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_double_fast.h"
14
15
ZSTD_fillDoubleHashTable(ZSTD_matchState_t * ms,void const * end,ZSTD_dictTableLoadMethod_e dtlm)16 void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
17 void const* end, ZSTD_dictTableLoadMethod_e dtlm)
18 {
19 const ZSTD_compressionParameters* const cParams = &ms->cParams;
20 U32* const hashLarge = ms->hashTable;
21 U32 const hBitsL = cParams->hashLog;
22 U32 const mls = cParams->minMatch;
23 U32* const hashSmall = ms->chainTable;
24 U32 const hBitsS = cParams->chainLog;
25 const BYTE* const base = ms->window.base;
26 const BYTE* ip = base + ms->nextToUpdate;
27 const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
28 const U32 fastHashFillStep = 3;
29
30 /* Always insert every fastHashFillStep position into the hash tables.
31 * Insert the other positions into the large hash table if their entry
32 * is empty.
33 */
34 for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
35 U32 const current = (U32)(ip - base);
36 U32 i;
37 for (i = 0; i < fastHashFillStep; ++i) {
38 size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
39 size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
40 if (i == 0)
41 hashSmall[smHash] = current + i;
42 if (i == 0 || hashLarge[lgHash] == 0)
43 hashLarge[lgHash] = current + i;
44 /* Only load extra positions for ZSTD_dtlm_full */
45 if (dtlm == ZSTD_dtlm_fast)
46 break;
47 } }
48 }
49
50
51 FORCE_INLINE_TEMPLATE
ZSTD_compressBlock_doubleFast_generic(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize,U32 const mls,ZSTD_dictMode_e const dictMode)52 size_t ZSTD_compressBlock_doubleFast_generic(
53 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
54 void const* src, size_t srcSize,
55 U32 const mls /* template */, ZSTD_dictMode_e const dictMode)
56 {
57 ZSTD_compressionParameters const* cParams = &ms->cParams;
58 U32* const hashLong = ms->hashTable;
59 const U32 hBitsL = cParams->hashLog;
60 U32* const hashSmall = ms->chainTable;
61 const U32 hBitsS = cParams->chainLog;
62 const BYTE* const base = ms->window.base;
63 const BYTE* const istart = (const BYTE*)src;
64 const BYTE* ip = istart;
65 const BYTE* anchor = istart;
66 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
67 /* presumes that, if there is a dictionary, it must be using Attach mode */
68 const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
69 const BYTE* const prefixLowest = base + prefixLowestIndex;
70 const BYTE* const iend = istart + srcSize;
71 const BYTE* const ilimit = iend - HASH_READ_SIZE;
72 U32 offset_1=rep[0], offset_2=rep[1];
73 U32 offsetSaved = 0;
74
75 const ZSTD_matchState_t* const dms = ms->dictMatchState;
76 const ZSTD_compressionParameters* const dictCParams =
77 dictMode == ZSTD_dictMatchState ?
78 &dms->cParams : NULL;
79 const U32* const dictHashLong = dictMode == ZSTD_dictMatchState ?
80 dms->hashTable : NULL;
81 const U32* const dictHashSmall = dictMode == ZSTD_dictMatchState ?
82 dms->chainTable : NULL;
83 const U32 dictStartIndex = dictMode == ZSTD_dictMatchState ?
84 dms->window.dictLimit : 0;
85 const BYTE* const dictBase = dictMode == ZSTD_dictMatchState ?
86 dms->window.base : NULL;
87 const BYTE* const dictStart = dictMode == ZSTD_dictMatchState ?
88 dictBase + dictStartIndex : NULL;
89 const BYTE* const dictEnd = dictMode == ZSTD_dictMatchState ?
90 dms->window.nextSrc : NULL;
91 const U32 dictIndexDelta = dictMode == ZSTD_dictMatchState ?
92 prefixLowestIndex - (U32)(dictEnd - dictBase) :
93 0;
94 const U32 dictHBitsL = dictMode == ZSTD_dictMatchState ?
95 dictCParams->hashLog : hBitsL;
96 const U32 dictHBitsS = dictMode == ZSTD_dictMatchState ?
97 dictCParams->chainLog : hBitsS;
98 const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictStart));
99
100 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_generic");
101
102 assert(dictMode == ZSTD_noDict || dictMode == ZSTD_dictMatchState);
103
104 /* if a dictionary is attached, it must be within window range */
105 if (dictMode == ZSTD_dictMatchState) {
106 assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
107 }
108
109 /* init */
110 ip += (dictAndPrefixLength == 0);
111 if (dictMode == ZSTD_noDict) {
112 U32 const current = (U32)(ip - base);
113 U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
114 U32 const maxRep = current - windowLow;
115 if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
116 if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
117 }
118 if (dictMode == ZSTD_dictMatchState) {
119 /* dictMatchState repCode checks don't currently handle repCode == 0
120 * disabling. */
121 assert(offset_1 <= dictAndPrefixLength);
122 assert(offset_2 <= dictAndPrefixLength);
123 }
124
125 /* Main Search Loop */
126 while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
127 size_t mLength;
128 U32 offset;
129 size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
130 size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
131 size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8);
132 size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls);
133 U32 const current = (U32)(ip-base);
134 U32 const matchIndexL = hashLong[h2];
135 U32 matchIndexS = hashSmall[h];
136 const BYTE* matchLong = base + matchIndexL;
137 const BYTE* match = base + matchIndexS;
138 const U32 repIndex = current + 1 - offset_1;
139 const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
140 && repIndex < prefixLowestIndex) ?
141 dictBase + (repIndex - dictIndexDelta) :
142 base + repIndex;
143 hashLong[h2] = hashSmall[h] = current; /* update hash tables */
144
145 /* check dictMatchState repcode */
146 if (dictMode == ZSTD_dictMatchState
147 && ((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
148 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
149 const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
150 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
151 ip++;
152 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
153 goto _match_stored;
154 }
155
156 /* check noDict repcode */
157 if ( dictMode == ZSTD_noDict
158 && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
159 mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
160 ip++;
161 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
162 goto _match_stored;
163 }
164
165 if (matchIndexL > prefixLowestIndex) {
166 /* check prefix long match */
167 if (MEM_read64(matchLong) == MEM_read64(ip)) {
168 mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
169 offset = (U32)(ip-matchLong);
170 while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
171 goto _match_found;
172 }
173 } else if (dictMode == ZSTD_dictMatchState) {
174 /* check dictMatchState long match */
175 U32 const dictMatchIndexL = dictHashLong[dictHL];
176 const BYTE* dictMatchL = dictBase + dictMatchIndexL;
177 assert(dictMatchL < dictEnd);
178
179 if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) {
180 mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8;
181 offset = (U32)(current - dictMatchIndexL - dictIndexDelta);
182 while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */
183 goto _match_found;
184 } }
185
186 if (matchIndexS > prefixLowestIndex) {
187 /* check prefix short match */
188 if (MEM_read32(match) == MEM_read32(ip)) {
189 goto _search_next_long;
190 }
191 } else if (dictMode == ZSTD_dictMatchState) {
192 /* check dictMatchState short match */
193 U32 const dictMatchIndexS = dictHashSmall[dictHS];
194 match = dictBase + dictMatchIndexS;
195 matchIndexS = dictMatchIndexS + dictIndexDelta;
196
197 if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) {
198 goto _search_next_long;
199 } }
200
201 ip += ((ip-anchor) >> kSearchStrength) + 1;
202 #if defined(__aarch64__)
203 PREFETCH_L1(ip+256);
204 #endif
205 continue;
206
207 _search_next_long:
208
209 { size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
210 size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8);
211 U32 const matchIndexL3 = hashLong[hl3];
212 const BYTE* matchL3 = base + matchIndexL3;
213 hashLong[hl3] = current + 1;
214
215 /* check prefix long +1 match */
216 if (matchIndexL3 > prefixLowestIndex) {
217 if (MEM_read64(matchL3) == MEM_read64(ip+1)) {
218 mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
219 ip++;
220 offset = (U32)(ip-matchL3);
221 while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
222 goto _match_found;
223 }
224 } else if (dictMode == ZSTD_dictMatchState) {
225 /* check dict long +1 match */
226 U32 const dictMatchIndexL3 = dictHashLong[dictHLNext];
227 const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3;
228 assert(dictMatchL3 < dictEnd);
229 if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) {
230 mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8;
231 ip++;
232 offset = (U32)(current + 1 - dictMatchIndexL3 - dictIndexDelta);
233 while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */
234 goto _match_found;
235 } } }
236
237 /* if no long +1 match, explore the short match we found */
238 if (dictMode == ZSTD_dictMatchState && matchIndexS < prefixLowestIndex) {
239 mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4;
240 offset = (U32)(current - matchIndexS);
241 while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
242 } else {
243 mLength = ZSTD_count(ip+4, match+4, iend) + 4;
244 offset = (U32)(ip - match);
245 while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
246 }
247
248 /* fall-through */
249
250 _match_found:
251 offset_2 = offset_1;
252 offset_1 = offset;
253
254 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
255
256 _match_stored:
257 /* match found */
258 ip += mLength;
259 anchor = ip;
260
261 if (ip <= ilimit) {
262 /* Complementary insertion */
263 /* done after iLimit test, as candidates could be > iend-8 */
264 { U32 const indexToInsert = current+2;
265 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
266 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
267 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
268 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
269 }
270
271 /* check immediate repcode */
272 if (dictMode == ZSTD_dictMatchState) {
273 while (ip <= ilimit) {
274 U32 const current2 = (U32)(ip-base);
275 U32 const repIndex2 = current2 - offset_2;
276 const BYTE* repMatch2 = dictMode == ZSTD_dictMatchState
277 && repIndex2 < prefixLowestIndex ?
278 dictBase + repIndex2 - dictIndexDelta :
279 base + repIndex2;
280 if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
281 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
282 const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend;
283 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4;
284 U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
285 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
286 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
287 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
288 ip += repLength2;
289 anchor = ip;
290 continue;
291 }
292 break;
293 } }
294
295 if (dictMode == ZSTD_noDict) {
296 while ( (ip <= ilimit)
297 && ( (offset_2>0)
298 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
299 /* store sequence */
300 size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
301 U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */
302 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
303 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
304 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, rLength-MINMATCH);
305 ip += rLength;
306 anchor = ip;
307 continue; /* faster when present ... (?) */
308 } } }
309 } /* while (ip < ilimit) */
310
311 /* save reps for next block */
312 rep[0] = offset_1 ? offset_1 : offsetSaved;
313 rep[1] = offset_2 ? offset_2 : offsetSaved;
314
315 /* Return the last literals size */
316 return (size_t)(iend - anchor);
317 }
318
319
ZSTD_compressBlock_doubleFast(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)320 size_t ZSTD_compressBlock_doubleFast(
321 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
322 void const* src, size_t srcSize)
323 {
324 const U32 mls = ms->cParams.minMatch;
325 switch(mls)
326 {
327 default: /* includes case 3 */
328 case 4 :
329 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_noDict);
330 case 5 :
331 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_noDict);
332 case 6 :
333 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_noDict);
334 case 7 :
335 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_noDict);
336 }
337 }
338
339
ZSTD_compressBlock_doubleFast_dictMatchState(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)340 size_t ZSTD_compressBlock_doubleFast_dictMatchState(
341 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
342 void const* src, size_t srcSize)
343 {
344 const U32 mls = ms->cParams.minMatch;
345 switch(mls)
346 {
347 default: /* includes case 3 */
348 case 4 :
349 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 4, ZSTD_dictMatchState);
350 case 5 :
351 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 5, ZSTD_dictMatchState);
352 case 6 :
353 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 6, ZSTD_dictMatchState);
354 case 7 :
355 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, 7, ZSTD_dictMatchState);
356 }
357 }
358
359
ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize,U32 const mls)360 static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
361 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
362 void const* src, size_t srcSize,
363 U32 const mls /* template */)
364 {
365 ZSTD_compressionParameters const* cParams = &ms->cParams;
366 U32* const hashLong = ms->hashTable;
367 U32 const hBitsL = cParams->hashLog;
368 U32* const hashSmall = ms->chainTable;
369 U32 const hBitsS = cParams->chainLog;
370 const BYTE* const istart = (const BYTE*)src;
371 const BYTE* ip = istart;
372 const BYTE* anchor = istart;
373 const BYTE* const iend = istart + srcSize;
374 const BYTE* const ilimit = iend - 8;
375 const BYTE* const base = ms->window.base;
376 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
377 const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
378 const U32 dictStartIndex = lowLimit;
379 const U32 dictLimit = ms->window.dictLimit;
380 const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit;
381 const BYTE* const prefixStart = base + prefixStartIndex;
382 const BYTE* const dictBase = ms->window.dictBase;
383 const BYTE* const dictStart = dictBase + dictStartIndex;
384 const BYTE* const dictEnd = dictBase + prefixStartIndex;
385 U32 offset_1=rep[0], offset_2=rep[1];
386
387 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize);
388
389 /* if extDict is invalidated due to maxDistance, switch to "regular" variant */
390 if (prefixStartIndex == dictStartIndex)
391 return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, src, srcSize, mls, ZSTD_noDict);
392
393 /* Search Loop */
394 while (ip < ilimit) { /* < instead of <=, because (ip+1) */
395 const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
396 const U32 matchIndex = hashSmall[hSmall];
397 const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
398 const BYTE* match = matchBase + matchIndex;
399
400 const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
401 const U32 matchLongIndex = hashLong[hLong];
402 const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base;
403 const BYTE* matchLong = matchLongBase + matchLongIndex;
404
405 const U32 current = (U32)(ip-base);
406 const U32 repIndex = current + 1 - offset_1; /* offset_1 expected <= current +1 */
407 const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
408 const BYTE* const repMatch = repBase + repIndex;
409 size_t mLength;
410 hashSmall[hSmall] = hashLong[hLong] = current; /* update hash table */
411
412 if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */
413 & (offset_1 < current+1 - dictStartIndex)) /* note: we are searching at current+1 */
414 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
415 const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
416 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
417 ip++;
418 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
419 } else {
420 if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
421 const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend;
422 const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart;
423 U32 offset;
424 mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8;
425 offset = current - matchLongIndex;
426 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
427 offset_2 = offset_1;
428 offset_1 = offset;
429 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
430
431 } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) {
432 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
433 U32 const matchIndex3 = hashLong[h3];
434 const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base;
435 const BYTE* match3 = match3Base + matchIndex3;
436 U32 offset;
437 hashLong[h3] = current + 1;
438 if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
439 const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend;
440 const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart;
441 mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8;
442 ip++;
443 offset = current+1 - matchIndex3;
444 while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
445 } else {
446 const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
447 const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
448 mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
449 offset = current - matchIndex;
450 while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
451 }
452 offset_2 = offset_1;
453 offset_1 = offset;
454 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
455
456 } else {
457 ip += ((ip-anchor) >> kSearchStrength) + 1;
458 continue;
459 } }
460
461 /* move to next sequence start */
462 ip += mLength;
463 anchor = ip;
464
465 if (ip <= ilimit) {
466 /* Complementary insertion */
467 /* done after iLimit test, as candidates could be > iend-8 */
468 { U32 const indexToInsert = current+2;
469 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
470 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
471 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
472 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
473 }
474
475 /* check immediate repcode */
476 while (ip <= ilimit) {
477 U32 const current2 = (U32)(ip-base);
478 U32 const repIndex2 = current2 - offset_2;
479 const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
480 if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */
481 & (offset_2 < current2 - dictStartIndex))
482 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
483 const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
484 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
485 U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
486 ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
487 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
488 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
489 ip += repLength2;
490 anchor = ip;
491 continue;
492 }
493 break;
494 } } }
495
496 /* save reps for next block */
497 rep[0] = offset_1;
498 rep[1] = offset_2;
499
500 /* Return the last literals size */
501 return (size_t)(iend - anchor);
502 }
503
504
ZSTD_compressBlock_doubleFast_extDict(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)505 size_t ZSTD_compressBlock_doubleFast_extDict(
506 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
507 void const* src, size_t srcSize)
508 {
509 U32 const mls = ms->cParams.minMatch;
510 switch(mls)
511 {
512 default: /* includes case 3 */
513 case 4 :
514 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 4);
515 case 5 :
516 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 5);
517 case 6 :
518 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 6);
519 case 7 :
520 return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, src, srcSize, 7);
521 }
522 }
523