlock_lazy_extDict_generic(
1206                         ZSTD_matchState_t* ms, seqStore_t* seqStore,
1207                         U32 rep[ZSTD_REP_NUM],
1208                         const void* src, size_t srcSize,
1209                         const searchMethod_e searchMethod, const U32 depth)
1210 {
1211     const BYTE* const istart = (const BYTE*)src;
1212     const BYTE* ip = istart;
1213     const BYTE* anchor = istart;
1214     const BYTE* const iend = istart + srcSize;
1215     const BYTE* const ilimit = iend - 8;
1216     const BYTE* const base = ms->window.base;
1217     const U32 dictLimit = ms->window.dictLimit;
1218     const BYTE* const prefixStart = base + dictLimit;
1219     const BYTE* const dictBase = ms->window.dictBase;
1220     const BYTE* const dictEnd  = dictBase + dictLimit;
1221     const BYTE* const dictStart  = dictBase + ms->window.lowLimit;
1222     const U32 windowLog = ms->cParams.windowLog;
1223 
1224     typedef size_t (*searchMax_f)(
1225                         ZSTD_matchState_t* ms,
1226                         const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
1227     searchMax_f searchMax = searchMethod==search_binaryTree ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
1228 
1229     U32 offset_1 = rep[0], offset_2 = rep[1];
1230 
1231     DEBUGLOG(5, "ZSTD_compressBlock_lazy_extDict_generic");
1232 
1233     /* init */
1234     ip += (ip == prefixStart);
1235 
1236     /* Match Loop */
1237 #if defined(__x86_64__)
1238     /* I've measured random a 5% speed loss on levels 5 & 6 (greedy) when the
1239      * code alignment is perturbed. To fix the instability align the loop on 32-bytes.
1240      */
1241     __asm__(".p2align 5");
1242 #endif
1243     while (ip < ilimit) {
1244         size_t matchLength=0;
1245         size_t offset=0;
1246         const BYTE* start=ip+1;
1247         U32 curr = (U32)(ip-base);
1248 
1249         /* check repCode */
1250         {   const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr+1, windowLog);
1251             const U32 repIndex = (U32)(curr+1 - offset_1);
1252             const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1253             const BYTE* const repMatch = repBase + repIndex;
1254             if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))   /* intentional overflow */
1255             if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
1256                 /* repcode detected we should take it */
1257                 const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1258                 matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1259                 if (depth==0) goto _storeSequence;
1260         }   }
1261 
1262         /* first search (depth 0) */
1263         {   size_t offsetFound = 999999999;
1264             size_t const ml2 = searchMax(ms, ip, iend, &offsetFound);
1265             if (ml2 > matchLength)
1266                 matchLength = ml2, start = ip, offset=offsetFound;
1267         }
1268 
1269         if (matchLength < 4) {
1270             ip += ((ip-anchor) >> kSearchStrength) + 1;   /* jump faster over incompressible sections */
1271             continue;
1272         }
1273 
1274         /* let's try to find a better solution */
1275         if (depth>=1)
1276         while (ip<ilimit) {
1277             ip ++;
1278             curr++;
1279             /* check repCode */
1280             if (offset) {
1281                 const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
1282                 const U32 repIndex = (U32)(curr - offset_1);
1283                 const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1284                 const BYTE* const repMatch = repBase + repIndex;
1285                 if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
1286                 if (MEM_read32(ip) == MEM_read32(repMatch)) {
1287                     /* repcode detected */
1288                     const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1289                     size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1290                     int const gain2 = (int)(repLength * 3);
1291                     int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
1292                     if ((repLength >= 4) && (gain2 > gain1))
1293                         matchLength = repLength, offset = 0, start = ip;
1294             }   }
1295 
1296             /* search match, depth 1 */
1297             {   size_t offset2=999999999;
1298                 size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1299                 int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
1300                 int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
1301                 if ((ml2 >= 4) && (gain2 > gain1)) {
1302                     matchLength = ml2, offset = offset2, start = ip;
1303                     continue;   /* search a better one */
1304             }   }
1305 
1306             /* let's find an even better one */
1307             if ((depth==2) && (ip<ilimit)) {
1308                 ip ++;
1309                 curr++;
1310                 /* check repCode */
1311                 if (offset) {
1312                     const U32 windowLow = ZSTD_getLowestMatchIndex(ms, curr, windowLog);
1313                     const U32 repIndex = (U32)(curr - offset_1);
1314                     const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1315                     const BYTE* const repMatch = repBase + repIndex;
1316                     if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
1317                     if (MEM_read32(ip) == MEM_read32(repMatch)) {
1318                         /* repcode detected */
1319                         const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1320                         size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1321                         int const gain2 = (int)(repLength * 4);
1322                         int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
1323                         if ((repLength >= 4) && (gain2 > gain1))
1324                             matchLength = repLength, offset = 0, start = ip;
1325                 }   }
1326 
1327                 /* search match, depth 2 */
1328                 {   size_t offset2=999999999;
1329                     size_t const ml2 = searchMax(ms, ip, iend, &offset2);
1330                     int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
1331                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
1332                     if ((ml2 >= 4) && (gain2 > gain1)) {
1333                         matchLength = ml2, offset = offset2, start = ip;
1334                         continue;
1335             }   }   }
1336             break;  /* nothing found : store previous solution */
1337         }
1338 
1339         /* catch up */
1340         if (offset) {
1341             U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
1342             const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
1343             const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
1344             while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
1345             offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
1346         }
1347 
1348         /* store sequence */
1349 _storeSequence:
1350         {   size_t const litLength = start - anchor;
1351             ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
1352             anchor = ip = start + matchLength;
1353         }
1354 
1355         /* check immediate repcode */
1356         while (ip <= ilimit) {
1357             const U32 repCurrent = (U32)(ip-base);
1358             const U32 windowLow = ZSTD_getLowestMatchIndex(ms, repCurrent, windowLog);
1359             const U32 repIndex = repCurrent - offset_2;
1360             const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
1361             const BYTE* const repMatch = repBase + repIndex;
1362             if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > windowLow))  /* intentional overflow */
1363             if (MEM_read32(ip) == MEM_read32(repMatch)) {
1364                 /* repcode detected we should take it */
1365                 const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
1366                 matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
1367                 offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset history */
1368                 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
1369                 ip += matchLength;
1370                 anchor = ip;
1371                 continue;   /* faster when present ... (?) */
1372             }
1373             break;
1374     }   }
1375 
1376     /* Save reps for next block */
1377     rep[0] = offset_1;
1378     rep[1] = offset_2;
1379 
1380     /* Return the last literals size */
1381     return (size_t)(iend - anchor);
1382 }
1383 
1384 
1385 size_t ZSTD_compressBlock_greedy_extDict(
1386         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1387         void const* src, size_t srcSize)
1388 {
1389     return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 0);
1390 }
1391 
1392 size_t ZSTD_compressBlock_lazy_extDict(
1393         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1394         void const* src, size_t srcSize)
1395 
1396 {
1397     return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 1);
1398 }
1399 
1400 size_t ZSTD_compressBlock_lazy2_extDict(
1401         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1402         void const* src, size_t srcSize)
1403 
1404 {
1405     return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_hashChain, 2);
1406 }
1407 
1408 size_t ZSTD_compressBlock_btlazy2_extDict(
1409         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
1410         void const* src, size_t srcSize)
1411 
1412 {
1413     return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, src, srcSize, search_binaryTree, 2);
1414 }
1415