zstd/lib/common/zstd_internal.h

256 lines
8.9 KiB
C
Raw Normal View History

/*
2016-08-30 17:04:33 +00:00
* Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
* All rights reserved.
*
* This source code is licensed under both the BSD-style license (found in the
* LICENSE file in the root directory of this source tree) and the GPLv2 (found
* in the COPYING file in the root directory of this source tree).
* You may select, at your option, one of the above-listed licenses.
2016-08-30 17:04:33 +00:00
*/
#ifndef ZSTD_CCOMMON_H_MODULE
#define ZSTD_CCOMMON_H_MODULE
/* this module contains definitions which must be identical
* across compression, decompression and dictBuilder.
* It also contains a few functions useful to at least 2 of them
* and which benefit from being inlined */
2016-09-06 13:05:19 +00:00
2016-02-11 23:07:30 +00:00
/*-*************************************
2016-02-04 14:28:14 +00:00
* Dependencies
***************************************/
2017-08-15 00:20:50 +00:00
#include "compiler.h"
#include "mem.h"
#include "debug.h" /* assert, DEBUGLOG, RAWLOG, g_debuglevel */
#include "error_private.h"
#define ZSTD_STATIC_LINKING_ONLY
#include "zstd.h"
#define FSE_STATIC_LINKING_ONLY
#include "fse.h"
#define HUF_STATIC_LINKING_ONLY
#include "huf.h"
#ifndef XXH_STATIC_LINKING_ONLY
# define XXH_STATIC_LINKING_ONLY /* XXH64_state_t */
#endif
#include "xxhash.h" /* XXH_reset, update, digest */
2017-08-25 01:09:50 +00:00
#if defined (__cplusplus)
extern "C" {
#endif
/* ---- static assert (debug) --- */
#define ZSTD_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c)
2016-02-11 23:07:30 +00:00
/*-*************************************
2016-09-06 13:36:19 +00:00
* shared macros
2015-11-11 20:38:21 +00:00
***************************************/
#undef MIN
#undef MAX
2015-10-31 11:57:14 +00:00
#define MIN(a,b) ((a)<(b) ? (a) : (b))
2015-11-11 20:38:21 +00:00
#define MAX(a,b) ((a)>(b) ? (a) : (b))
2016-09-06 13:36:19 +00:00
#define CHECK_F(f) { size_t const errcod = f; if (ERR_isError(errcod)) return errcod; } /* check and Forward error code */
2016-09-06 14:38:51 +00:00
#define CHECK_E(f, e) { size_t const errcod = f; if (ERR_isError(errcod)) return ERROR(e); } /* check and send Error code */
2015-10-31 11:57:14 +00:00
2016-02-11 23:07:30 +00:00
/*-*************************************
2015-11-11 20:38:21 +00:00
* Common constants
***************************************/
2016-03-02 14:56:24 +00:00
#define ZSTD_OPT_NUM (1<<12)
2015-11-25 13:42:45 +00:00
#define ZSTD_REP_NUM 3 /* number of repcodes */
#define ZSTD_REP_MOVE (ZSTD_REP_NUM-1)
2016-06-13 23:49:25 +00:00
static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
2015-11-25 13:42:45 +00:00
2015-11-11 20:38:21 +00:00
#define KB *(1 <<10)
#define MB *(1 <<20)
#define GB *(1U<<30)
2015-11-11 20:38:21 +00:00
#define BIT7 128
#define BIT6 64
#define BIT5 32
#define BIT4 16
#define BIT1 2
#define BIT0 1
#define ZSTD_WINDOWLOG_ABSOLUTEMIN 10
#define ZSTD_WINDOWLOG_DEFAULTMAX 27 /* Default maximum allowed window log */
static const size_t ZSTD_fcs_fieldSize[4] = { 0, 2, 4, 8 };
2016-05-29 03:01:04 +00:00
static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 };
2016-03-19 14:11:42 +00:00
#define ZSTD_FRAMEIDSIZE 4 /* magic number size */
#define ZSTD_BLOCKHEADERSIZE 3 /* C standard doesn't allow `static const` variable to be init using another `static const` variable */
2016-03-19 14:11:42 +00:00
static const size_t ZSTD_blockHeaderSize = ZSTD_BLOCKHEADERSIZE;
2016-07-27 22:55:43 +00:00
typedef enum { bt_raw, bt_rle, bt_compressed, bt_reserved } blockType_e;
2016-03-19 14:11:42 +00:00
#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */
#define HufLog 12
2016-07-23 14:31:49 +00:00
typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingType_e;
2015-11-05 16:32:18 +00:00
2016-03-19 14:11:42 +00:00
#define LONGNBSEQ 0x7F00
2016-04-06 07:46:01 +00:00
#define MINMATCH 3
2015-11-11 20:38:21 +00:00
2016-02-03 17:47:30 +00:00
#define Litbits 8
2016-02-03 21:56:55 +00:00
#define MaxLit ((1<<Litbits) - 1)
#define MaxML 52
#define MaxLL 35
2017-09-18 23:54:53 +00:00
#define DefaultMaxOff 28
#define MaxOff 31
2016-03-18 21:23:49 +00:00
#define MaxSeq MAX(MaxLL, MaxML) /* Assumption : MaxOff < MaxLL,MaxML */
#define MLFSELog 9
2016-03-20 23:07:42 +00:00
#define LLFSELog 9
2016-03-24 01:31:27 +00:00
#define OffFSELog 8
#define MaxFSELog MAX(MAX(MLFSELog, LLFSELog), OffFSELog)
2016-03-04 19:04:25 +00:00
static const U32 LL_bits[MaxLL+1] = { 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 1, 1, 2, 2, 3, 3,
4, 6, 7, 8, 9,10,11,12,
13,14,15,16 };
static const S16 LL_defaultNorm[MaxLL+1] = { 4, 3, 2, 2, 2, 2, 2, 2,
2, 2, 2, 2, 2, 1, 1, 1,
2, 2, 2, 2, 2, 2, 2, 2,
2, 3, 2, 1, 1, 1, 1, 1,
-1,-1,-1,-1 };
#define LL_DEFAULTNORMLOG 6 /* for static allocation */
static const U32 LL_defaultNormLog = LL_DEFAULTNORMLOG;
static const U32 ML_bits[MaxML+1] = { 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
1, 1, 1, 1, 2, 2, 3, 3,
4, 4, 5, 7, 8, 9,10,11,
12,13,14,15,16 };
static const S16 ML_defaultNorm[MaxML+1] = { 1, 4, 3, 2, 2, 2, 2, 2,
2, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1,-1,-1,
-1,-1,-1,-1,-1 };
#define ML_DEFAULTNORMLOG 6 /* for static allocation */
static const U32 ML_defaultNormLog = ML_DEFAULTNORMLOG;
static const S16 OF_defaultNorm[DefaultMaxOff+1] = { 1, 1, 1, 1, 1, 1, 2, 2,
2, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1,
-1,-1,-1,-1,-1 };
#define OF_DEFAULTNORMLOG 5 /* for static allocation */
static const U32 OF_defaultNormLog = OF_DEFAULTNORMLOG;
2016-01-27 23:18:06 +00:00
/*-*******************************************
2015-11-11 20:38:21 +00:00
* Shared functions to include for inlining
2016-01-27 23:18:06 +00:00
*********************************************/
2015-11-11 20:38:21 +00:00
static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
#define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
2016-02-04 14:28:14 +00:00
/*! ZSTD_wildcopy() :
2017-11-17 19:40:08 +00:00
* custom version of memcpy(), can overwrite up to WILDCOPY_OVERLENGTH bytes (if length==0) */
2016-03-19 14:11:42 +00:00
#define WILDCOPY_OVERLENGTH 8
MEM_STATIC void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length)
2015-11-11 20:38:21 +00:00
{
const BYTE* ip = (const BYTE*)src;
BYTE* op = (BYTE*)dst;
BYTE* const oend = op + length;
do
COPY8(op, ip)
2016-08-07 22:44:00 +00:00
while (op < oend);
}
MEM_STATIC void ZSTD_wildcopy_e(void* dst, const void* src, void* dstEnd) /* should be faster for decoding, but strangely, not verified on all platform */
{
const BYTE* ip = (const BYTE*)src;
BYTE* op = (BYTE*)dst;
BYTE* const oend = (BYTE*)dstEnd;
do
COPY8(op, ip)
2015-11-11 20:38:21 +00:00
while (op < oend);
}
2016-02-11 23:07:30 +00:00
/*-*******************************************
* Private declarations
2016-02-11 23:07:30 +00:00
*********************************************/
2016-07-29 19:22:17 +00:00
typedef struct seqDef_s {
U32 offset;
U16 litLength;
U16 matchLength;
} seqDef;
2016-02-11 23:07:30 +00:00
typedef struct {
2016-07-29 22:55:13 +00:00
seqDef* sequencesStart;
2016-07-29 19:22:17 +00:00
seqDef* sequences;
2016-02-11 23:07:30 +00:00
BYTE* litStart;
BYTE* lit;
2016-07-29 19:22:17 +00:00
BYTE* llCode;
BYTE* mlCode;
BYTE* ofCode;
2016-04-07 15:19:00 +00:00
U32 longLengthID; /* 0 == no longLength; 1 == Lit.longLength; 2 == Match.longLength; */
U32 longLengthPos;
} seqStore_t;
const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx); /* compress & dictBuilder */
void ZSTD_seqToCodes(const seqStore_t* seqStorePtr); /* compress, dictBuilder, decodeCorpus (shouldn't get its definition from here) */
/* custom memory allocation functions */
void* ZSTD_malloc(size_t size, ZSTD_customMem customMem);
void* ZSTD_calloc(size_t size, ZSTD_customMem customMem);
void ZSTD_free(void* ptr, ZSTD_customMem customMem);
MEM_STATIC U32 ZSTD_highbit32(U32 val) /* compress, dictBuilder, decodeCorpus */
{
2017-08-22 18:55:42 +00:00
assert(val != 0);
{
# if defined(_MSC_VER) /* Visual */
2017-08-22 18:55:42 +00:00
unsigned long r=0;
_BitScanReverse(&r, val);
return (unsigned)r;
# elif defined(__GNUC__) && (__GNUC__ >= 3) /* GCC Intrinsic */
2017-08-22 18:55:42 +00:00
return 31 - __builtin_clz(val);
# else /* Software version */
zstd_opt: changed cost formula There was a flaw in the formula which compared literal cost with match cost : at a given position, a non-null literal suite is going to be part of next sequence, while if position ends a previous match, to immediately start another match, next sequence will have a litlength of zero. A litlength of zero has a non-null cost. It follows that literals cost should be compared to match cost + litlength==0. Not doing so gave a structural advantage to matches, which would be selected more often. I believe that's what led to the creation of the strange heuristic which added a complex cost to matches. The heuristic was actually compensating. It was probably created through multiple trials, settling for best outcome on a given scenario (I suspect silesia.tar). The problem with this heuristic is that it's hard to understand, and unfortunately, any future change in the parser would impact the way it should be calculated and its effects. The "proper" formula makes it possible to remove this heuristic. Now, the problem is : in a head to head comparison, it's sometimes better, sometimes worse. Note that all differences are small (< 0.01 ratio). In general, the newer formula is better for smaller files (for example, calgary.tar and enwik7). I suspect that's because starting statistics are pretty poor (another area of improvement). However, for silesia.tar specifically, it's worse at level 22 (while being better at level 17, so even compression level has an impact ...). It's a pity that zstd -22 gets worse on silesia.tar. That being said, I like that the new code gets rid of strange variables, which were introducing complexity for any future evolution (faster variants being in mind). Therefore, in spite of this detrimental side effect, I tend to be in favor of it.
2017-11-28 22:07:03 +00:00
static const U32 DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
2017-08-22 18:55:42 +00:00
U32 v = val;
v |= v >> 1;
v |= v >> 2;
v |= v >> 4;
v |= v >> 8;
v |= v >> 16;
zstd_opt: changed cost formula There was a flaw in the formula which compared literal cost with match cost : at a given position, a non-null literal suite is going to be part of next sequence, while if position ends a previous match, to immediately start another match, next sequence will have a litlength of zero. A litlength of zero has a non-null cost. It follows that literals cost should be compared to match cost + litlength==0. Not doing so gave a structural advantage to matches, which would be selected more often. I believe that's what led to the creation of the strange heuristic which added a complex cost to matches. The heuristic was actually compensating. It was probably created through multiple trials, settling for best outcome on a given scenario (I suspect silesia.tar). The problem with this heuristic is that it's hard to understand, and unfortunately, any future change in the parser would impact the way it should be calculated and its effects. The "proper" formula makes it possible to remove this heuristic. Now, the problem is : in a head to head comparison, it's sometimes better, sometimes worse. Note that all differences are small (< 0.01 ratio). In general, the newer formula is better for smaller files (for example, calgary.tar and enwik7). I suspect that's because starting statistics are pretty poor (another area of improvement). However, for silesia.tar specifically, it's worse at level 22 (while being better at level 17, so even compression level has an impact ...). It's a pity that zstd -22 gets worse on silesia.tar. That being said, I like that the new code gets rid of strange variables, which were introducing complexity for any future evolution (faster variants being in mind). Therefore, in spite of this detrimental side effect, I tend to be in favor of it.
2017-11-28 22:07:03 +00:00
return DeBruijnClz[(v * 0x07C4ACDDU) >> 27];
# endif
2017-08-22 18:55:42 +00:00
}
}
/* ZSTD_invalidateRepCodes() :
* ensures next compression will not use repcodes from previous block.
* Note : only works with regular variant;
* do not use with extDict variant ! */
void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx); /* zstdmt, adaptive_compression (shouldn't get this definition from here) */
typedef struct {
blockType_e blockType;
U32 lastBlock;
U32 origSize;
} blockProperties_t;
/*! ZSTD_getcBlockSize() :
* Provides the size of compressed block from block header `src` */
/* Used by: decompress, fullbench (does not get its definition from here) */
size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
blockProperties_t* bpPtr);
2017-08-25 01:09:50 +00:00
#if defined (__cplusplus)
}
#endif
#endif /* ZSTD_CCOMMON_H_MODULE */