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_double_fast.h"
13
14
15 void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
16 void const* end, ZSTD_dictTableLoadMethod_e dtlm)
17 {
18 const ZSTD_compressionParameters* const cParams = &ms->cParams;
19 U32* const hashLarge = ms->hashTable;
20 U32 const hBitsL = cParams->hashLog;
21 U32 const mls = cParams->minMatch;
22 U32* const hashSmall = ms->chainTable;
23 U32 const hBitsS = cParams->chainLog;
24 const BYTE* const base = ms->window.base;
25 const BYTE* ip = base + ms->nextToUpdate;
26 const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
27 const U32 fastHashFillStep = 3;
28
29 /* Always insert every fastHashFillStep position into the hash tables.
30 * Insert the other positions into the large hash table if their entry
31 * is empty.
32 */
33 for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
34 U32 const curr = (U32)(ip - base);
35 U32 i;
36 for (i = 0; i < fastHashFillStep; ++i) {
37 size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
38 size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
39 if (i == 0)
40 hashSmall[smHash] = curr + i;
41 if (i == 0 || hashLarge[lgHash] == 0)
42 hashLarge[lgHash] = curr + i;
43 /* Only load extra positions for ZSTD_dtlm_full */
44 if (dtlm == ZSTD_dtlm_fast)
45 break;
46 } }
47 }
48
49
50 FORCE_INLINE_TEMPLATE
51 size_t ZSTD_compressBlock_doubleFast_noDict_generic(
52 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
53 void const* src, size_t srcSize, U32 const mls /* template */)
54 {
55 ZSTD_compressionParameters const* cParams = &ms->cParams;
56 U32* const hashLong = ms->hashTable;
57 const U32 hBitsL = cParams->hashLog;
58 U32* const hashSmall = ms->chainTable;
59 const U32 hBitsS = cParams->chainLog;
60 const BYTE* const base = ms->window.base;
61 const BYTE* const istart = (const BYTE*)src;
62 const BYTE* anchor = istart;
63 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
64 /* presumes that, if there is a dictionary, it must be using Attach mode */
65 const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
66 const BYTE* const prefixLowest = base + prefixLowestIndex;
67 const BYTE* const iend = istart + srcSize;
68 const BYTE* const ilimit = iend - HASH_READ_SIZE;
69 U32 offset_1=rep[0], offset_2=rep[1];
70 U32 offsetSaved = 0;
71
72 size_t mLength;
73 U32 offset;
74 U32 curr;
75
76 /* how many positions to search before increasing step size */
77 const size_t kStepIncr = 1 << kSearchStrength;
78 /* the position at which to increment the step size if no match is found */
79 const BYTE* nextStep;
80 size_t step; /* the current step size */
81
82 size_t hl0; /* the long hash at ip */
83 size_t hl1; /* the long hash at ip1 */
84
85 U32 idxl0; /* the long match index for ip */
86 U32 idxl1; /* the long match index for ip1 */
87
88 const BYTE* matchl0; /* the long match for ip */
89 const BYTE* matchs0; /* the short match for ip */
90 const BYTE* matchl1; /* the long match for ip1 */
91
92 const BYTE* ip = istart; /* the current position */
93 const BYTE* ip1; /* the next position */
94
95 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_noDict_generic");
96
97 /* init */
98 ip += ((ip - prefixLowest) == 0);
99 {
100 U32 const current = (U32)(ip - base);
101 U32 const windowLow = ZSTD_getLowestPrefixIndex(ms, current, cParams->windowLog);
102 U32 const maxRep = current - windowLow;
103 if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
104 if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
105 }
106
107 /* Outer Loop: one iteration per match found and stored */
108 while (1) {
109 step = 1;
110 nextStep = ip + kStepIncr;
111 ip1 = ip + step;
112
113 if (ip1 > ilimit) {
114 goto _cleanup;
115 }
116
117 hl0 = ZSTD_hashPtr(ip, hBitsL, 8);
118 idxl0 = hashLong[hl0];
119 matchl0 = base + idxl0;
120
121 /* Inner Loop: one iteration per search / position */
122 do {
123 const size_t hs0 = ZSTD_hashPtr(ip, hBitsS, mls);
124 const U32 idxs0 = hashSmall[hs0];
125 curr = (U32)(ip-base);
126 matchs0 = base + idxs0;
127
128 hashLong[hl0] = hashSmall[hs0] = curr; /* update hash tables */
129
130 /* check noDict repcode */
131 if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
132 mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
133 ip++;
134 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
135 goto _match_stored;
136 }
137
138 hl1 = ZSTD_hashPtr(ip1, hBitsL, 8);
139
140 if (idxl0 > prefixLowestIndex) {
141 /* check prefix long match */
142 if (MEM_read64(matchl0) == MEM_read64(ip)) {
143 mLength = ZSTD_count(ip+8, matchl0+8, iend) + 8;
144 offset = (U32)(ip-matchl0);
145 while (((ip>anchor) & (matchl0>prefixLowest)) && (ip[-1] == matchl0[-1])) { ip--; matchl0--; mLength++; } /* catch up */
146 goto _match_found;
147 }
148 }
149
150 idxl1 = hashLong[hl1];
151 matchl1 = base + idxl1;
152
153 if (idxs0 > prefixLowestIndex) {
154 /* check prefix short match */
155 if (MEM_read32(matchs0) == MEM_read32(ip)) {
156 goto _search_next_long;
157 }
158 }
159
160 if (ip1 >= nextStep) {
161 PREFETCH_L1(ip1 + 64);
162 PREFETCH_L1(ip1 + 128);
163 step++;
164 nextStep += kStepIncr;
165 }
166 ip = ip1;
167 ip1 += step;
168
169 hl0 = hl1;
170 idxl0 = idxl1;
171 matchl0 = matchl1;
172 #if defined(__aarch64__)
173 PREFETCH_L1(ip+256);
174 #endif
175 } while (ip1 <= ilimit);
176
177 _cleanup:
178 /* save reps for next block */
179 rep[0] = offset_1 ? offset_1 : offsetSaved;
180 rep[1] = offset_2 ? offset_2 : offsetSaved;
181
182 /* Return the last literals size */
183 return (size_t)(iend - anchor);
184
185 _search_next_long:
186
187 /* check prefix long +1 match */
188 if (idxl1 > prefixLowestIndex) {
189 if (MEM_read64(matchl1) == MEM_read64(ip1)) {
190 ip = ip1;
191 mLength = ZSTD_count(ip+8, matchl1+8, iend) + 8;
192 offset = (U32)(ip-matchl1);
193 while (((ip>anchor) & (matchl1>prefixLowest)) && (ip[-1] == matchl1[-1])) { ip--; matchl1--; mLength++; } /* catch up */
194 goto _match_found;
195 }
196 }
197
198 /* if no long +1 match, explore the short match we found */
199 mLength = ZSTD_count(ip+4, matchs0+4, iend) + 4;
200 offset = (U32)(ip - matchs0);
201 while (((ip>anchor) & (matchs0>prefixLowest)) && (ip[-1] == matchs0[-1])) { ip--; matchs0--; mLength++; } /* catch up */
202
203 /* fall-through */
204
205 _match_found: /* requires ip, offset, mLength */
206 offset_2 = offset_1;
207 offset_1 = offset;
208
209 if (step < 4) {
210 /* It is unsafe to write this value back to the hashtable when ip1 is
211 * greater than or equal to the new ip we will have after we're done
212 * processing this match. Rather than perform that test directly
213 * (ip1 >= ip + mLength), which costs speed in practice, we do a simpler
214 * more predictable test. The minmatch even if we take a short match is
215 * 4 bytes, so as long as step, the distance between ip and ip1
216 * (initially) is less than 4, we know ip1 < new ip. */
217 hashLong[hl1] = (U32)(ip1 - base);
218 }
219
220 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
221
222 _match_stored:
223 /* match found */
224 ip += mLength;
225 anchor = ip;
226
227 if (ip <= ilimit) {
228 /* Complementary insertion */
229 /* done after iLimit test, as candidates could be > iend-8 */
230 { U32 const indexToInsert = curr+2;
231 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
232 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
233 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
234 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
235 }
236
237 /* check immediate repcode */
238 while ( (ip <= ilimit)
239 && ( (offset_2>0)
240 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
241 /* store sequence */
242 size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
243 U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; /* swap offset_2 <=> offset_1 */
244 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
245 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
246 ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, rLength);
247 ip += rLength;
248 anchor = ip;
249 continue; /* faster when present ... (?) */
250 }
251 }
252 }
253 }
254
255
256 FORCE_INLINE_TEMPLATE
257 size_t ZSTD_compressBlock_doubleFast_dictMatchState_generic(
258 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
259 void const* src, size_t srcSize,
260 U32 const mls /* template */)
261 {
262 ZSTD_compressionParameters const* cParams = &ms->cParams;
263 U32* const hashLong = ms->hashTable;
264 const U32 hBitsL = cParams->hashLog;
265 U32* const hashSmall = ms->chainTable;
266 const U32 hBitsS = cParams->chainLog;
267 const BYTE* const base = ms->window.base;
268 const BYTE* const istart = (const BYTE*)src;
269 const BYTE* ip = istart;
270 const BYTE* anchor = istart;
271 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
272 /* presumes that, if there is a dictionary, it must be using Attach mode */
273 const U32 prefixLowestIndex = ZSTD_getLowestPrefixIndex(ms, endIndex, cParams->windowLog);
274 const BYTE* const prefixLowest = base + prefixLowestIndex;
275 const BYTE* const iend = istart + srcSize;
276 const BYTE* const ilimit = iend - HASH_READ_SIZE;
277 U32 offset_1=rep[0], offset_2=rep[1];
278 U32 offsetSaved = 0;
279
280 const ZSTD_matchState_t* const dms = ms->dictMatchState;
281 const ZSTD_compressionParameters* const dictCParams = &dms->cParams;
282 const U32* const dictHashLong = dms->hashTable;
283 const U32* const dictHashSmall = dms->chainTable;
284 const U32 dictStartIndex = dms->window.dictLimit;
285 const BYTE* const dictBase = dms->window.base;
286 const BYTE* const dictStart = dictBase + dictStartIndex;
287 const BYTE* const dictEnd = dms->window.nextSrc;
288 const U32 dictIndexDelta = prefixLowestIndex - (U32)(dictEnd - dictBase);
289 const U32 dictHBitsL = dictCParams->hashLog;
290 const U32 dictHBitsS = dictCParams->chainLog;
291 const U32 dictAndPrefixLength = (U32)((ip - prefixLowest) + (dictEnd - dictStart));
292
293 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_dictMatchState_generic");
294
295 /* if a dictionary is attached, it must be within window range */
296 assert(ms->window.dictLimit + (1U << cParams->windowLog) >= endIndex);
297
298 /* init */
299 ip += (dictAndPrefixLength == 0);
300
301 /* dictMatchState repCode checks don't currently handle repCode == 0
302 * disabling. */
303 assert(offset_1 <= dictAndPrefixLength);
304 assert(offset_2 <= dictAndPrefixLength);
305
306 /* Main Search Loop */
307 while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
308 size_t mLength;
309 U32 offset;
310 size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
311 size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
312 size_t const dictHL = ZSTD_hashPtr(ip, dictHBitsL, 8);
313 size_t const dictHS = ZSTD_hashPtr(ip, dictHBitsS, mls);
314 U32 const curr = (U32)(ip-base);
315 U32 const matchIndexL = hashLong[h2];
316 U32 matchIndexS = hashSmall[h];
317 const BYTE* matchLong = base + matchIndexL;
318 const BYTE* match = base + matchIndexS;
319 const U32 repIndex = curr + 1 - offset_1;
320 const BYTE* repMatch = (repIndex < prefixLowestIndex) ?
321 dictBase + (repIndex - dictIndexDelta) :
322 base + repIndex;
323 hashLong[h2] = hashSmall[h] = curr; /* update hash tables */
324
325 /* check repcode */
326 if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
327 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
328 const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
329 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
330 ip++;
331 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
332 goto _match_stored;
333 }
334
335 if (matchIndexL > prefixLowestIndex) {
336 /* check prefix long match */
337 if (MEM_read64(matchLong) == MEM_read64(ip)) {
338 mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
339 offset = (U32)(ip-matchLong);
340 while (((ip>anchor) & (matchLong>prefixLowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
341 goto _match_found;
342 }
343 } else {
344 /* check dictMatchState long match */
345 U32 const dictMatchIndexL = dictHashLong[dictHL];
346 const BYTE* dictMatchL = dictBase + dictMatchIndexL;
347 assert(dictMatchL < dictEnd);
348
349 if (dictMatchL > dictStart && MEM_read64(dictMatchL) == MEM_read64(ip)) {
350 mLength = ZSTD_count_2segments(ip+8, dictMatchL+8, iend, dictEnd, prefixLowest) + 8;
351 offset = (U32)(curr - dictMatchIndexL - dictIndexDelta);
352 while (((ip>anchor) & (dictMatchL>dictStart)) && (ip[-1] == dictMatchL[-1])) { ip--; dictMatchL--; mLength++; } /* catch up */
353 goto _match_found;
354 } }
355
356 if (matchIndexS > prefixLowestIndex) {
357 /* check prefix short match */
358 if (MEM_read32(match) == MEM_read32(ip)) {
359 goto _search_next_long;
360 }
361 } else {
362 /* check dictMatchState short match */
363 U32 const dictMatchIndexS = dictHashSmall[dictHS];
364 match = dictBase + dictMatchIndexS;
365 matchIndexS = dictMatchIndexS + dictIndexDelta;
366
367 if (match > dictStart && MEM_read32(match) == MEM_read32(ip)) {
368 goto _search_next_long;
369 } }
370
371 ip += ((ip-anchor) >> kSearchStrength) + 1;
372 #if defined(__aarch64__)
373 PREFETCH_L1(ip+256);
374 #endif
375 continue;
376
377 _search_next_long:
378
379 { size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
380 size_t const dictHLNext = ZSTD_hashPtr(ip+1, dictHBitsL, 8);
381 U32 const matchIndexL3 = hashLong[hl3];
382 const BYTE* matchL3 = base + matchIndexL3;
383 hashLong[hl3] = curr + 1;
384
385 /* check prefix long +1 match */
386 if (matchIndexL3 > prefixLowestIndex) {
387 if (MEM_read64(matchL3) == MEM_read64(ip+1)) {
388 mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
389 ip++;
390 offset = (U32)(ip-matchL3);
391 while (((ip>anchor) & (matchL3>prefixLowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
392 goto _match_found;
393 }
394 } else {
395 /* check dict long +1 match */
396 U32 const dictMatchIndexL3 = dictHashLong[dictHLNext];
397 const BYTE* dictMatchL3 = dictBase + dictMatchIndexL3;
398 assert(dictMatchL3 < dictEnd);
399 if (dictMatchL3 > dictStart && MEM_read64(dictMatchL3) == MEM_read64(ip+1)) {
400 mLength = ZSTD_count_2segments(ip+1+8, dictMatchL3+8, iend, dictEnd, prefixLowest) + 8;
401 ip++;
402 offset = (U32)(curr + 1 - dictMatchIndexL3 - dictIndexDelta);
403 while (((ip>anchor) & (dictMatchL3>dictStart)) && (ip[-1] == dictMatchL3[-1])) { ip--; dictMatchL3--; mLength++; } /* catch up */
404 goto _match_found;
405 } } }
406
407 /* if no long +1 match, explore the short match we found */
408 if (matchIndexS < prefixLowestIndex) {
409 mLength = ZSTD_count_2segments(ip+4, match+4, iend, dictEnd, prefixLowest) + 4;
410 offset = (U32)(curr - matchIndexS);
411 while (((ip>anchor) & (match>dictStart)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
412 } else {
413 mLength = ZSTD_count(ip+4, match+4, iend) + 4;
414 offset = (U32)(ip - match);
415 while (((ip>anchor) & (match>prefixLowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
416 }
417
418 _match_found:
419 offset_2 = offset_1;
420 offset_1 = offset;
421
422 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
423
424 _match_stored:
425 /* match found */
426 ip += mLength;
427 anchor = ip;
428
429 if (ip <= ilimit) {
430 /* Complementary insertion */
431 /* done after iLimit test, as candidates could be > iend-8 */
432 { U32 const indexToInsert = curr+2;
433 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
434 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
435 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
436 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
437 }
438
439 /* check immediate repcode */
440 while (ip <= ilimit) {
441 U32 const current2 = (U32)(ip-base);
442 U32 const repIndex2 = current2 - offset_2;
443 const BYTE* repMatch2 = repIndex2 < prefixLowestIndex ?
444 dictBase + repIndex2 - dictIndexDelta :
445 base + repIndex2;
446 if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex2) >= 3 /* intentional overflow */)
447 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
448 const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend;
449 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4;
450 U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
451 ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, repLength2);
452 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
453 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
454 ip += repLength2;
455 anchor = ip;
456 continue;
457 }
458 break;
459 }
460 }
461 } /* while (ip < ilimit) */
462
463 /* save reps for next block */
464 rep[0] = offset_1 ? offset_1 : offsetSaved;
465 rep[1] = offset_2 ? offset_2 : offsetSaved;
466
467 /* Return the last literals size */
468 return (size_t)(iend - anchor);
469 }
470
471 #define ZSTD_GEN_DFAST_FN(dictMode, mls) \
472 static size_t ZSTD_compressBlock_doubleFast_##dictMode##_##mls( \
473 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM], \
474 void const* src, size_t srcSize) \
475 { \
476 return ZSTD_compressBlock_doubleFast_##dictMode##_generic(ms, seqStore, rep, src, srcSize, mls); \
477 }
478
479 ZSTD_GEN_DFAST_FN(noDict, 4)
480 ZSTD_GEN_DFAST_FN(noDict, 5)
481 ZSTD_GEN_DFAST_FN(noDict, 6)
482 ZSTD_GEN_DFAST_FN(noDict, 7)
483
484 ZSTD_GEN_DFAST_FN(dictMatchState, 4)
485 ZSTD_GEN_DFAST_FN(dictMatchState, 5)
486 ZSTD_GEN_DFAST_FN(dictMatchState, 6)
487 ZSTD_GEN_DFAST_FN(dictMatchState, 7)
488
489
490 size_t ZSTD_compressBlock_doubleFast(
491 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
492 void const* src, size_t srcSize)
493 {
494 const U32 mls = ms->cParams.minMatch;
495 switch(mls)
496 {
497 default: /* includes case 3 */
498 case 4 :
499 return ZSTD_compressBlock_doubleFast_noDict_4(ms, seqStore, rep, src, srcSize);
500 case 5 :
501 return ZSTD_compressBlock_doubleFast_noDict_5(ms, seqStore, rep, src, srcSize);
502 case 6 :
503 return ZSTD_compressBlock_doubleFast_noDict_6(ms, seqStore, rep, src, srcSize);
504 case 7 :
505 return ZSTD_compressBlock_doubleFast_noDict_7(ms, seqStore, rep, src, srcSize);
506 }
507 }
508
509
510 size_t ZSTD_compressBlock_doubleFast_dictMatchState(
511 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
512 void const* src, size_t srcSize)
513 {
514 const U32 mls = ms->cParams.minMatch;
515 switch(mls)
516 {
517 default: /* includes case 3 */
518 case 4 :
519 return ZSTD_compressBlock_doubleFast_dictMatchState_4(ms, seqStore, rep, src, srcSize);
520 case 5 :
521 return ZSTD_compressBlock_doubleFast_dictMatchState_5(ms, seqStore, rep, src, srcSize);
522 case 6 :
523 return ZSTD_compressBlock_doubleFast_dictMatchState_6(ms, seqStore, rep, src, srcSize);
524 case 7 :
525 return ZSTD_compressBlock_doubleFast_dictMatchState_7(ms, seqStore, rep, src, srcSize);
526 }
527 }
528
529
530 static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
531 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
532 void const* src, size_t srcSize,
533 U32 const mls /* template */)
534 {
535 ZSTD_compressionParameters const* cParams = &ms->cParams;
536 U32* const hashLong = ms->hashTable;
537 U32 const hBitsL = cParams->hashLog;
538 U32* const hashSmall = ms->chainTable;
539 U32 const hBitsS = cParams->chainLog;
540 const BYTE* const istart = (const BYTE*)src;
541 const BYTE* ip = istart;
542 const BYTE* anchor = istart;
543 const BYTE* const iend = istart + srcSize;
544 const BYTE* const ilimit = iend - 8;
545 const BYTE* const base = ms->window.base;
546 const U32 endIndex = (U32)((size_t)(istart - base) + srcSize);
547 const U32 lowLimit = ZSTD_getLowestMatchIndex(ms, endIndex, cParams->windowLog);
548 const U32 dictStartIndex = lowLimit;
549 const U32 dictLimit = ms->window.dictLimit;
550 const U32 prefixStartIndex = (dictLimit > lowLimit) ? dictLimit : lowLimit;
551 const BYTE* const prefixStart = base + prefixStartIndex;
552 const BYTE* const dictBase = ms->window.dictBase;
553 const BYTE* const dictStart = dictBase + dictStartIndex;
554 const BYTE* const dictEnd = dictBase + prefixStartIndex;
555 U32 offset_1=rep[0], offset_2=rep[1];
556
557 DEBUGLOG(5, "ZSTD_compressBlock_doubleFast_extDict_generic (srcSize=%zu)", srcSize);
558
559 /* if extDict is invalidated due to maxDistance, switch to "regular" variant */
560 if (prefixStartIndex == dictStartIndex)
561 return ZSTD_compressBlock_doubleFast(ms, seqStore, rep, src, srcSize);
562
563 /* Search Loop */
564 while (ip < ilimit) { /* < instead of <=, because (ip+1) */
565 const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
566 const U32 matchIndex = hashSmall[hSmall];
567 const BYTE* const matchBase = matchIndex < prefixStartIndex ? dictBase : base;
568 const BYTE* match = matchBase + matchIndex;
569
570 const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
571 const U32 matchLongIndex = hashLong[hLong];
572 const BYTE* const matchLongBase = matchLongIndex < prefixStartIndex ? dictBase : base;
573 const BYTE* matchLong = matchLongBase + matchLongIndex;
574
575 const U32 curr = (U32)(ip-base);
576 const U32 repIndex = curr + 1 - offset_1; /* offset_1 expected <= curr +1 */
577 const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
578 const BYTE* const repMatch = repBase + repIndex;
579 size_t mLength;
580 hashSmall[hSmall] = hashLong[hLong] = curr; /* update hash table */
581
582 if ((((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow : ensure repIndex doesn't overlap dict + prefix */
583 & (offset_1 <= curr+1 - dictStartIndex)) /* note: we are searching at curr+1 */
584 && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
585 const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
586 mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
587 ip++;
588 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_REPCODE_1, mLength);
589 } else {
590 if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
591 const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend;
592 const BYTE* const lowMatchPtr = matchLongIndex < prefixStartIndex ? dictStart : prefixStart;
593 U32 offset;
594 mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, prefixStart) + 8;
595 offset = curr - matchLongIndex;
596 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
597 offset_2 = offset_1;
598 offset_1 = offset;
599 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
600
601 } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) {
602 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
603 U32 const matchIndex3 = hashLong[h3];
604 const BYTE* const match3Base = matchIndex3 < prefixStartIndex ? dictBase : base;
605 const BYTE* match3 = match3Base + matchIndex3;
606 U32 offset;
607 hashLong[h3] = curr + 1;
608 if ( (matchIndex3 > dictStartIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
609 const BYTE* const matchEnd = matchIndex3 < prefixStartIndex ? dictEnd : iend;
610 const BYTE* const lowMatchPtr = matchIndex3 < prefixStartIndex ? dictStart : prefixStart;
611 mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, prefixStart) + 8;
612 ip++;
613 offset = curr+1 - matchIndex3;
614 while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
615 } else {
616 const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
617 const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
618 mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
619 offset = curr - matchIndex;
620 while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
621 }
622 offset_2 = offset_1;
623 offset_1 = offset;
624 ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, STORE_OFFSET(offset), mLength);
625
626 } else {
627 ip += ((ip-anchor) >> kSearchStrength) + 1;
628 continue;
629 } }
630
631 /* move to next sequence start */
632 ip += mLength;
633 anchor = ip;
634
635 if (ip <= ilimit) {
636 /* Complementary insertion */
637 /* done after iLimit test, as candidates could be > iend-8 */
638 { U32 const indexToInsert = curr+2;
639 hashLong[ZSTD_hashPtr(base+indexToInsert, hBitsL, 8)] = indexToInsert;
640 hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
641 hashSmall[ZSTD_hashPtr(base+indexToInsert, hBitsS, mls)] = indexToInsert;
642 hashSmall[ZSTD_hashPtr(ip-1, hBitsS, mls)] = (U32)(ip-1-base);
643 }
644
645 /* check immediate repcode */
646 while (ip <= ilimit) {
647 U32 const current2 = (U32)(ip-base);
648 U32 const repIndex2 = current2 - offset_2;
649 const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
650 if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) /* intentional overflow : ensure repIndex2 doesn't overlap dict + prefix */
651 & (offset_2 <= current2 - dictStartIndex))
652 && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
653 const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
654 size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
655 U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
656 ZSTD_storeSeq(seqStore, 0, anchor, iend, STORE_REPCODE_1, repLength2);
657 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
658 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
659 ip += repLength2;
660 anchor = ip;
661 continue;
662 }
663 break;
664 } } }
665
666 /* save reps for next block */
667 rep[0] = offset_1;
668 rep[1] = offset_2;
669
670 /* Return the last literals size */
671 return (size_t)(iend - anchor);
672 }
673
674 ZSTD_GEN_DFAST_FN(extDict, 4)
675 ZSTD_GEN_DFAST_FN(extDict, 5)
676 ZSTD_GEN_DFAST_FN(extDict, 6)
677 ZSTD_GEN_DFAST_FN(extDict, 7)
678
679 size_t ZSTD_compressBlock_doubleFast_extDict(
680 ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
681 void const* src, size_t srcSize)
682 {
683 U32 const mls = ms->cParams.minMatch;
684 switch(mls)
685 {
686 default: /* includes case 3 */
687 case 4 :
688 return ZSTD_compressBlock_doubleFast_extDict_4(ms, seqStore, rep, src, srcSize);
689 case 5 :
690 return ZSTD_compressBlock_doubleFast_extDict_5(ms, seqStore, rep, src, srcSize);
691 case 6 :
692 return ZSTD_compressBlock_doubleFast_extDict_6(ms, seqStore, rep, src, srcSize);
693 case 7 :
694 return ZSTD_compressBlock_doubleFast_extDict_7(ms, seqStore, rep, src, srcSize);
695 }
696 }
697
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.