extzstd 0.3 → 0.3.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/HISTORY.ja.md +8 -0
- data/README.md +1 -1
- data/contrib/zstd/CHANGELOG +94 -0
- data/contrib/zstd/CONTRIBUTING.md +351 -1
- data/contrib/zstd/Makefile +32 -10
- data/contrib/zstd/README.md +33 -10
- data/contrib/zstd/TESTING.md +2 -2
- data/contrib/zstd/appveyor.yml +42 -4
- data/contrib/zstd/lib/Makefile +128 -60
- data/contrib/zstd/lib/README.md +47 -16
- data/contrib/zstd/lib/common/bitstream.h +38 -39
- data/contrib/zstd/lib/common/compiler.h +40 -5
- data/contrib/zstd/lib/common/cpu.h +1 -1
- data/contrib/zstd/lib/common/debug.c +11 -31
- data/contrib/zstd/lib/common/debug.h +11 -31
- data/contrib/zstd/lib/common/entropy_common.c +13 -33
- data/contrib/zstd/lib/common/error_private.c +2 -1
- data/contrib/zstd/lib/common/error_private.h +6 -2
- data/contrib/zstd/lib/common/fse.h +12 -32
- data/contrib/zstd/lib/common/fse_decompress.c +12 -35
- data/contrib/zstd/lib/common/huf.h +15 -33
- data/contrib/zstd/lib/common/mem.h +75 -2
- data/contrib/zstd/lib/common/pool.c +8 -4
- data/contrib/zstd/lib/common/pool.h +2 -2
- data/contrib/zstd/lib/common/threading.c +50 -4
- data/contrib/zstd/lib/common/threading.h +36 -4
- data/contrib/zstd/lib/common/xxhash.c +23 -35
- data/contrib/zstd/lib/common/xxhash.h +11 -31
- data/contrib/zstd/lib/common/zstd_common.c +1 -1
- data/contrib/zstd/lib/common/zstd_errors.h +2 -1
- data/contrib/zstd/lib/common/zstd_internal.h +154 -26
- data/contrib/zstd/lib/compress/fse_compress.c +17 -40
- data/contrib/zstd/lib/compress/hist.c +15 -35
- data/contrib/zstd/lib/compress/hist.h +12 -32
- data/contrib/zstd/lib/compress/huf_compress.c +92 -92
- data/contrib/zstd/lib/compress/zstd_compress.c +1191 -1330
- data/contrib/zstd/lib/compress/zstd_compress_internal.h +317 -55
- data/contrib/zstd/lib/compress/zstd_compress_literals.c +158 -0
- data/contrib/zstd/lib/compress/zstd_compress_literals.h +29 -0
- data/contrib/zstd/lib/compress/zstd_compress_sequences.c +419 -0
- data/contrib/zstd/lib/compress/zstd_compress_sequences.h +54 -0
- data/contrib/zstd/lib/compress/zstd_compress_superblock.c +845 -0
- data/contrib/zstd/lib/compress/zstd_compress_superblock.h +32 -0
- data/contrib/zstd/lib/compress/zstd_cwksp.h +525 -0
- data/contrib/zstd/lib/compress/zstd_double_fast.c +65 -43
- data/contrib/zstd/lib/compress/zstd_double_fast.h +2 -2
- data/contrib/zstd/lib/compress/zstd_fast.c +92 -66
- data/contrib/zstd/lib/compress/zstd_fast.h +2 -2
- data/contrib/zstd/lib/compress/zstd_lazy.c +74 -42
- data/contrib/zstd/lib/compress/zstd_lazy.h +1 -1
- data/contrib/zstd/lib/compress/zstd_ldm.c +32 -10
- data/contrib/zstd/lib/compress/zstd_ldm.h +7 -2
- data/contrib/zstd/lib/compress/zstd_opt.c +81 -114
- data/contrib/zstd/lib/compress/zstd_opt.h +1 -1
- data/contrib/zstd/lib/compress/zstdmt_compress.c +95 -51
- data/contrib/zstd/lib/compress/zstdmt_compress.h +3 -2
- data/contrib/zstd/lib/decompress/huf_decompress.c +76 -60
- data/contrib/zstd/lib/decompress/zstd_ddict.c +12 -8
- data/contrib/zstd/lib/decompress/zstd_ddict.h +2 -2
- data/contrib/zstd/lib/decompress/zstd_decompress.c +292 -172
- data/contrib/zstd/lib/decompress/zstd_decompress_block.c +459 -338
- data/contrib/zstd/lib/decompress/zstd_decompress_block.h +3 -3
- data/contrib/zstd/lib/decompress/zstd_decompress_internal.h +18 -4
- data/contrib/zstd/lib/deprecated/zbuff.h +9 -8
- data/contrib/zstd/lib/deprecated/zbuff_common.c +2 -2
- data/contrib/zstd/lib/deprecated/zbuff_compress.c +1 -1
- data/contrib/zstd/lib/deprecated/zbuff_decompress.c +1 -1
- data/contrib/zstd/lib/dictBuilder/cover.c +164 -54
- data/contrib/zstd/lib/dictBuilder/cover.h +52 -7
- data/contrib/zstd/lib/dictBuilder/fastcover.c +60 -43
- data/contrib/zstd/lib/dictBuilder/zdict.c +43 -19
- data/contrib/zstd/lib/dictBuilder/zdict.h +56 -28
- data/contrib/zstd/lib/legacy/zstd_legacy.h +8 -4
- data/contrib/zstd/lib/legacy/zstd_v01.c +110 -110
- data/contrib/zstd/lib/legacy/zstd_v01.h +1 -1
- data/contrib/zstd/lib/legacy/zstd_v02.c +23 -13
- data/contrib/zstd/lib/legacy/zstd_v02.h +1 -1
- data/contrib/zstd/lib/legacy/zstd_v03.c +23 -13
- data/contrib/zstd/lib/legacy/zstd_v03.h +1 -1
- data/contrib/zstd/lib/legacy/zstd_v04.c +30 -17
- data/contrib/zstd/lib/legacy/zstd_v04.h +1 -1
- data/contrib/zstd/lib/legacy/zstd_v05.c +113 -102
- data/contrib/zstd/lib/legacy/zstd_v05.h +2 -2
- data/contrib/zstd/lib/legacy/zstd_v06.c +20 -18
- data/contrib/zstd/lib/legacy/zstd_v06.h +1 -1
- data/contrib/zstd/lib/legacy/zstd_v07.c +25 -19
- data/contrib/zstd/lib/legacy/zstd_v07.h +1 -1
- data/contrib/zstd/lib/libzstd.pc.in +3 -2
- data/contrib/zstd/lib/zstd.h +265 -88
- data/ext/extzstd.h +1 -1
- data/ext/libzstd_conf.h +8 -0
- data/ext/zstd_common.c +1 -3
- data/ext/zstd_compress.c +3 -3
- data/ext/zstd_decompress.c +1 -5
- data/ext/zstd_dictbuilder.c +2 -3
- data/ext/zstd_dictbuilder_fastcover.c +1 -3
- data/ext/zstd_legacy_v01.c +2 -0
- data/ext/zstd_legacy_v02.c +2 -0
- data/ext/zstd_legacy_v03.c +2 -0
- data/ext/zstd_legacy_v04.c +2 -0
- data/ext/zstd_legacy_v05.c +2 -0
- data/ext/zstd_legacy_v06.c +2 -0
- data/ext/zstd_legacy_v07.c +2 -0
- data/lib/extzstd.rb +18 -10
- data/lib/extzstd/version.rb +1 -1
- metadata +15 -6
@@ -1,5 +1,5 @@
|
|
1
1
|
/*
|
2
|
-
* Copyright (c) 2016-
|
2
|
+
* Copyright (c) 2016-2020, Yann Collet, Facebook, Inc.
|
3
3
|
* All rights reserved.
|
4
4
|
*
|
5
5
|
* This source code is licensed under both the BSD-style license (found in the
|
@@ -18,7 +18,8 @@
|
|
18
18
|
/*-*************************************
|
19
19
|
* Dependencies
|
20
20
|
***************************************/
|
21
|
-
#include "zstd_internal.h"
|
21
|
+
#include "../common/zstd_internal.h"
|
22
|
+
#include "zstd_cwksp.h"
|
22
23
|
#ifdef ZSTD_MULTITHREAD
|
23
24
|
# include "zstdmt_compress.h"
|
24
25
|
#endif
|
@@ -33,13 +34,13 @@ extern "C" {
|
|
33
34
|
***************************************/
|
34
35
|
#define kSearchStrength 8
|
35
36
|
#define HASH_READ_SIZE 8
|
36
|
-
#define ZSTD_DUBT_UNSORTED_MARK 1 /* For btlazy2 strategy, index 1
|
37
|
+
#define ZSTD_DUBT_UNSORTED_MARK 1 /* For btlazy2 strategy, index ZSTD_DUBT_UNSORTED_MARK==1 means "unsorted".
|
37
38
|
It could be confused for a real successor at index "1", if sorted as larger than its predecessor.
|
38
39
|
It's not a big deal though : candidate will just be sorted again.
|
39
40
|
Additionally, candidate position 1 will be lost.
|
40
41
|
But candidate 1 cannot hide a large tree of candidates, so it's a minimal loss.
|
41
|
-
The benefit is that ZSTD_DUBT_UNSORTED_MARK cannot be mishandled after table re-use with a different strategy
|
42
|
-
|
42
|
+
The benefit is that ZSTD_DUBT_UNSORTED_MARK cannot be mishandled after table re-use with a different strategy.
|
43
|
+
This constant is required by ZSTD_compressBlock_btlazy2() and ZSTD_reduceTable_internal() */
|
43
44
|
|
44
45
|
|
45
46
|
/*-*************************************
|
@@ -128,21 +129,26 @@ typedef struct {
|
|
128
129
|
BYTE const* base; /* All regular indexes relative to this position */
|
129
130
|
BYTE const* dictBase; /* extDict indexes relative to this position */
|
130
131
|
U32 dictLimit; /* below that point, need extDict */
|
131
|
-
U32 lowLimit; /* below that point, no more data */
|
132
|
+
U32 lowLimit; /* below that point, no more valid data */
|
132
133
|
} ZSTD_window_t;
|
133
134
|
|
134
135
|
typedef struct ZSTD_matchState_t ZSTD_matchState_t;
|
135
136
|
struct ZSTD_matchState_t {
|
136
137
|
ZSTD_window_t window; /* State for window round buffer management */
|
137
|
-
U32 loadedDictEnd; /* index of end of dictionary
|
138
|
+
U32 loadedDictEnd; /* index of end of dictionary, within context's referential.
|
139
|
+
* When loadedDictEnd != 0, a dictionary is in use, and still valid.
|
140
|
+
* This relies on a mechanism to set loadedDictEnd=0 when dictionary is no longer within distance.
|
141
|
+
* Such mechanism is provided within ZSTD_window_enforceMaxDist() and ZSTD_checkDictValidity().
|
142
|
+
* When dict referential is copied into active context (i.e. not attached),
|
143
|
+
* loadedDictEnd == dictSize, since referential starts from zero.
|
144
|
+
*/
|
138
145
|
U32 nextToUpdate; /* index from which to continue table update */
|
139
|
-
U32
|
140
|
-
U32 hashLog3; /* dispatch table : larger == faster, more memory */
|
146
|
+
U32 hashLog3; /* dispatch table for matches of len==3 : larger == faster, more memory */
|
141
147
|
U32* hashTable;
|
142
148
|
U32* hashTable3;
|
143
149
|
U32* chainTable;
|
144
150
|
optState_t opt; /* optimal parser state */
|
145
|
-
const ZSTD_matchState_t
|
151
|
+
const ZSTD_matchState_t* dictMatchState;
|
146
152
|
ZSTD_compressionParameters cParams;
|
147
153
|
};
|
148
154
|
|
@@ -160,6 +166,7 @@ typedef struct {
|
|
160
166
|
typedef struct {
|
161
167
|
ZSTD_window_t window; /* State for the window round buffer management */
|
162
168
|
ldmEntry_t* hashTable;
|
169
|
+
U32 loadedDictEnd;
|
163
170
|
BYTE* bucketOffsets; /* Next position in bucket to insert entry */
|
164
171
|
U64 hashPower; /* Used to compute the rolling hash.
|
165
172
|
* Depends on ldmParams.minMatchLength */
|
@@ -187,6 +194,13 @@ typedef struct {
|
|
187
194
|
size_t capacity; /* The capacity starting from `seq` pointer */
|
188
195
|
} rawSeqStore_t;
|
189
196
|
|
197
|
+
typedef struct {
|
198
|
+
int collectSequences;
|
199
|
+
ZSTD_Sequence* seqStart;
|
200
|
+
size_t seqIndex;
|
201
|
+
size_t maxSequences;
|
202
|
+
} SeqCollector;
|
203
|
+
|
190
204
|
struct ZSTD_CCtx_params_s {
|
191
205
|
ZSTD_format_e format;
|
192
206
|
ZSTD_compressionParameters cParams;
|
@@ -195,6 +209,12 @@ struct ZSTD_CCtx_params_s {
|
|
195
209
|
int compressionLevel;
|
196
210
|
int forceWindow; /* force back-references to respect limit of
|
197
211
|
* 1<<wLog, even for dictionary */
|
212
|
+
size_t targetCBlockSize; /* Tries to fit compressed block size to be around targetCBlockSize.
|
213
|
+
* No target when targetCBlockSize == 0.
|
214
|
+
* There is no guarantee on compressed block size */
|
215
|
+
int srcSizeHint; /* User's best guess of source size.
|
216
|
+
* Hint is not valid when srcSizeHint == 0.
|
217
|
+
* There is no guarantee that hint is close to actual source size */
|
198
218
|
|
199
219
|
ZSTD_dictAttachPref_e attachDictPref;
|
200
220
|
ZSTD_literalCompressionMode_e literalCompressionMode;
|
@@ -220,9 +240,7 @@ struct ZSTD_CCtx_s {
|
|
220
240
|
ZSTD_CCtx_params appliedParams;
|
221
241
|
U32 dictID;
|
222
242
|
|
223
|
-
|
224
|
-
void* workSpace;
|
225
|
-
size_t workSpaceSize;
|
243
|
+
ZSTD_cwksp workspace; /* manages buffer for dynamic allocations */
|
226
244
|
size_t blockSize;
|
227
245
|
unsigned long long pledgedSrcSizePlusOne; /* this way, 0 (default) == unknown */
|
228
246
|
unsigned long long consumedSrcSize;
|
@@ -230,6 +248,9 @@ struct ZSTD_CCtx_s {
|
|
230
248
|
XXH64_state_t xxhState;
|
231
249
|
ZSTD_customMem customMem;
|
232
250
|
size_t staticSize;
|
251
|
+
SeqCollector seqCollector;
|
252
|
+
int isFirstBlock;
|
253
|
+
int initialized;
|
233
254
|
|
234
255
|
seqStore_t seqStore; /* sequences storage ptrs */
|
235
256
|
ldmState_t ldmState; /* long distance matching state */
|
@@ -305,26 +326,145 @@ MEM_STATIC U32 ZSTD_MLcode(U32 mlBase)
|
|
305
326
|
return (mlBase > 127) ? ZSTD_highbit32(mlBase) + ML_deltaCode : ML_Code[mlBase];
|
306
327
|
}
|
307
328
|
|
329
|
+
typedef struct repcodes_s {
|
330
|
+
U32 rep[3];
|
331
|
+
} repcodes_t;
|
332
|
+
|
333
|
+
MEM_STATIC repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
|
334
|
+
{
|
335
|
+
repcodes_t newReps;
|
336
|
+
if (offset >= ZSTD_REP_NUM) { /* full offset */
|
337
|
+
newReps.rep[2] = rep[1];
|
338
|
+
newReps.rep[1] = rep[0];
|
339
|
+
newReps.rep[0] = offset - ZSTD_REP_MOVE;
|
340
|
+
} else { /* repcode */
|
341
|
+
U32 const repCode = offset + ll0;
|
342
|
+
if (repCode > 0) { /* note : if repCode==0, no change */
|
343
|
+
U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
|
344
|
+
newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
|
345
|
+
newReps.rep[1] = rep[0];
|
346
|
+
newReps.rep[0] = currentOffset;
|
347
|
+
} else { /* repCode == 0 */
|
348
|
+
memcpy(&newReps, rep, sizeof(newReps));
|
349
|
+
}
|
350
|
+
}
|
351
|
+
return newReps;
|
352
|
+
}
|
353
|
+
|
354
|
+
/* ZSTD_cParam_withinBounds:
|
355
|
+
* @return 1 if value is within cParam bounds,
|
356
|
+
* 0 otherwise */
|
357
|
+
MEM_STATIC int ZSTD_cParam_withinBounds(ZSTD_cParameter cParam, int value)
|
358
|
+
{
|
359
|
+
ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam);
|
360
|
+
if (ZSTD_isError(bounds.error)) return 0;
|
361
|
+
if (value < bounds.lowerBound) return 0;
|
362
|
+
if (value > bounds.upperBound) return 0;
|
363
|
+
return 1;
|
364
|
+
}
|
365
|
+
|
366
|
+
/* ZSTD_noCompressBlock() :
|
367
|
+
* Writes uncompressed block to dst buffer from given src.
|
368
|
+
* Returns the size of the block */
|
369
|
+
MEM_STATIC size_t ZSTD_noCompressBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize, U32 lastBlock)
|
370
|
+
{
|
371
|
+
U32 const cBlockHeader24 = lastBlock + (((U32)bt_raw)<<1) + (U32)(srcSize << 3);
|
372
|
+
RETURN_ERROR_IF(srcSize + ZSTD_blockHeaderSize > dstCapacity,
|
373
|
+
dstSize_tooSmall, "dst buf too small for uncompressed block");
|
374
|
+
MEM_writeLE24(dst, cBlockHeader24);
|
375
|
+
memcpy((BYTE*)dst + ZSTD_blockHeaderSize, src, srcSize);
|
376
|
+
return ZSTD_blockHeaderSize + srcSize;
|
377
|
+
}
|
378
|
+
|
379
|
+
MEM_STATIC size_t ZSTD_rleCompressBlock (void* dst, size_t dstCapacity, BYTE src, size_t srcSize, U32 lastBlock)
|
380
|
+
{
|
381
|
+
BYTE* const op = (BYTE*)dst;
|
382
|
+
U32 const cBlockHeader = lastBlock + (((U32)bt_rle)<<1) + (U32)(srcSize << 3);
|
383
|
+
RETURN_ERROR_IF(dstCapacity < 4, dstSize_tooSmall, "");
|
384
|
+
MEM_writeLE24(op, cBlockHeader);
|
385
|
+
op[3] = src;
|
386
|
+
return 4;
|
387
|
+
}
|
388
|
+
|
389
|
+
|
390
|
+
/* ZSTD_minGain() :
|
391
|
+
* minimum compression required
|
392
|
+
* to generate a compress block or a compressed literals section.
|
393
|
+
* note : use same formula for both situations */
|
394
|
+
MEM_STATIC size_t ZSTD_minGain(size_t srcSize, ZSTD_strategy strat)
|
395
|
+
{
|
396
|
+
U32 const minlog = (strat>=ZSTD_btultra) ? (U32)(strat) - 1 : 6;
|
397
|
+
ZSTD_STATIC_ASSERT(ZSTD_btultra == 8);
|
398
|
+
assert(ZSTD_cParam_withinBounds(ZSTD_c_strategy, strat));
|
399
|
+
return (srcSize >> minlog) + 2;
|
400
|
+
}
|
401
|
+
|
402
|
+
MEM_STATIC int ZSTD_disableLiteralsCompression(const ZSTD_CCtx_params* cctxParams)
|
403
|
+
{
|
404
|
+
switch (cctxParams->literalCompressionMode) {
|
405
|
+
case ZSTD_lcm_huffman:
|
406
|
+
return 0;
|
407
|
+
case ZSTD_lcm_uncompressed:
|
408
|
+
return 1;
|
409
|
+
default:
|
410
|
+
assert(0 /* impossible: pre-validated */);
|
411
|
+
/* fall-through */
|
412
|
+
case ZSTD_lcm_auto:
|
413
|
+
return (cctxParams->cParams.strategy == ZSTD_fast) && (cctxParams->cParams.targetLength > 0);
|
414
|
+
}
|
415
|
+
}
|
416
|
+
|
417
|
+
/*! ZSTD_safecopyLiterals() :
|
418
|
+
* memcpy() function that won't read beyond more than WILDCOPY_OVERLENGTH bytes past ilimit_w.
|
419
|
+
* Only called when the sequence ends past ilimit_w, so it only needs to be optimized for single
|
420
|
+
* large copies.
|
421
|
+
*/
|
422
|
+
static void ZSTD_safecopyLiterals(BYTE* op, BYTE const* ip, BYTE const* const iend, BYTE const* ilimit_w) {
|
423
|
+
assert(iend > ilimit_w);
|
424
|
+
if (ip <= ilimit_w) {
|
425
|
+
ZSTD_wildcopy(op, ip, ilimit_w - ip, ZSTD_no_overlap);
|
426
|
+
op += ilimit_w - ip;
|
427
|
+
ip = ilimit_w;
|
428
|
+
}
|
429
|
+
while (ip < iend) *op++ = *ip++;
|
430
|
+
}
|
431
|
+
|
308
432
|
/*! ZSTD_storeSeq() :
|
309
|
-
* Store a sequence (
|
310
|
-
* `
|
433
|
+
* Store a sequence (litlen, litPtr, offCode and mlBase) into seqStore_t.
|
434
|
+
* `offCode` : distance to match + ZSTD_REP_MOVE (values <= ZSTD_REP_MOVE are repCodes).
|
311
435
|
* `mlBase` : matchLength - MINMATCH
|
436
|
+
* Allowed to overread literals up to litLimit.
|
312
437
|
*/
|
313
|
-
|
438
|
+
HINT_INLINE UNUSED_ATTR
|
439
|
+
void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const BYTE* literals, const BYTE* litLimit, U32 offCode, size_t mlBase)
|
314
440
|
{
|
441
|
+
BYTE const* const litLimit_w = litLimit - WILDCOPY_OVERLENGTH;
|
442
|
+
BYTE const* const litEnd = literals + litLength;
|
315
443
|
#if defined(DEBUGLEVEL) && (DEBUGLEVEL >= 6)
|
316
444
|
static const BYTE* g_start = NULL;
|
317
445
|
if (g_start==NULL) g_start = (const BYTE*)literals; /* note : index only works for compression within a single segment */
|
318
446
|
{ U32 const pos = (U32)((const BYTE*)literals - g_start);
|
319
447
|
DEBUGLOG(6, "Cpos%7u :%3u literals, match%4u bytes at offCode%7u",
|
320
|
-
pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)
|
448
|
+
pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)offCode);
|
321
449
|
}
|
322
450
|
#endif
|
323
451
|
assert((size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart) < seqStorePtr->maxNbSeq);
|
324
452
|
/* copy Literals */
|
325
453
|
assert(seqStorePtr->maxNbLit <= 128 KB);
|
326
454
|
assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + seqStorePtr->maxNbLit);
|
327
|
-
|
455
|
+
assert(literals + litLength <= litLimit);
|
456
|
+
if (litEnd <= litLimit_w) {
|
457
|
+
/* Common case we can use wildcopy.
|
458
|
+
* First copy 16 bytes, because literals are likely short.
|
459
|
+
*/
|
460
|
+
assert(WILDCOPY_OVERLENGTH >= 16);
|
461
|
+
ZSTD_copy16(seqStorePtr->lit, literals);
|
462
|
+
if (litLength > 16) {
|
463
|
+
ZSTD_wildcopy(seqStorePtr->lit+16, literals+16, (ptrdiff_t)litLength-16, ZSTD_no_overlap);
|
464
|
+
}
|
465
|
+
} else {
|
466
|
+
ZSTD_safecopyLiterals(seqStorePtr->lit, literals, litEnd, litLimit_w);
|
467
|
+
}
|
328
468
|
seqStorePtr->lit += litLength;
|
329
469
|
|
330
470
|
/* literal Length */
|
@@ -336,7 +476,7 @@ MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const v
|
|
336
476
|
seqStorePtr->sequences[0].litLength = (U16)litLength;
|
337
477
|
|
338
478
|
/* match offset */
|
339
|
-
seqStorePtr->sequences[0].offset =
|
479
|
+
seqStorePtr->sequences[0].offset = offCode + 1;
|
340
480
|
|
341
481
|
/* match Length */
|
342
482
|
if (mlBase>0xFFFF) {
|
@@ -359,8 +499,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
|
|
359
499
|
if (MEM_64bits()) {
|
360
500
|
# if defined(_MSC_VER) && defined(_WIN64)
|
361
501
|
unsigned long r = 0;
|
362
|
-
_BitScanForward64( &r, (U64)val );
|
363
|
-
return (unsigned)(r>>3);
|
502
|
+
return _BitScanForward64( &r, (U64)val ) ? (unsigned)(r >> 3) : 0;
|
364
503
|
# elif defined(__GNUC__) && (__GNUC__ >= 4)
|
365
504
|
return (__builtin_ctzll((U64)val) >> 3);
|
366
505
|
# else
|
@@ -377,8 +516,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
|
|
377
516
|
} else { /* 32 bits */
|
378
517
|
# if defined(_MSC_VER)
|
379
518
|
unsigned long r=0;
|
380
|
-
_BitScanForward( &r, (U32)val );
|
381
|
-
return (unsigned)(r>>3);
|
519
|
+
return _BitScanForward( &r, (U32)val ) ? (unsigned)(r >> 3) : 0;
|
382
520
|
# elif defined(__GNUC__) && (__GNUC__ >= 3)
|
383
521
|
return (__builtin_ctz((U32)val) >> 3);
|
384
522
|
# else
|
@@ -393,8 +531,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
|
|
393
531
|
if (MEM_64bits()) {
|
394
532
|
# if defined(_MSC_VER) && defined(_WIN64)
|
395
533
|
unsigned long r = 0;
|
396
|
-
_BitScanReverse64( &r, val );
|
397
|
-
return (unsigned)(r>>3);
|
534
|
+
return _BitScanReverse64( &r, val ) ? (unsigned)(r >> 3) : 0;
|
398
535
|
# elif defined(__GNUC__) && (__GNUC__ >= 4)
|
399
536
|
return (__builtin_clzll(val) >> 3);
|
400
537
|
# else
|
@@ -408,8 +545,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
|
|
408
545
|
} else { /* 32 bits */
|
409
546
|
# if defined(_MSC_VER)
|
410
547
|
unsigned long r = 0;
|
411
|
-
_BitScanReverse( &r, (unsigned long)val );
|
412
|
-
return (unsigned)(r>>3);
|
548
|
+
return _BitScanReverse( &r, (unsigned long)val ) ? (unsigned)(r >> 3) : 0;
|
413
549
|
# elif defined(__GNUC__) && (__GNUC__ >= 3)
|
414
550
|
return (__builtin_clz((U32)val) >> 3);
|
415
551
|
# else
|
@@ -564,6 +700,9 @@ MEM_STATIC U64 ZSTD_rollingHash_rotate(U64 hash, BYTE toRemove, BYTE toAdd, U64
|
|
564
700
|
/*-*************************************
|
565
701
|
* Round buffer management
|
566
702
|
***************************************/
|
703
|
+
#if (ZSTD_WINDOWLOG_MAX_64 > 31)
|
704
|
+
# error "ZSTD_WINDOWLOG_MAX is too large : would overflow ZSTD_CURRENT_MAX"
|
705
|
+
#endif
|
567
706
|
/* Max current allowed */
|
568
707
|
#define ZSTD_CURRENT_MAX ((3U << 29) + (1U << ZSTD_WINDOWLOG_MAX))
|
569
708
|
/* Maximum chunk size before overflow correction needs to be called again */
|
@@ -653,7 +792,10 @@ MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
|
|
653
792
|
*/
|
654
793
|
U32 const cycleMask = (1U << cycleLog) - 1;
|
655
794
|
U32 const current = (U32)((BYTE const*)src - window->base);
|
656
|
-
U32 const
|
795
|
+
U32 const currentCycle0 = current & cycleMask;
|
796
|
+
/* Exclude zero so that newCurrent - maxDist >= 1. */
|
797
|
+
U32 const currentCycle1 = currentCycle0 == 0 ? (1U << cycleLog) : currentCycle0;
|
798
|
+
U32 const newCurrent = currentCycle1 + maxDist;
|
657
799
|
U32 const correction = current - newCurrent;
|
658
800
|
assert((maxDist & cycleMask) == 0);
|
659
801
|
assert(current > newCurrent);
|
@@ -662,8 +804,17 @@ MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
|
|
662
804
|
|
663
805
|
window->base += correction;
|
664
806
|
window->dictBase += correction;
|
665
|
-
window->lowLimit
|
666
|
-
window->
|
807
|
+
if (window->lowLimit <= correction) window->lowLimit = 1;
|
808
|
+
else window->lowLimit -= correction;
|
809
|
+
if (window->dictLimit <= correction) window->dictLimit = 1;
|
810
|
+
else window->dictLimit -= correction;
|
811
|
+
|
812
|
+
/* Ensure we can still reference the full window. */
|
813
|
+
assert(newCurrent >= maxDist);
|
814
|
+
assert(newCurrent - maxDist >= 1);
|
815
|
+
/* Ensure that lowLimit and dictLimit didn't underflow. */
|
816
|
+
assert(window->lowLimit <= newCurrent);
|
817
|
+
assert(window->dictLimit <= newCurrent);
|
667
818
|
|
668
819
|
DEBUGLOG(4, "Correction of 0x%x bytes to lowLimit=0x%x", correction,
|
669
820
|
window->lowLimit);
|
@@ -675,31 +826,49 @@ MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
|
|
675
826
|
* Updates lowLimit so that:
|
676
827
|
* (srcEnd - base) - lowLimit == maxDist + loadedDictEnd
|
677
828
|
*
|
678
|
-
*
|
679
|
-
* This must be called before a block compression call
|
680
|
-
*
|
829
|
+
* It ensures index is valid as long as index >= lowLimit.
|
830
|
+
* This must be called before a block compression call.
|
831
|
+
*
|
832
|
+
* loadedDictEnd is only defined if a dictionary is in use for current compression.
|
833
|
+
* As the name implies, loadedDictEnd represents the index at end of dictionary.
|
834
|
+
* The value lies within context's referential, it can be directly compared to blockEndIdx.
|
681
835
|
*
|
682
|
-
* If loadedDictEndPtr is
|
683
|
-
*
|
684
|
-
*
|
685
|
-
*
|
686
|
-
*
|
836
|
+
* If loadedDictEndPtr is NULL, no dictionary is in use, and we use loadedDictEnd == 0.
|
837
|
+
* If loadedDictEndPtr is not NULL, we set it to zero after updating lowLimit.
|
838
|
+
* This is because dictionaries are allowed to be referenced fully
|
839
|
+
* as long as the last byte of the dictionary is in the window.
|
840
|
+
* Once input has progressed beyond window size, dictionary cannot be referenced anymore.
|
687
841
|
*
|
688
|
-
* In normal dict mode, the
|
689
|
-
* dictMatchState mode, lowLimit and dictLimit are the same,
|
690
|
-
* is below them.
|
842
|
+
* In normal dict mode, the dictionary lies between lowLimit and dictLimit.
|
843
|
+
* In dictMatchState mode, lowLimit and dictLimit are the same,
|
844
|
+
* and the dictionary is below them.
|
845
|
+
* forceWindow and dictMatchState are therefore incompatible.
|
691
846
|
*/
|
692
847
|
MEM_STATIC void
|
693
848
|
ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
|
694
|
-
|
695
|
-
U32
|
696
|
-
U32*
|
849
|
+
const void* blockEnd,
|
850
|
+
U32 maxDist,
|
851
|
+
U32* loadedDictEndPtr,
|
697
852
|
const ZSTD_matchState_t** dictMatchStatePtr)
|
698
853
|
{
|
699
|
-
U32 const blockEndIdx = (U32)((BYTE const*)
|
700
|
-
U32 loadedDictEnd = (loadedDictEndPtr != NULL) ? *loadedDictEndPtr : 0;
|
701
|
-
DEBUGLOG(5, "ZSTD_window_enforceMaxDist: blockEndIdx=%u, maxDist=%u",
|
702
|
-
(unsigned)blockEndIdx, (unsigned)maxDist);
|
854
|
+
U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
|
855
|
+
U32 const loadedDictEnd = (loadedDictEndPtr != NULL) ? *loadedDictEndPtr : 0;
|
856
|
+
DEBUGLOG(5, "ZSTD_window_enforceMaxDist: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
|
857
|
+
(unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
|
858
|
+
|
859
|
+
/* - When there is no dictionary : loadedDictEnd == 0.
|
860
|
+
In which case, the test (blockEndIdx > maxDist) is merely to avoid
|
861
|
+
overflowing next operation `newLowLimit = blockEndIdx - maxDist`.
|
862
|
+
- When there is a standard dictionary :
|
863
|
+
Index referential is copied from the dictionary,
|
864
|
+
which means it starts from 0.
|
865
|
+
In which case, loadedDictEnd == dictSize,
|
866
|
+
and it makes sense to compare `blockEndIdx > maxDist + dictSize`
|
867
|
+
since `blockEndIdx` also starts from zero.
|
868
|
+
- When there is an attached dictionary :
|
869
|
+
loadedDictEnd is expressed within the referential of the context,
|
870
|
+
so it can be directly compared against blockEndIdx.
|
871
|
+
*/
|
703
872
|
if (blockEndIdx > maxDist + loadedDictEnd) {
|
704
873
|
U32 const newLowLimit = blockEndIdx - maxDist;
|
705
874
|
if (window->lowLimit < newLowLimit) window->lowLimit = newLowLimit;
|
@@ -708,11 +877,54 @@ ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
|
|
708
877
|
(unsigned)window->dictLimit, (unsigned)window->lowLimit);
|
709
878
|
window->dictLimit = window->lowLimit;
|
710
879
|
}
|
711
|
-
|
880
|
+
/* On reaching window size, dictionaries are invalidated */
|
881
|
+
if (loadedDictEndPtr) *loadedDictEndPtr = 0;
|
882
|
+
if (dictMatchStatePtr) *dictMatchStatePtr = NULL;
|
883
|
+
}
|
884
|
+
}
|
885
|
+
|
886
|
+
/* Similar to ZSTD_window_enforceMaxDist(),
|
887
|
+
* but only invalidates dictionary
|
888
|
+
* when input progresses beyond window size.
|
889
|
+
* assumption : loadedDictEndPtr and dictMatchStatePtr are valid (non NULL)
|
890
|
+
* loadedDictEnd uses same referential as window->base
|
891
|
+
* maxDist is the window size */
|
892
|
+
MEM_STATIC void
|
893
|
+
ZSTD_checkDictValidity(const ZSTD_window_t* window,
|
894
|
+
const void* blockEnd,
|
895
|
+
U32 maxDist,
|
896
|
+
U32* loadedDictEndPtr,
|
897
|
+
const ZSTD_matchState_t** dictMatchStatePtr)
|
898
|
+
{
|
899
|
+
assert(loadedDictEndPtr != NULL);
|
900
|
+
assert(dictMatchStatePtr != NULL);
|
901
|
+
{ U32 const blockEndIdx = (U32)((BYTE const*)blockEnd - window->base);
|
902
|
+
U32 const loadedDictEnd = *loadedDictEndPtr;
|
903
|
+
DEBUGLOG(5, "ZSTD_checkDictValidity: blockEndIdx=%u, maxDist=%u, loadedDictEnd=%u",
|
904
|
+
(unsigned)blockEndIdx, (unsigned)maxDist, (unsigned)loadedDictEnd);
|
905
|
+
assert(blockEndIdx >= loadedDictEnd);
|
906
|
+
|
907
|
+
if (blockEndIdx > loadedDictEnd + maxDist) {
|
908
|
+
/* On reaching window size, dictionaries are invalidated.
|
909
|
+
* For simplification, if window size is reached anywhere within next block,
|
910
|
+
* the dictionary is invalidated for the full block.
|
911
|
+
*/
|
912
|
+
DEBUGLOG(6, "invalidating dictionary for current block (distance > windowSize)");
|
712
913
|
*loadedDictEndPtr = 0;
|
713
|
-
if (dictMatchStatePtr)
|
714
914
|
*dictMatchStatePtr = NULL;
|
715
|
-
|
915
|
+
} else {
|
916
|
+
if (*loadedDictEndPtr != 0) {
|
917
|
+
DEBUGLOG(6, "dictionary considered valid for current block");
|
918
|
+
} } }
|
919
|
+
}
|
920
|
+
|
921
|
+
MEM_STATIC void ZSTD_window_init(ZSTD_window_t* window) {
|
922
|
+
memset(window, 0, sizeof(*window));
|
923
|
+
window->base = (BYTE const*)"";
|
924
|
+
window->dictBase = (BYTE const*)"";
|
925
|
+
window->dictLimit = 1; /* start from 1, so that 1st position is valid */
|
926
|
+
window->lowLimit = 1; /* it ensures first and later CCtx usages compress the same */
|
927
|
+
window->nextSrc = window->base + 1; /* see issue #1241 */
|
716
928
|
}
|
717
929
|
|
718
930
|
/**
|
@@ -728,6 +940,10 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
|
|
728
940
|
BYTE const* const ip = (BYTE const*)src;
|
729
941
|
U32 contiguous = 1;
|
730
942
|
DEBUGLOG(5, "ZSTD_window_update");
|
943
|
+
if (srcSize == 0)
|
944
|
+
return contiguous;
|
945
|
+
assert(window->base != NULL);
|
946
|
+
assert(window->dictBase != NULL);
|
731
947
|
/* Check if blocks follow each other */
|
732
948
|
if (src != window->nextSrc) {
|
733
949
|
/* not contiguous */
|
@@ -738,7 +954,7 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
|
|
738
954
|
window->dictLimit = (U32)distanceFromBase;
|
739
955
|
window->dictBase = window->base;
|
740
956
|
window->base = ip - distanceFromBase;
|
741
|
-
|
957
|
+
/* ms->nextToUpdate = window->dictLimit; */
|
742
958
|
if (window->dictLimit - window->lowLimit < HASH_READ_SIZE) window->lowLimit = window->dictLimit; /* too small extDict */
|
743
959
|
contiguous = 0;
|
744
960
|
}
|
@@ -754,6 +970,33 @@ MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
|
|
754
970
|
return contiguous;
|
755
971
|
}
|
756
972
|
|
973
|
+
/**
|
974
|
+
* Returns the lowest allowed match index. It may either be in the ext-dict or the prefix.
|
975
|
+
*/
|
976
|
+
MEM_STATIC U32 ZSTD_getLowestMatchIndex(const ZSTD_matchState_t* ms, U32 current, unsigned windowLog)
|
977
|
+
{
|
978
|
+
U32 const maxDistance = 1U << windowLog;
|
979
|
+
U32 const lowestValid = ms->window.lowLimit;
|
980
|
+
U32 const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
|
981
|
+
U32 const isDictionary = (ms->loadedDictEnd != 0);
|
982
|
+
U32 const matchLowest = isDictionary ? lowestValid : withinWindow;
|
983
|
+
return matchLowest;
|
984
|
+
}
|
985
|
+
|
986
|
+
/**
|
987
|
+
* Returns the lowest allowed match index in the prefix.
|
988
|
+
*/
|
989
|
+
MEM_STATIC U32 ZSTD_getLowestPrefixIndex(const ZSTD_matchState_t* ms, U32 current, unsigned windowLog)
|
990
|
+
{
|
991
|
+
U32 const maxDistance = 1U << windowLog;
|
992
|
+
U32 const lowestValid = ms->window.dictLimit;
|
993
|
+
U32 const withinWindow = (current - lowestValid > maxDistance) ? current - maxDistance : lowestValid;
|
994
|
+
U32 const isDictionary = (ms->loadedDictEnd != 0);
|
995
|
+
U32 const matchLowest = isDictionary ? lowestValid : withinWindow;
|
996
|
+
return matchLowest;
|
997
|
+
}
|
998
|
+
|
999
|
+
|
757
1000
|
|
758
1001
|
/* debug functions */
|
759
1002
|
#if (DEBUGLEVEL>=2)
|
@@ -791,6 +1034,21 @@ MEM_STATIC void ZSTD_debugTable(const U32* table, U32 max)
|
|
791
1034
|
}
|
792
1035
|
#endif
|
793
1036
|
|
1037
|
+
/* ===============================================================
|
1038
|
+
* Shared internal declarations
|
1039
|
+
* These prototypes may be called from sources not in lib/compress
|
1040
|
+
* =============================================================== */
|
1041
|
+
|
1042
|
+
/* ZSTD_loadCEntropy() :
|
1043
|
+
* dict : must point at beginning of a valid zstd dictionary.
|
1044
|
+
* return : size of dictionary header (size of magic number + dict ID + entropy tables)
|
1045
|
+
* assumptions : magic number supposed already checked
|
1046
|
+
* and dictSize >= 8 */
|
1047
|
+
size_t ZSTD_loadCEntropy(ZSTD_compressedBlockState_t* bs, void* workspace,
|
1048
|
+
short* offcodeNCount, unsigned* offcodeMaxValue,
|
1049
|
+
const void* const dict, size_t dictSize);
|
1050
|
+
|
1051
|
+
void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs);
|
794
1052
|
|
795
1053
|
/* ==============================================================
|
796
1054
|
* Private declarations
|
@@ -800,6 +1058,7 @@ MEM_STATIC void ZSTD_debugTable(const U32* table, U32 max)
|
|
800
1058
|
/* ZSTD_getCParamsFromCCtxParams() :
|
801
1059
|
* cParams are built depending on compressionLevel, src size hints,
|
802
1060
|
* LDM and manually set compression parameters.
|
1061
|
+
* Note: srcSizeHint == 0 means 0!
|
803
1062
|
*/
|
804
1063
|
ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
|
805
1064
|
const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize);
|
@@ -812,7 +1071,7 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
|
|
812
1071
|
size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
|
813
1072
|
const void* dict, size_t dictSize,
|
814
1073
|
const ZSTD_CDict* cdict,
|
815
|
-
ZSTD_CCtx_params
|
1074
|
+
const ZSTD_CCtx_params* params, unsigned long long pledgedSrcSize);
|
816
1075
|
|
817
1076
|
void ZSTD_resetSeqStore(seqStore_t* ssPtr);
|
818
1077
|
|
@@ -827,7 +1086,7 @@ size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
|
|
827
1086
|
ZSTD_dictContentType_e dictContentType,
|
828
1087
|
ZSTD_dictTableLoadMethod_e dtlm,
|
829
1088
|
const ZSTD_CDict* cdict,
|
830
|
-
ZSTD_CCtx_params params,
|
1089
|
+
const ZSTD_CCtx_params* params,
|
831
1090
|
unsigned long long pledgedSrcSize);
|
832
1091
|
|
833
1092
|
/* ZSTD_compress_advanced_internal() :
|
@@ -836,7 +1095,7 @@ size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
|
|
836
1095
|
void* dst, size_t dstCapacity,
|
837
1096
|
const void* src, size_t srcSize,
|
838
1097
|
const void* dict,size_t dictSize,
|
839
|
-
ZSTD_CCtx_params params);
|
1098
|
+
const ZSTD_CCtx_params* params);
|
840
1099
|
|
841
1100
|
|
842
1101
|
/* ZSTD_writeLastEmptyBlock() :
|
@@ -859,5 +1118,8 @@ size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity);
|
|
859
1118
|
*/
|
860
1119
|
size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq);
|
861
1120
|
|
1121
|
+
/** ZSTD_cycleLog() :
|
1122
|
+
* condition for correct operation : hashLog > 1 */
|
1123
|
+
U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat);
|
862
1124
|
|
863
1125
|
#endif /* ZSTD_COMPRESS_H */
|