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 
ZSTD_fillDoubleHashTable(ZSTD_matchState_t * ms,void const * end,ZSTD_dictTableLoadMethod_e dtlm)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
ZSTD_compressBlock_doubleFast_noDict_generic(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize,U32 const mls)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
ZSTD_compressBlock_doubleFast_dictMatchState_generic(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize,U32 const mls)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 
ZSTD_compressBlock_doubleFast(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)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 
ZSTD_compressBlock_doubleFast_dictMatchState(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)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 
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)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 
ZSTD_compressBlock_doubleFast_extDict(ZSTD_matchState_t * ms,seqStore_t * seqStore,U32 rep[ZSTD_REP_NUM],void const * src,size_t srcSize)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