hertog/godot-module-template
Template
2
1
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update zstd to 1.3.4

Browse source
This commit is contained in:
volzhs 2018-05-16 02:45:22 +09:00
parent 3b8bd50b41
commit 5c5918a52d
32 changed files with 5648 additions and 3257 deletions
Download patch file Download diff file Expand all files Collapse all files

30
thirdparty/zstd/compress/fse_compress.c vendored
View file

@ -248,7 +248,7 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
bitCount -= (count<max);
previous0 = (count==1);
if (remaining<1) return ERROR(GENERIC);
while (remaining<threshold) nbBits--, threshold>>=1;
while (remaining<threshold) { nbBits--; threshold>>=1; }
}
if (bitCount>16) {
if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall); /* Buffer overflow */
@ -292,7 +292,7 @@ size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalized
It doesn't use any additional memory.
But this function is unsafe : it doesn't check that all values within `src` can fit into `count`.
For this reason, prefer using a table `count` with 256 elements.
@return : count of most numerous element
@return : count of most numerous element.
*/
size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
const void* src, size_t srcSize)
@ -305,7 +305,10 @@ size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
memset(count, 0, (maxSymbolValue+1)*sizeof(*count));
if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
while (ip<end) count[*ip++]++;
while (ip<end) {
assert(*ip <= maxSymbolValue);
count[*ip++]++;
}
while (!count[maxSymbolValue]) maxSymbolValue--;
*maxSymbolValuePtr = maxSymbolValue;
@ -318,7 +321,8 @@ size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
/* FSE_count_parallel_wksp() :
* Same as FSE_count_parallel(), but using an externally provided scratch buffer.
* `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`` */
* `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`.
* @return : largest histogram frequency, or an error code (notably when histogram would be larger than *maxSymbolValuePtr). */
static size_t FSE_count_parallel_wksp(
unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize,
@ -333,7 +337,7 @@ static size_t FSE_count_parallel_wksp(
U32* const Counting3 = Counting2 + 256;
U32* const Counting4 = Counting3 + 256;
memset(Counting1, 0, 4*256*sizeof(unsigned));
memset(workSpace, 0, 4*256*sizeof(unsigned));
/* safety checks */
if (!sourceSize) {
@ -379,7 +383,9 @@ static size_t FSE_count_parallel_wksp(
if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
} }
{ U32 s; for (s=0; s<=maxSymbolValue; s++) {
{ U32 s;
if (maxSymbolValue > 255) maxSymbolValue = 255;
for (s=0; s<=maxSymbolValue; s++) {
count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
if (count[s] > max) max = count[s];
} }
@ -393,9 +399,11 @@ static size_t FSE_count_parallel_wksp(
* Same as FSE_countFast(), but using an externally provided scratch buffer.
* `workSpace` size must be table of >= `1024` unsigned */
size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
const void* source, size_t sourceSize, unsigned* workSpace)
const void* source, size_t sourceSize,
unsigned* workSpace)
{
if (sourceSize < 1500) return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
if (sourceSize < 1500) /* heuristic threshold */
return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 0, workSpace);
}
@ -540,7 +548,7 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
find max, then give all remaining points to max */
U32 maxV = 0, maxC = 0;
for (s=0; s<=maxSymbolValue; s++)
if (count[s] > maxC) maxV=s, maxC=count[s];
if (count[s] > maxC) { maxV=s; maxC=count[s]; }
norm[maxV] += (short)ToDistribute;
return 0;
}
@ -548,7 +556,7 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
if (total == 0) {
/* all of the symbols were low enough for the lowOne or lowThreshold */
for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
if (norm[s] > 0) ToDistribute--, norm[s]++;
if (norm[s] > 0) { ToDistribute--; norm[s]++; }
return 0;
}
@ -604,7 +612,7 @@ size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
U64 restToBeat = vStep * rtbTable[proba];
proba += (count[s]*step) - ((U64)proba<<scale) > restToBeat;
}
if (proba > largestP) largestP=proba, largest=s;
if (proba > largestP) { largestP=proba; largest=s; }
normalizedCounter[s] = proba;
stillToDistribute -= proba;
} }

222
thirdparty/zstd/compress/huf_compress.c vendored
View file

@ -46,6 +46,7 @@
#include <string.h> /* memcpy, memset */
#include <stdio.h> /* printf (debug) */
#include "bitstream.h"
#include "compiler.h"
#define FSE_STATIC_LINKING_ONLY /* FSE_optimalTableLog_internal */
#include "fse.h" /* header compression */
#define HUF_STATIC_LINKING_ONLY
@ -322,7 +323,10 @@ static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
U32 const c = count[n];
U32 const r = BIT_highbit32(c+1) + 1;
U32 pos = rank[r].current++;
while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) huffNode[pos]=huffNode[pos-1], pos--;
while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) {
huffNode[pos] = huffNode[pos-1];
pos--;
}
huffNode[pos].count = c;
huffNode[pos].byte = (BYTE)n;
}
@ -331,10 +335,10 @@ static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
/** HUF_buildCTable_wksp() :
* Same as HUF_buildCTable(), but using externally allocated scratch buffer.
* `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of 1024 unsigned.
* `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of HUF_CTABLE_WORKSPACE_SIZE_U32 unsigned.
*/
#define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
typedef nodeElt huffNodeTable[2*HUF_SYMBOLVALUE_MAX+1 +1];
typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
{
nodeElt* const huffNode0 = (nodeElt*)workSpace;
@ -345,9 +349,10 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValu
U32 nodeRoot;
/* safety checks */
if (wkspSize < sizeof(huffNodeTable)) return ERROR(GENERIC); /* workSpace is not large enough */
if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (wkspSize < sizeof(huffNodeTable)) return ERROR(workSpace_tooSmall);
if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(GENERIC);
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
memset(huffNode0, 0, sizeof(huffNodeTable));
/* sort, decreasing order */
@ -405,6 +410,7 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValu
}
/** HUF_buildCTable() :
* @return : maxNbBits
* Note : count is used before tree is written, so they can safely overlap
*/
size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits)
@ -432,13 +438,14 @@ static int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, uns
return !bad;
}
static void HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
FORCE_INLINE_TEMPLATE void
HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
{
BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
}
size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
#define HUF_FLUSHBITS(s) BIT_flushBits(s)
#define HUF_FLUSHBITS_1(stream) \
@ -447,7 +454,10 @@ size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
#define HUF_FLUSHBITS_2(stream) \
if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*4+7) HUF_FLUSHBITS(stream)
size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
FORCE_INLINE_TEMPLATE size_t
HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
const BYTE* ip = (const BYTE*) src;
BYTE* const ostart = (BYTE*)dst;
@ -491,8 +501,58 @@ size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, si
return BIT_closeCStream(&bitC);
}
#if DYNAMIC_BMI2
size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
static TARGET_ATTRIBUTE("bmi2") size_t
HUF_compress1X_usingCTable_internal_bmi2(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
static size_t
HUF_compress1X_usingCTable_internal_default(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
static size_t
HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, const int bmi2)
{
if (bmi2) {
return HUF_compress1X_usingCTable_internal_bmi2(dst, dstSize, src, srcSize, CTable);
}
return HUF_compress1X_usingCTable_internal_default(dst, dstSize, src, srcSize, CTable);
}
#else
static size_t
HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, const int bmi2)
{
(void)bmi2;
return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
}
#endif
size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
{
return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
}
static size_t
HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
const void* src, size_t srcSize,
const HUF_CElt* CTable, int bmi2)
{
size_t const segmentSize = (srcSize+3)/4; /* first 3 segments */
const BYTE* ip = (const BYTE*) src;
@ -505,28 +565,31 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
if (srcSize < 12) return 0; /* no saving possible : too small input */
op += 6; /* jumpTable */
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart+2, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
if (cSize==0) return 0;
assert(cSize <= 65535);
MEM_writeLE16(ostart+4, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, iend-ip, CTable) );
{ CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, iend-ip, CTable, bmi2) );
if (cSize==0) return 0;
op += cSize;
}
@ -534,15 +597,20 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
return op-ostart;
}
size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
{
return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
}
static size_t HUF_compressCTable_internal(
BYTE* const ostart, BYTE* op, BYTE* const oend,
const void* src, size_t srcSize,
unsigned singleStream, const HUF_CElt* CTable)
unsigned singleStream, const HUF_CElt* CTable, const int bmi2)
{
size_t const cSize = singleStream ?
HUF_compress1X_usingCTable(op, oend - op, src, srcSize, CTable) :
HUF_compress4X_usingCTable(op, oend - op, src, srcSize, CTable);
HUF_compress1X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2) :
HUF_compress4X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2);
if (HUF_isError(cSize)) { return cSize; }
if (cSize==0) { return 0; } /* uncompressible */
op += cSize;
@ -551,86 +619,98 @@ static size_t HUF_compressCTable_internal(
return op-ostart;
}
typedef struct {
U32 count[HUF_SYMBOLVALUE_MAX + 1];
HUF_CElt CTable[HUF_SYMBOLVALUE_MAX + 1];
huffNodeTable nodeTable;
} HUF_compress_tables_t;
/* `workSpace` must a table of at least 1024 unsigned */
/* HUF_compress_internal() :
* `workSpace` must a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
static size_t HUF_compress_internal (
void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
unsigned singleStream,
void* workSpace, size_t wkspSize,
HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat)
HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat,
const int bmi2)
{
HUF_compress_tables_t* const table = (HUF_compress_tables_t*)workSpace;
BYTE* const ostart = (BYTE*)dst;
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart;
U32* count;
size_t const countSize = sizeof(U32) * (HUF_SYMBOLVALUE_MAX + 1);
HUF_CElt* CTable;
size_t const CTableSize = sizeof(HUF_CElt) * (HUF_SYMBOLVALUE_MAX + 1);
/* checks & inits */
if (wkspSize < sizeof(huffNodeTable) + countSize + CTableSize) return ERROR(GENERIC);
if (!srcSize) return 0; /* Uncompressed (note : 1 means rle, so first byte must be correct) */
if (!dstSize) return 0; /* cannot fit within dst budget */
if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC); /* must be aligned on 4-bytes boundaries */
if (wkspSize < sizeof(*table)) return ERROR(workSpace_tooSmall);
if (!srcSize) return 0; /* Uncompressed */
if (!dstSize) return 0; /* cannot fit anything within dst budget */
if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong); /* current block size limit */
if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
count = (U32*)workSpace;
workSpace = (BYTE*)workSpace + countSize;
wkspSize -= countSize;
CTable = (HUF_CElt*)workSpace;
workSpace = (BYTE*)workSpace + CTableSize;
wkspSize -= CTableSize;
/* Heuristic : If we don't need to check the validity of the old table use the old table for small inputs */
/* Heuristic : If old table is valid, use it for small inputs */
if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
singleStream, oldHufTable, bmi2);
}
/* Scan input and build symbol stats */
{ CHECK_V_F(largest, FSE_count_wksp (count, &maxSymbolValue, (const BYTE*)src, srcSize, (U32*)workSpace) );
{ CHECK_V_F(largest, FSE_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, table->count) );
if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */
if (largest <= (srcSize >> 7)+1) return 0; /* Fast heuristic : not compressible enough */
if (largest <= (srcSize >> 7)+1) return 0; /* heuristic : probably not compressible enough */
}
/* Check validity of previous table */
if (repeat && *repeat == HUF_repeat_check && !HUF_validateCTable(oldHufTable, count, maxSymbolValue)) {
if ( repeat
&& *repeat == HUF_repeat_check
&& !HUF_validateCTable(oldHufTable, table->count, maxSymbolValue)) {
*repeat = HUF_repeat_none;
}
/* Heuristic : use existing table for small inputs */
if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
singleStream, oldHufTable, bmi2);
}
/* Build Huffman Tree */
huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
{ CHECK_V_F(maxBits, HUF_buildCTable_wksp (CTable, count, maxSymbolValue, huffLog, workSpace, wkspSize) );
{ CHECK_V_F(maxBits, HUF_buildCTable_wksp(table->CTable, table->count,
maxSymbolValue, huffLog,
table->nodeTable, sizeof(table->nodeTable)) );
huffLog = (U32)maxBits;
/* Zero the unused symbols so we can check it for validity */
memset(CTable + maxSymbolValue + 1, 0, CTableSize - (maxSymbolValue + 1) * sizeof(HUF_CElt));
/* Zero unused symbols in CTable, so we can check it for validity */
memset(table->CTable + (maxSymbolValue + 1), 0,
sizeof(table->CTable) - ((maxSymbolValue + 1) * sizeof(HUF_CElt)));
}
/* Write table description header */
{ CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, CTable, maxSymbolValue, huffLog) );
/* Check if using the previous table will be beneficial */
{ CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, table->CTable, maxSymbolValue, huffLog) );
/* Check if using previous huffman table is beneficial */
if (repeat && *repeat != HUF_repeat_none) {
size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, count, maxSymbolValue);
size_t const newSize = HUF_estimateCompressedSize(CTable, count, maxSymbolValue);
size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, table->count, maxSymbolValue);
size_t const newSize = HUF_estimateCompressedSize(table->CTable, table->count, maxSymbolValue);
if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
}
}
/* Use the new table */
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
singleStream, oldHufTable, bmi2);
} }
/* Use the new huffman table */
if (hSize + 12ul >= srcSize) { return 0; }
op += hSize;
if (repeat) { *repeat = HUF_repeat_none; }
if (oldHufTable) { memcpy(oldHufTable, CTable, CTableSize); } /* Save the new table */
if (oldHufTable)
memcpy(oldHufTable, table->CTable, sizeof(table->CTable)); /* Save new table */
}
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, CTable);
return HUF_compressCTable_internal(ostart, op, oend,
src, srcSize,
singleStream, table->CTable, bmi2);
}
@ -639,52 +719,70 @@ size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, NULL, NULL, 0);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, 1 /*single stream*/,
workSpace, wkspSize,
NULL, NULL, 0, 0 /*bmi2*/);
}
size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, 1 /*single stream*/,
workSpace, wkspSize, hufTable,
repeat, preferRepeat, bmi2);
}
size_t HUF_compress1X (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog)
{
unsigned workSpace[1024];
unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
}
/* HUF_compress4X_repeat():
* compress input using 4 streams.
* provide workspace to generate compression tables */
size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, NULL, NULL, 0);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, 0 /*4 streams*/,
workSpace, wkspSize,
NULL, NULL, 0, 0 /*bmi2*/);
}
/* HUF_compress4X_repeat():
* compress input using 4 streams.
* re-use an existing huffman compression table */
size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, 0 /* 4 streams */,
workSpace, wkspSize,
hufTable, repeat, preferRepeat, bmi2);
}
size_t HUF_compress2 (void* dst, size_t dstSize,
const void* src, size_t srcSize,
unsigned maxSymbolValue, unsigned huffLog)
{
unsigned workSpace[1024];
unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
}
size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
{
return HUF_compress2(dst, maxDstSize, src, (U32)srcSize, 255, HUF_TABLELOG_DEFAULT);
return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT);
}

1615
thirdparty/zstd/compress/zstd_compress.c vendored
View file

File diff suppressed because it is too large Load diff

315
thirdparty/zstd/compress/zstd_compress_internal.h vendored
View file

@ -30,8 +30,14 @@ extern "C" {
/*-*************************************
* Constants
***************************************/
static const U32 g_searchStrength = 8;
#define HASH_READ_SIZE 8
#define kSearchStrength 8
#define HASH_READ_SIZE 8
#define ZSTD_DUBT_UNSORTED_MARK 1 /* For btlazy2 strategy, index 1 now means "unsorted".
It could be confused for a real successor at index "1", if sorted as larger than its predecessor.
It's not a big deal though : candidate will just be sorted again.
Additionnally, candidate position 1 will be lost.
But candidate 1 cannot hide a large tree of candidates, so it's a minimal loss.
The benefit is that ZSTD_DUBT_UNSORTED_MARK cannot be misdhandled after table re-use with a different strategy */
/*-*************************************
@ -43,7 +49,7 @@ typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
typedef struct ZSTD_prefixDict_s {
const void* dict;
size_t dictSize;
ZSTD_dictMode_e dictMode;
ZSTD_dictContentType_e dictContentType;
} ZSTD_prefixDict;
typedef struct {
@ -51,7 +57,6 @@ typedef struct {
FSE_CTable offcodeCTable[FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
FSE_CTable litlengthCTable[FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
U32 workspace[HUF_WORKSPACE_SIZE_U32];
HUF_repeat hufCTable_repeatMode;
FSE_repeat offcode_repeatMode;
FSE_repeat matchlength_repeatMode;
@ -93,12 +98,44 @@ typedef struct {
U32 staticPrices; /* prices follow a pre-defined cost structure, statistics are irrelevant */
} optState_t;
typedef struct {
ZSTD_entropyCTables_t entropy;
U32 rep[ZSTD_REP_NUM];
} ZSTD_compressedBlockState_t;
typedef struct {
BYTE const* nextSrc; /* next block here to continue on current prefix */
BYTE const* base; /* All regular indexes relative to this position */
BYTE const* dictBase; /* extDict indexes relative to this position */
U32 dictLimit; /* below that point, need extDict */
U32 lowLimit; /* below that point, no more data */
} ZSTD_window_t;
typedef struct {
ZSTD_window_t window; /* State for window round buffer management */
U32 loadedDictEnd; /* index of end of dictionary */
U32 nextToUpdate; /* index from which to continue table update */
U32 nextToUpdate3; /* index from which to continue table update */
U32 hashLog3; /* dispatch table : larger == faster, more memory */
U32* hashTable;
U32* hashTable3;
U32* chainTable;
optState_t opt; /* optimal parser state */
} ZSTD_matchState_t;
typedef struct {
ZSTD_compressedBlockState_t* prevCBlock;
ZSTD_compressedBlockState_t* nextCBlock;
ZSTD_matchState_t matchState;
} ZSTD_blockState_t;
typedef struct {
U32 offset;
U32 checksum;
} ldmEntry_t;
typedef struct {
ZSTD_window_t window; /* State for the window round buffer management */
ldmEntry_t* hashTable;
BYTE* bucketOffsets; /* Next position in bucket to insert entry */
U64 hashPower; /* Used to compute the rolling hash.
@ -111,60 +148,68 @@ typedef struct {
U32 bucketSizeLog; /* Log bucket size for collision resolution, at most 8 */
U32 minMatchLength; /* Minimum match length */
U32 hashEveryLog; /* Log number of entries to skip */
U32 windowLog; /* Window log for the LDM */
} ldmParams_t;
typedef struct {
U32 offset;
U32 litLength;
U32 matchLength;
} rawSeq;
typedef struct {
rawSeq* seq; /* The start of the sequences */
size_t pos; /* The position where reading stopped. <= size. */
size_t size; /* The number of sequences. <= capacity. */
size_t capacity; /* The capacity of the `seq` pointer */
} rawSeqStore_t;
struct ZSTD_CCtx_params_s {
ZSTD_format_e format;
ZSTD_compressionParameters cParams;
ZSTD_frameParameters fParams;
int compressionLevel;
U32 forceWindow; /* force back-references to respect limit of
int disableLiteralCompression;
int forceWindow; /* force back-references to respect limit of
* 1<<wLog, even for dictionary */
/* Multithreading: used to pass parameters to mtctx */
U32 nbThreads;
unsigned nbWorkers;
unsigned jobSize;
unsigned overlapSizeLog;
/* Long distance matching parameters */
ldmParams_t ldmParams;
/* For use with createCCtxParams() and freeCCtxParams() only */
/* Internal use, for createCCtxParams() and freeCCtxParams() only */
ZSTD_customMem customMem;
}; /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
struct ZSTD_CCtx_s {
const BYTE* nextSrc; /* next block here to continue on current prefix */
const BYTE* base; /* All regular indexes relative to this position */
const BYTE* dictBase; /* extDict indexes relative to this position */
U32 dictLimit; /* below that point, need extDict */
U32 lowLimit; /* below that point, no more data */
U32 nextToUpdate; /* index from which to continue dictionary update */
U32 nextToUpdate3; /* index from which to continue dictionary update */
U32 hashLog3; /* dispatch table : larger == faster, more memory */
U32 loadedDictEnd; /* index of end of dictionary */
ZSTD_compressionStage_e stage;
U32 dictID;
int cParamsChanged; /* == 1 if cParams(except wlog) or compression level are changed in requestedParams. Triggers transmission of new params to ZSTDMT (if available) then reset to 0. */
int bmi2; /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
ZSTD_CCtx_params requestedParams;
ZSTD_CCtx_params appliedParams;
U32 dictID;
void* workSpace;
size_t workSpaceSize;
size_t blockSize;
U64 pledgedSrcSizePlusOne; /* this way, 0 (default) == unknown */
U64 consumedSrcSize;
unsigned long long pledgedSrcSizePlusOne; /* this way, 0 (default) == unknown */
unsigned long long consumedSrcSize;
unsigned long long producedCSize;
XXH64_state_t xxhState;
ZSTD_customMem customMem;
size_t staticSize;
seqStore_t seqStore; /* sequences storage ptrs */
optState_t optState;
ldmState_t ldmState; /* long distance matching state */
U32* hashTable;
U32* hashTable3;
U32* chainTable;
ZSTD_entropyCTables_t* entropy;
seqStore_t seqStore; /* sequences storage ptrs */
ldmState_t ldmState; /* long distance matching state */
rawSeq* ldmSequences; /* Storage for the ldm output sequences */
size_t maxNbLdmSequences;
rawSeqStore_t externSeqStore; /* Mutable reference to external sequences */
ZSTD_blockState_t blockState;
U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */
/* streaming */
char* inBuff;
@ -191,6 +236,12 @@ struct ZSTD_CCtx_s {
};
typedef size_t (*ZSTD_blockCompressor) (
ZSTD_matchState_t* bs, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict);
MEM_STATIC U32 ZSTD_LLcode(U32 litLength)
{
static const BYTE LL_Code[64] = { 0, 1, 2, 3, 4, 5, 6, 7,
@ -359,10 +410,12 @@ MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* co
}
/** ZSTD_count_2segments() :
* can count match length with `ip` & `match` in 2 different segments.
* convention : on reaching mEnd, match count continue starting from iStart
*/
MEM_STATIC size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
* can count match length with `ip` & `match` in 2 different segments.
* convention : on reaching mEnd, match count continue starting from iStart
*/
MEM_STATIC size_t
ZSTD_count_2segments(const BYTE* ip, const BYTE* match,
const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
{
const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
size_t const matchLength = ZSTD_count(ip, match, vEnd);
@ -372,8 +425,8 @@ MEM_STATIC size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const
/*-*************************************
* Hashes
***************************************/
* Hashes
***************************************/
static const U32 prime3bytes = 506832829U;
static U32 ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes) >> (32-h) ; }
MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
@ -411,6 +464,171 @@ MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
}
}
/*-*************************************
* Round buffer management
***************************************/
/* Max current allowed */
#define ZSTD_CURRENT_MAX ((3U << 29) + (1U << ZSTD_WINDOWLOG_MAX))
/* Maximum chunk size before overflow correction needs to be called again */
#define ZSTD_CHUNKSIZE_MAX \
( ((U32)-1) /* Maximum ending current index */ \
- ZSTD_CURRENT_MAX) /* Maximum beginning lowLimit */
/**
* ZSTD_window_clear():
* Clears the window containing the history by simply setting it to empty.
*/
MEM_STATIC void ZSTD_window_clear(ZSTD_window_t* window)
{
size_t const endT = (size_t)(window->nextSrc - window->base);
U32 const end = (U32)endT;
window->lowLimit = end;
window->dictLimit = end;
}
/**
* ZSTD_window_hasExtDict():
* Returns non-zero if the window has a non-empty extDict.
*/
MEM_STATIC U32 ZSTD_window_hasExtDict(ZSTD_window_t const window)
{
return window.lowLimit < window.dictLimit;
}
/**
* ZSTD_window_needOverflowCorrection():
* Returns non-zero if the indices are getting too large and need overflow
* protection.
*/
MEM_STATIC U32 ZSTD_window_needOverflowCorrection(ZSTD_window_t const window,
void const* srcEnd)
{
U32 const current = (U32)((BYTE const*)srcEnd - window.base);
return current > ZSTD_CURRENT_MAX;
}
/**
* ZSTD_window_correctOverflow():
* Reduces the indices to protect from index overflow.
* Returns the correction made to the indices, which must be applied to every
* stored index.
*
* The least significant cycleLog bits of the indices must remain the same,
* which may be 0. Every index up to maxDist in the past must be valid.
* NOTE: (maxDist & cycleMask) must be zero.
*/
MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
U32 maxDist, void const* src)
{
/* preemptive overflow correction:
* 1. correction is large enough:
* lowLimit > (3<<29) ==> current > 3<<29 + 1<<windowLog
* 1<<windowLog <= newCurrent < 1<<chainLog + 1<<windowLog
*
* current - newCurrent
* > (3<<29 + 1<<windowLog) - (1<<windowLog + 1<<chainLog)
* > (3<<29) - (1<<chainLog)
* > (3<<29) - (1<<30) (NOTE: chainLog <= 30)
* > 1<<29
*
* 2. (ip+ZSTD_CHUNKSIZE_MAX - cctx->base) doesn't overflow:
* After correction, current is less than (1<<chainLog + 1<<windowLog).
* In 64-bit mode we are safe, because we have 64-bit ptrdiff_t.
* In 32-bit mode we are safe, because (chainLog <= 29), so
* ip+ZSTD_CHUNKSIZE_MAX - cctx->base < 1<<32.
* 3. (cctx->lowLimit + 1<<windowLog) < 1<<32:
* windowLog <= 31 ==> 3<<29 + 1<<windowLog < 7<<29 < 1<<32.
*/
U32 const cycleMask = (1U << cycleLog) - 1;
U32 const current = (U32)((BYTE const*)src - window->base);
U32 const newCurrent = (current & cycleMask) + maxDist;
U32 const correction = current - newCurrent;
assert((maxDist & cycleMask) == 0);
assert(current > newCurrent);
/* Loose bound, should be around 1<<29 (see above) */
assert(correction > 1<<28);
window->base += correction;
window->dictBase += correction;
window->lowLimit -= correction;
window->dictLimit -= correction;
DEBUGLOG(4, "Correction of 0x%x bytes to lowLimit=0x%x", correction,
window->lowLimit);
return correction;
}
/**
* ZSTD_window_enforceMaxDist():
* Updates lowLimit so that:
* (srcEnd - base) - lowLimit == maxDist + loadedDictEnd
* This allows a simple check that index >= lowLimit to see if index is valid.
* This must be called before a block compression call, with srcEnd as the block
* source end.
* If loadedDictEndPtr is not NULL, we set it to zero once we update lowLimit.
* This is because dictionaries are allowed to be referenced as long as the last
* byte of the dictionary is in the window, but once they are out of range,
* they cannot be referenced. If loadedDictEndPtr is NULL, we use
* loadedDictEnd == 0.
*/
MEM_STATIC void ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
void const* srcEnd, U32 maxDist,
U32* loadedDictEndPtr)
{
U32 const current = (U32)((BYTE const*)srcEnd - window->base);
U32 loadedDictEnd = loadedDictEndPtr != NULL ? *loadedDictEndPtr : 0;
if (current > maxDist + loadedDictEnd) {
U32 const newLowLimit = current - maxDist;
if (window->lowLimit < newLowLimit) window->lowLimit = newLowLimit;
if (window->dictLimit < window->lowLimit) {
DEBUGLOG(5, "Update dictLimit from %u to %u", window->dictLimit,
window->lowLimit);
window->dictLimit = window->lowLimit;
}
if (loadedDictEndPtr)
*loadedDictEndPtr = 0;
}
}
/**
* ZSTD_window_update():
* Updates the window by appending [src, src + srcSize) to the window.
* If it is not contiguous, the current prefix becomes the extDict, and we
* forget about the extDict. Handles overlap of the prefix and extDict.
* Returns non-zero if the segment is contiguous.
*/
MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
void const* src, size_t srcSize)
{
BYTE const* const ip = (BYTE const*)src;
U32 contiguous = 1;
/* Check if blocks follow each other */
if (src != window->nextSrc) {
/* not contiguous */
size_t const distanceFromBase = (size_t)(window->nextSrc - window->base);
DEBUGLOG(5, "Non contiguous blocks, new segment starts at %u",
window->dictLimit);
window->lowLimit = window->dictLimit;
assert(distanceFromBase == (size_t)(U32)distanceFromBase); /* should never overflow */
window->dictLimit = (U32)distanceFromBase;
window->dictBase = window->base;
window->base = ip - distanceFromBase;
// ms->nextToUpdate = window->dictLimit;
if (window->dictLimit - window->lowLimit < HASH_READ_SIZE) window->lowLimit = window->dictLimit; /* too small extDict */
contiguous = 0;
}
window->nextSrc = ip + srcSize;
/* if input and dictionary overlap : reduce dictionary (area presumed modified by input) */
if ( (ip+srcSize > window->dictBase + window->lowLimit)
& (ip < window->dictBase + window->dictLimit)) {
ptrdiff_t const highInputIdx = (ip + srcSize) - window->dictBase;
U32 const lowLimitMax = (highInputIdx > (ptrdiff_t)window->dictLimit) ? window->dictLimit : (U32)highInputIdx;
window->lowLimit = lowLimitMax;
}
return contiguous;
}
#if defined (__cplusplus)
}
#endif
@ -421,6 +639,13 @@ MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
* These prototypes shall only be called from within lib/compress
* ============================================================== */
/* ZSTD_getCParamsFromCCtxParams() :
* cParams are built depending on compressionLevel, src size hints,
* LDM and manually set compression parameters.
*/
ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize);
/*! ZSTD_initCStream_internal() :
* Private use only. Init streaming operation.
* expects params to be valid.
@ -446,7 +671,7 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
* Private use only. To be called from zstdmt_compress.c. */
size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
const void* dict, size_t dictSize,
ZSTD_dictMode_e dictMode,
ZSTD_dictContentType_e dictContentType,
const ZSTD_CDict* cdict,
ZSTD_CCtx_params params,
unsigned long long pledgedSrcSize);
@ -459,4 +684,26 @@ size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
const void* dict,size_t dictSize,
ZSTD_CCtx_params params);
/* ZSTD_writeLastEmptyBlock() :
* output an empty Block with end-of-frame mark to complete a frame
* @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))
* or an error code if `dstCapcity` is too small (<ZSTD_blockHeaderSize)
*/
size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity);
/* ZSTD_referenceExternalSequences() :
* Must be called before starting a compression operation.
* seqs must parse a prefix of the source.
* This cannot be used when long range matching is enabled.
* Zstd will use these sequences, and pass the literals to a secondary block
* compressor.
* @return : An error code on failure.
* NOTE: seqs are not verified! Invalid sequences can cause out-of-bounds memory
* access and data corruption.
*/
size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq);
#endif /* ZSTD_COMPRESS_H */

140
thirdparty/zstd/compress/zstd_double_fast.c vendored
View file

@ -12,44 +12,58 @@
#include "zstd_double_fast.h"
void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls)
void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
ZSTD_compressionParameters const* cParams,
void const* end)
{
U32* const hashLarge = cctx->hashTable;
U32 const hBitsL = cctx->appliedParams.cParams.hashLog;
U32* const hashSmall = cctx->chainTable;
U32 const hBitsS = cctx->appliedParams.cParams.chainLog;
const BYTE* const base = cctx->base;
const BYTE* ip = base + cctx->nextToUpdate;
U32* const hashLarge = ms->hashTable;
U32 const hBitsL = cParams->hashLog;
U32 const mls = cParams->searchLength;
U32* const hashSmall = ms->chainTable;
U32 const hBitsS = cParams->chainLog;
const BYTE* const base = ms->window.base;
const BYTE* ip = base + ms->nextToUpdate;
const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
const size_t fastHashFillStep = 3;
const U32 fastHashFillStep = 3;
while(ip <= iend) {
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip - base);
hashLarge[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip - base);
ip += fastHashFillStep;
/* Always insert every fastHashFillStep position into the hash tables.
* Insert the other positions into the large hash table if their entry
* is empty.
*/
for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
U32 const current = (U32)(ip - base);
U32 i;
for (i = 0; i < fastHashFillStep; ++i) {
size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
if (i == 0)
hashSmall[smHash] = current + i;
if (i == 0 || hashLarge[lgHash] == 0)
hashLarge[lgHash] = current + i;
}
}
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
const void* src, size_t srcSize,
const U32 mls)
size_t ZSTD_compressBlock_doubleFast_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
U32 const mls /* template */)
{
U32* const hashLong = cctx->hashTable;
const U32 hBitsL = cctx->appliedParams.cParams.hashLog;
U32* const hashSmall = cctx->chainTable;
const U32 hBitsS = cctx->appliedParams.cParams.chainLog;
seqStore_t* seqStorePtr = &(cctx->seqStore);
const BYTE* const base = cctx->base;
U32* const hashLong = ms->hashTable;
const U32 hBitsL = cParams->hashLog;
U32* const hashSmall = ms->chainTable;
const U32 hBitsS = cParams->chainLog;
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = cctx->dictLimit;
const U32 lowestIndex = ms->window.dictLimit;
const BYTE* const lowest = base + lowestIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - HASH_READ_SIZE;
U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
U32 offset_1=rep[0], offset_2=rep[1];
U32 offsetSaved = 0;
/* init */
@ -76,7 +90,7 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
/* favor repcode */
mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
ip++;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
} else {
U32 offset;
if ( (matchIndexL > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip)) ) {
@ -99,14 +113,14 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
}
} else {
ip += ((ip-anchor) >> g_searchStrength) + 1;
ip += ((ip-anchor) >> kSearchStrength) + 1;
continue;
}
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
}
/* match found */
@ -129,61 +143,63 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
{ U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
ip += rLength;
anchor = ip;
continue; /* faster when present ... (?) */
} } }
/* save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
rep[0] = offset_1 ? offset_1 : offsetSaved;
rep[1] = offset_2 ? offset_2 : offsetSaved;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_doubleFast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
const U32 mls = ctx->appliedParams.cParams.searchLength;
const U32 mls = cParams->searchLength;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 4);
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 5);
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 6);
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 7);
return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 7);
}
}
static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
const void* src, size_t srcSize,
const U32 mls)
static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
U32 const mls /* template */)
{
U32* const hashLong = ctx->hashTable;
U32 const hBitsL = ctx->appliedParams.cParams.hashLog;
U32* const hashSmall = ctx->chainTable;
U32 const hBitsS = ctx->appliedParams.cParams.chainLog;
seqStore_t* seqStorePtr = &(ctx->seqStore);
const BYTE* const base = ctx->base;
const BYTE* const dictBase = ctx->dictBase;
U32* const hashLong = ms->hashTable;
U32 const hBitsL = cParams->hashLog;
U32* const hashSmall = ms->chainTable;
U32 const hBitsS = cParams->chainLog;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = ctx->lowLimit;
const U32 lowestIndex = ms->window.lowLimit;
const BYTE* const dictStart = dictBase + lowestIndex;
const U32 dictLimit = ctx->dictLimit;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const lowPrefixPtr = base + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
U32 offset_1=rep[0], offset_2=rep[1];
/* Search Loop */
while (ip < ilimit) { /* < instead of <=, because (ip+1) */
@ -209,7 +225,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
ip++;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
} else {
if ((matchLongIndex > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
const BYTE* matchEnd = matchLongIndex < dictLimit ? dictEnd : iend;
@ -220,7 +236,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} else if ((matchIndex > lowestIndex) && (MEM_read32(match) == MEM_read32(ip))) {
size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
@ -245,10 +261,10 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
}
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} else {
ip += ((ip-anchor) >> g_searchStrength) + 1;
ip += ((ip-anchor) >> kSearchStrength) + 1;
continue;
} }
@ -272,7 +288,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
ip += repLength2;
@ -283,27 +299,29 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
} } }
/* save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
rep[0] = offset_1;
rep[1] = offset_2;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
size_t ZSTD_compressBlock_doubleFast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
U32 const mls = ctx->appliedParams.cParams.searchLength;
U32 const mls = cParams->searchLength;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 4);
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 4);
case 5 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 5);
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 5);
case 6 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 6);
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 6);
case 7 :
return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 7);
return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 7);
}
}

15
thirdparty/zstd/compress/zstd_double_fast.h vendored
View file

@ -16,11 +16,18 @@ extern "C" {
#endif
#include "mem.h" /* U32 */
#include "zstd.h" /* ZSTD_CCtx, size_t */
#include "zstd_compress_internal.h" /* ZSTD_CCtx, size_t */
void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
ZSTD_compressionParameters const* cParams,
void const* end);
size_t ZSTD_compressBlock_doubleFast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_doubleFast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls);
size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
#if defined (__cplusplus)
}

158
thirdparty/zstd/compress/zstd_fast.c vendored
View file

@ -12,39 +12,48 @@
#include "zstd_fast.h"
void ZSTD_fillHashTable (ZSTD_CCtx* zc, const void* end, const U32 mls)
void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
ZSTD_compressionParameters const* cParams,
void const* end)
{
U32* const hashTable = zc->hashTable;
U32 const hBits = zc->appliedParams.cParams.hashLog;
const BYTE* const base = zc->base;
const BYTE* ip = base + zc->nextToUpdate;
U32* const hashTable = ms->hashTable;
U32 const hBits = cParams->hashLog;
U32 const mls = cParams->searchLength;
const BYTE* const base = ms->window.base;
const BYTE* ip = base + ms->nextToUpdate;
const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
const size_t fastHashFillStep = 3;
const U32 fastHashFillStep = 3;
while(ip <= iend) {
hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip - base);
ip += fastHashFillStep;
/* Always insert every fastHashFillStep position into the hash table.
* Insert the other positions if their hash entry is empty.
*/
for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
U32 const current = (U32)(ip - base);
U32 i;
for (i = 0; i < fastHashFillStep; ++i) {
size_t const hash = ZSTD_hashPtr(ip + i, hBits, mls);
if (i == 0 || hashTable[hash] == 0)
hashTable[hash] = current + i;
}
}
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
const void* src, size_t srcSize,
const U32 mls)
size_t ZSTD_compressBlock_fast_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const hlog, U32 const stepSize, U32 const mls)
{
U32* const hashTable = cctx->hashTable;
U32 const hBits = cctx->appliedParams.cParams.hashLog;
seqStore_t* seqStorePtr = &(cctx->seqStore);
const BYTE* const base = cctx->base;
U32* const hashTable = ms->hashTable;
const BYTE* const base = ms->window.base;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = cctx->dictLimit;
const U32 lowestIndex = ms->window.dictLimit;
const BYTE* const lowest = base + lowestIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - HASH_READ_SIZE;
U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
U32 offset_1=rep[0], offset_2=rep[1];
U32 offsetSaved = 0;
/* init */
@ -57,7 +66,7 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
/* Main Search Loop */
while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
size_t mLength;
size_t const h = ZSTD_hashPtr(ip, hBits, mls);
size_t const h = ZSTD_hashPtr(ip, hlog, mls);
U32 const current = (U32)(ip-base);
U32 const matchIndex = hashTable[h];
const BYTE* match = base + matchIndex;
@ -66,21 +75,21 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
ip++;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
} else {
U32 offset;
if ( (matchIndex <= lowestIndex) || (MEM_read32(match) != MEM_read32(ip)) ) {
ip += ((ip-anchor) >> g_searchStrength) + 1;
if ( (matchIndex <= lowestIndex)
|| (MEM_read32(match) != MEM_read32(ip)) ) {
assert(stepSize >= 1);
ip += ((ip-anchor) >> kSearchStrength) + stepSize;
continue;
}
mLength = ZSTD_count(ip+4, match+4, iend) + 4;
offset = (U32)(ip-match);
while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
}
{ U32 const offset = (U32)(ip-match);
while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} }
/* match found */
ip += mLength;
@ -88,8 +97,8 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
if (ip <= ilimit) {
/* Fill Table */
hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2; /* here because current+2 could be > iend-8 */
hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2; /* here because current+2 could be > iend-8 */
hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
/* check immediate repcode */
while ( (ip <= ilimit)
&& ( (offset_2>0)
@ -97,65 +106,67 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
/* store sequence */
size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
{ U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip-base);
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
hashTable[ZSTD_hashPtr(ip, hlog, mls)] = (U32)(ip-base);
ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
ip += rLength;
anchor = ip;
continue; /* faster when present ... (?) */
} } }
/* save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
rep[0] = offset_1 ? offset_1 : offsetSaved;
rep[1] = offset_2 ? offset_2 : offsetSaved;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
size_t ZSTD_compressBlock_fast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
const U32 mls = ctx->appliedParams.cParams.searchLength;
U32 const hlog = cParams->hashLog;
U32 const mls = cParams->searchLength;
U32 const stepSize = cParams->targetLength;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 4);
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 4);
case 5 :
return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 5);
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 5);
case 6 :
return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 6);
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 6);
case 7 :
return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 7);
return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 7);
}
}
static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
const void* src, size_t srcSize,
const U32 mls)
static size_t ZSTD_compressBlock_fast_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
void const* src, size_t srcSize,
U32 const hlog, U32 const stepSize, U32 const mls)
{
U32* hashTable = ctx->hashTable;
const U32 hBits = ctx->appliedParams.cParams.hashLog;
seqStore_t* seqStorePtr = &(ctx->seqStore);
const BYTE* const base = ctx->base;
const BYTE* const dictBase = ctx->dictBase;
U32* hashTable = ms->hashTable;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = ctx->lowLimit;
const U32 lowestIndex = ms->window.lowLimit;
const BYTE* const dictStart = dictBase + lowestIndex;
const U32 dictLimit = ctx->dictLimit;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const lowPrefixPtr = base + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
U32 offset_1=rep[0], offset_2=rep[1];
/* Search Loop */
while (ip < ilimit) { /* < instead of <=, because (ip+1) */
const size_t h = ZSTD_hashPtr(ip, hBits, mls);
const size_t h = ZSTD_hashPtr(ip, hlog, mls);
const U32 matchIndex = hashTable[h];
const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
const BYTE* match = matchBase + matchIndex;
@ -171,11 +182,12 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
ip++;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
} else {
if ( (matchIndex < lowestIndex) ||
(MEM_read32(match) != MEM_read32(ip)) ) {
ip += ((ip-anchor) >> g_searchStrength) + 1;
assert(stepSize >= 1);
ip += ((ip-anchor) >> kSearchStrength) + stepSize;
continue;
}
{ const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
@ -186,7 +198,7 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
offset = current - matchIndex;
offset_2 = offset_1;
offset_1 = offset;
ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
} }
/* found a match : store it */
@ -195,8 +207,8 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
if (ip <= ilimit) {
/* Fill Table */
hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;
hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2;
hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
/* check immediate repcode */
while (ip <= ilimit) {
U32 const current2 = (U32)(ip-base);
@ -207,8 +219,8 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; /* swap offset_2 <=> offset_1 */
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
hashTable[ZSTD_hashPtr(ip, hBits, mls)] = current2;
ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
ip += repLength2;
anchor = ip;
continue;
@ -217,27 +229,31 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
} } }
/* save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
rep[0] = offset_1;
rep[1] = offset_2;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
size_t ZSTD_compressBlock_fast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
U32 const mls = ctx->appliedParams.cParams.searchLength;
U32 const hlog = cParams->hashLog;
U32 const mls = cParams->searchLength;
U32 const stepSize = cParams->targetLength;
switch(mls)
{
default: /* includes case 3 */
case 4 :
return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 4);
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 4);
case 5 :
return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 5);
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 5);
case 6 :
return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 6);
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 6);
case 7 :
return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 7);
return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 7);
}
}

16
thirdparty/zstd/compress/zstd_fast.h vendored
View file

@ -16,13 +16,17 @@ extern "C" {
#endif
#include "mem.h" /* U32 */
#include "zstd.h" /* ZSTD_CCtx, size_t */
#include "zstd_compress_internal.h"
void ZSTD_fillHashTable(ZSTD_CCtx* zc, const void* end, const U32 mls);
size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
const void* src, size_t srcSize);
size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
const void* src, size_t srcSize);
void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
ZSTD_compressionParameters const* cParams,
void const* end);
size_t ZSTD_compressBlock_fast(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_fast_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
#if defined (__cplusplus)
}

613
thirdparty/zstd/compress/zstd_lazy.c vendored
View file

@ -15,76 +15,90 @@
/*-*************************************
* Binary Tree search
***************************************/
/** ZSTD_insertBt1() : add one or multiple positions to tree.
* ip : assumed <= iend-8 .
* @return : nb of positions added */
static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
const BYTE* const ip, const BYTE* const iend,
U32 nbCompares, U32 const mls, U32 const extDict)
void ZSTD_updateDUBT(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* iend,
U32 mls)
{
U32* const hashTable = zc->hashTable;
U32 const hashLog = zc->appliedParams.cParams.hashLog;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32* const bt = zc->chainTable;
U32 const btLog = zc->appliedParams.cParams.chainLog - 1;
U32* const hashTable = ms->hashTable;
U32 const hashLog = cParams->hashLog;
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
const BYTE* const base = ms->window.base;
U32 const target = (U32)(ip - base);
U32 idx = ms->nextToUpdate;
if (idx != target)
DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
idx, target, ms->window.dictLimit);
assert(ip + 8 <= iend); /* condition for ZSTD_hashPtr */
(void)iend;
assert(idx >= ms->window.dictLimit); /* condition for valid base+idx */
for ( ; idx < target ; idx++) {
size_t const h = ZSTD_hashPtr(base + idx, hashLog, mls); /* assumption : ip + 8 <= iend */
U32 const matchIndex = hashTable[h];
U32* const nextCandidatePtr = bt + 2*(idx&btMask);
U32* const sortMarkPtr = nextCandidatePtr + 1;
DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
hashTable[h] = idx; /* Update Hash Table */
*nextCandidatePtr = matchIndex; /* update BT like a chain */
*sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
}
ms->nextToUpdate = target;
}
/** ZSTD_insertDUBT1() :
* sort one already inserted but unsorted position
* assumption : current >= btlow == (current - btmask)
* doesn't fail */
static void ZSTD_insertDUBT1(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
U32 current, const BYTE* inputEnd,
U32 nbCompares, U32 btLow, int extDict)
{
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
U32 matchIndex = hashTable[h];
size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const base = zc->base;
const BYTE* const dictBase = zc->dictBase;
const U32 dictLimit = zc->dictLimit;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const ip = (current>=dictLimit) ? base + current : dictBase + current;
const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* match;
const U32 current = (U32)(ip-base);
const U32 btLow = btMask >= current ? 0 : current - btMask;
U32* smallerPtr = bt + 2*(current&btMask);
U32* largerPtr = smallerPtr + 1;
U32 matchIndex = *smallerPtr;
U32 dummy32; /* to be nullified at the end */
U32 const windowLow = zc->lowLimit;
U32 matchEndIdx = current+8+1;
size_t bestLength = 8;
#ifdef ZSTD_C_PREDICT
U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
predictedSmall += (predictedSmall>0);
predictedLarge += (predictedLarge>0);
#endif /* ZSTD_C_PREDICT */
U32 const windowLow = ms->window.lowLimit;
DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
assert(ip <= iend-8); /* required for h calculation */
hashTable[h] = current; /* Update Hash Table */
DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
current, dictLimit, windowLow);
assert(current >= btLow);
assert(ip < iend); /* condition for ZSTD_count */
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
assert(matchIndex < current);
#ifdef ZSTD_C_PREDICT /* note : can create issues when hlog small <= 11 */
const U32* predictPtr = bt + 2*((matchIndex-1) & btMask); /* written this way, as bt is a roll buffer */
if (matchIndex == predictedSmall) {
/* no need to check length, result known */
*smallerPtr = matchIndex;
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
predictedSmall = predictPtr[1] + (predictPtr[1]>0);
continue;
}
if (matchIndex == predictedLarge) {
*largerPtr = matchIndex;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
predictedLarge = predictPtr[0] + (predictPtr[0]>0);
continue;
}
#endif
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
assert(matchIndex+matchLength >= dictLimit); /* might be wrong if extDict is incorrectly set to 0 */
match = base + matchIndex;
if ( (!extDict)
|| (matchIndex+matchLength >= dictLimit) /* both in current segment*/
|| (current < dictLimit) /* both in extDict */) {
const BYTE* const mBase = !extDict || ((matchIndex+matchLength) >= dictLimit) ? base : dictBase;
assert( (matchIndex+matchLength >= dictLimit) /* might be wrong if extDict is incorrectly set to 0 */
|| (current < dictLimit) );
match = mBase + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
@ -93,11 +107,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
}
if (matchLength > bestLength) {
bestLength = matchLength;
if (matchLength > matchEndIdx - matchIndex)
matchEndIdx = matchIndex + (U32)matchLength;
}
DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
current, matchIndex, (U32)matchLength);
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
@ -108,6 +119,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
matchIndex, btLow, nextPtr[1]);
smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
} else {
@ -115,184 +128,205 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
matchIndex, btLow, nextPtr[0]);
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
if (bestLength > 384) return MIN(192, (U32)(bestLength - 384)); /* speed optimization */
assert(matchEndIdx > current + 8);
return matchEndIdx - (current + 8);
}
FORCE_INLINE_TEMPLATE
void ZSTD_updateTree_internal(ZSTD_CCtx* zc,
const BYTE* const ip, const BYTE* const iend,
const U32 nbCompares, const U32 mls, const U32 extDict)
{
const BYTE* const base = zc->base;
U32 const target = (U32)(ip - base);
U32 idx = zc->nextToUpdate;
DEBUGLOG(7, "ZSTD_updateTree_internal, from %u to %u (extDict:%u)",
idx, target, extDict);
while(idx < target)
idx += ZSTD_insertBt1(zc, base+idx, iend, nbCompares, mls, extDict);
zc->nextToUpdate = target;
}
void ZSTD_updateTree(ZSTD_CCtx* zc,
const BYTE* const ip, const BYTE* const iend,
const U32 nbCompares, const U32 mls)
{
ZSTD_updateTree_internal(zc, ip, iend, nbCompares, mls, 0 /*extDict*/);
}
void ZSTD_updateTree_extDict(ZSTD_CCtx* zc,
const BYTE* const ip, const BYTE* const iend,
const U32 nbCompares, const U32 mls)
{
ZSTD_updateTree_internal(zc, ip, iend, nbCompares, mls, 1 /*extDict*/);
}
static size_t ZSTD_insertBtAndFindBestMatch (
ZSTD_CCtx* zc,
const BYTE* const ip, const BYTE* const iend,
size_t* offsetPtr,
U32 nbCompares, const U32 mls,
U32 extDict)
static size_t ZSTD_DUBT_findBestMatch (
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iend,
size_t* offsetPtr,
U32 const mls,
U32 const extDict)
{
U32* const hashTable = zc->hashTable;
U32 const hashLog = zc->appliedParams.cParams.hashLog;
U32* const hashTable = ms->hashTable;
U32 const hashLog = cParams->hashLog;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32* const bt = zc->chainTable;
U32 const btLog = zc->appliedParams.cParams.chainLog - 1;
U32 matchIndex = hashTable[h];
const BYTE* const base = ms->window.base;
U32 const current = (U32)(ip-base);
U32 const windowLow = ms->window.lowLimit;
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
U32 matchIndex = hashTable[h];
size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const base = zc->base;
const BYTE* const dictBase = zc->dictBase;
const U32 dictLimit = zc->dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const U32 current = (U32)(ip-base);
const U32 btLow = btMask >= current ? 0 : current - btMask;
const U32 windowLow = zc->lowLimit;
U32* smallerPtr = bt + 2*(current&btMask);
U32* largerPtr = bt + 2*(current&btMask) + 1;
U32 matchEndIdx = current+8+1;
U32 dummy32; /* to be nullified at the end */
size_t bestLength = 0;
U32 const btLow = (btMask >= current) ? 0 : current - btMask;
U32 const unsortLimit = MAX(btLow, windowLow);
U32* nextCandidate = bt + 2*(matchIndex&btMask);
U32* unsortedMark = bt + 2*(matchIndex&btMask) + 1;
U32 nbCompares = 1U << cParams->searchLog;
U32 nbCandidates = nbCompares;
U32 previousCandidate = 0;
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", current);
assert(ip <= iend-8); /* required for h calculation */
hashTable[h] = current; /* Update Hash Table */
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
const BYTE* match;
/* reach end of unsorted candidates list */
while ( (matchIndex > unsortLimit)
&& (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
&& (nbCandidates > 1) ) {
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
matchIndex);
*unsortedMark = previousCandidate;
previousCandidate = matchIndex;
matchIndex = *nextCandidate;
nextCandidate = bt + 2*(matchIndex&btMask);
unsortedMark = bt + 2*(matchIndex&btMask) + 1;
nbCandidates --;
}
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
match = base + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (matchIndex+matchLength >= dictLimit)
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
}
if ( (matchIndex > unsortLimit)
&& (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
matchIndex);
*nextCandidate = *unsortedMark = 0; /* nullify next candidate if it's still unsorted (note : simplification, detrimental to compression ratio, beneficial for speed) */
}
if (matchLength > bestLength) {
if (matchLength > matchEndIdx - matchIndex)
matchEndIdx = matchIndex + (U32)matchLength;
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
break; /* drop, to guarantee consistency (miss a little bit of compression) */
/* batch sort stacked candidates */
matchIndex = previousCandidate;
while (matchIndex) { /* will end on matchIndex == 0 */
U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
U32 const nextCandidateIdx = *nextCandidateIdxPtr;
ZSTD_insertDUBT1(ms, cParams, matchIndex, iend,
nbCandidates, unsortLimit, extDict);
matchIndex = nextCandidateIdx;
nbCandidates++;
}
/* find longest match */
{ size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
U32* smallerPtr = bt + 2*(current&btMask);
U32* largerPtr = bt + 2*(current&btMask) + 1;
U32 matchEndIdx = current+8+1;
U32 dummy32; /* to be nullified at the end */
size_t bestLength = 0;
matchIndex = hashTable[h];
hashTable[h] = current; /* Update Hash Table */
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
const BYTE* match;
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
match = base + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (matchIndex+matchLength >= dictLimit)
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
}
if (matchLength > bestLength) {
if (matchLength > matchEndIdx - matchIndex)
matchEndIdx = matchIndex + (U32)matchLength;
if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
break; /* drop, to guarantee consistency (miss a little bit of compression) */
}
}
if (match[matchLength] < ip[matchLength]) {
/* match is smaller than current */
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
} else {
/* match is larger than current */
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */
ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
if (bestLength >= MINMATCH) {
U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
current, (U32)bestLength, (U32)*offsetPtr, mIndex);
}
if (match[matchLength] < ip[matchLength]) {
/* match is smaller than current */
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
} else {
/* match is larger than current */
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
assert(matchEndIdx > current+8);
zc->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
return bestLength;
return bestLength;
}
}
/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
static size_t ZSTD_BtFindBestMatch (
ZSTD_CCtx* zc,
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 mls)
const U32 mls /* template */)
{
if (ip < zc->base + zc->nextToUpdate) return 0; /* skipped area */
ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 0);
DEBUGLOG(7, "ZSTD_BtFindBestMatch");
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls);
return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 0);
}
static size_t ZSTD_BtFindBestMatch_selectMLS (
ZSTD_CCtx* zc, /* Index table will be updated */
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 matchLengthSearch)
size_t* offsetPtr)
{
switch(matchLengthSearch)
switch(cParams->searchLength)
{
default : /* includes case 3 */
case 4 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
case 5 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4);
case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5);
case 7 :
case 6 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6);
}
}
/** Tree updater, providing best match */
static size_t ZSTD_BtFindBestMatch_extDict (
ZSTD_CCtx* zc,
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 mls)
const U32 mls)
{
if (ip < zc->base + zc->nextToUpdate) return 0; /* skipped area */
ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 1);
DEBUGLOG(7, "ZSTD_BtFindBestMatch_extDict");
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls);
return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 1);
}
static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
ZSTD_CCtx* zc, /* Index table will be updated */
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 matchLengthSearch)
size_t* offsetPtr)
{
switch(matchLengthSearch)
switch(cParams->searchLength)
{
default : /* includes case 3 */
case 4 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
case 5 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
case 4 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 4);
case 5 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 5);
case 7 :
case 6 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
case 6 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 6);
}
}
@ -305,15 +339,17 @@ static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
/* Update chains up to ip (excluded)
Assumption : always within prefix (i.e. not within extDict) */
U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
static U32 ZSTD_insertAndFindFirstIndex_internal(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, U32 const mls)
{
U32* const hashTable = zc->hashTable;
const U32 hashLog = zc->appliedParams.cParams.hashLog;
U32* const chainTable = zc->chainTable;
const U32 chainMask = (1 << zc->appliedParams.cParams.chainLog) - 1;
const BYTE* const base = zc->base;
U32* const hashTable = ms->hashTable;
const U32 hashLog = cParams->hashLog;
U32* const chainTable = ms->chainTable;
const U32 chainMask = (1 << cParams->chainLog) - 1;
const BYTE* const base = ms->window.base;
const U32 target = (U32)(ip - base);
U32 idx = zc->nextToUpdate;
U32 idx = ms->nextToUpdate;
while(idx < target) { /* catch up */
size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
@ -322,35 +358,42 @@ U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
idx++;
}
zc->nextToUpdate = target;
ms->nextToUpdate = target;
return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
}
U32 ZSTD_insertAndFindFirstIndex(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip)
{
return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, cParams->searchLength);
}
/* inlining is important to hardwire a hot branch (template emulation) */
FORCE_INLINE_TEMPLATE
size_t ZSTD_HcFindBestMatch_generic (
ZSTD_CCtx* zc, /* Index table will be updated */
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 mls, const U32 extDict)
const U32 mls, const U32 extDict)
{
U32* const chainTable = zc->chainTable;
const U32 chainSize = (1 << zc->appliedParams.cParams.chainLog);
U32* const chainTable = ms->chainTable;
const U32 chainSize = (1 << cParams->chainLog);
const U32 chainMask = chainSize-1;
const BYTE* const base = zc->base;
const BYTE* const dictBase = zc->dictBase;
const U32 dictLimit = zc->dictLimit;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const U32 lowLimit = zc->lowLimit;
const U32 lowLimit = ms->window.lowLimit;
const U32 current = (U32)(ip-base);
const U32 minChain = current > chainSize ? current - chainSize : 0;
int nbAttempts=maxNbAttempts;
U32 nbAttempts = 1U << cParams->searchLog;
size_t ml=4-1;
/* HC4 match finder */
U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
U32 matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
size_t currentMl=0;
@ -381,35 +424,33 @@ size_t ZSTD_HcFindBestMatch_generic (
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
ZSTD_CCtx* zc,
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* const iLimit,
size_t* offsetPtr,
const U32 maxNbAttempts, const U32 matchLengthSearch)
size_t* offsetPtr)
{
switch(matchLengthSearch)
switch(cParams->searchLength)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 0);
case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 0);
case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 0);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 0);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 0);
case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 0);
}
}
FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
ZSTD_CCtx* const zc,
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* const iLimit,
size_t* const offsetPtr,
U32 const maxNbAttempts, U32 const matchLengthSearch)
size_t* const offsetPtr)
{
switch(matchLengthSearch)
switch(cParams->searchLength)
{
default : /* includes case 3 */
case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 1);
case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 1);
case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 1);
case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 1);
case 7 :
case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 1);
case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 1);
}
}
@ -418,30 +459,29 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
* Common parser - lazy strategy
*********************************/
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
const void* src, size_t srcSize,
const U32 searchMethod, const U32 depth)
size_t ZSTD_compressBlock_lazy_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore,
U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams,
const void* src, size_t srcSize,
const U32 searchMethod, const U32 depth)
{
seqStore_t* seqStorePtr = &(ctx->seqStore);
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
const BYTE* const base = ctx->base + ctx->dictLimit;
const BYTE* const base = ms->window.base + ms->window.dictLimit;
U32 const maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
U32 const mls = ctx->appliedParams.cParams.searchLength;
typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
size_t* offsetPtr,
U32 maxNbAttempts, U32 matchLengthSearch);
typedef size_t (*searchMax_f)(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS;
U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1], savedOffset=0;
U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
/* init */
ip += (ip==base);
ctx->nextToUpdate3 = ctx->nextToUpdate;
ms->nextToUpdate3 = ms->nextToUpdate;
{ U32 const maxRep = (U32)(ip-base);
if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
@ -462,13 +502,13 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
/* first search (depth 0) */
{ size_t offsetFound = 99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound);
if (ml2 > matchLength)
matchLength = ml2, start = ip, offset=offsetFound;
}
if (matchLength < 4) {
ip += ((ip-anchor) >> g_searchStrength) + 1; /* jump faster over incompressible sections */
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
continue;
}
@ -484,7 +524,7 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
matchLength = mlRep, offset = 0, start = ip;
}
{ size_t offset2=99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
if ((ml2 >= 4) && (gain2 > gain1)) {
@ -503,7 +543,7 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
matchLength = ml2, offset = 0, start = ip;
}
{ size_t offset2=99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
if ((ml2 >= 4) && (gain2 > gain1)) {
@ -528,7 +568,7 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
/* store sequence */
_storeSequence:
{ size_t const litLength = start - anchor;
ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
anchor = ip = start + matchLength;
}
@ -538,73 +578,80 @@ _storeSequence:
/* store sequence */
matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
ip += matchLength;
anchor = ip;
continue; /* faster when present ... (?) */
} }
/* Save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : savedOffset;
seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : savedOffset;
rep[0] = offset_1 ? offset_1 : savedOffset;
rep[1] = offset_2 ? offset_2 : savedOffset;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btlazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 1, 2);
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2);
}
size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_lazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 2);
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2);
}
size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_lazy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 1);
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1);
}
size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_greedy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 0);
return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0);
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
const void* src, size_t srcSize,
const U32 searchMethod, const U32 depth)
size_t ZSTD_compressBlock_lazy_extDict_generic(
ZSTD_matchState_t* ms, seqStore_t* seqStore,
U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams,
const void* src, size_t srcSize,
const U32 searchMethod, const U32 depth)
{
seqStore_t* seqStorePtr = &(ctx->seqStore);
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
const BYTE* const base = ctx->base;
const U32 dictLimit = ctx->dictLimit;
const U32 lowestIndex = ctx->lowLimit;
const BYTE* const base = ms->window.base;
const U32 dictLimit = ms->window.dictLimit;
const U32 lowestIndex = ms->window.lowLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* const dictBase = ctx->dictBase;
const BYTE* const dictBase = ms->window.dictBase;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const dictStart = dictBase + ctx->lowLimit;
const BYTE* const dictStart = dictBase + lowestIndex;
const U32 maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
const U32 mls = ctx->appliedParams.cParams.searchLength;
typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
size_t* offsetPtr,
U32 maxNbAttempts, U32 matchLengthSearch);
typedef size_t (*searchMax_f)(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1];
U32 offset_1 = rep[0], offset_2 = rep[1];
/* init */
ctx->nextToUpdate3 = ctx->nextToUpdate;
ms->nextToUpdate3 = ms->nextToUpdate;
ip += (ip == prefixStart);
/* Match Loop */
@ -628,13 +675,13 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
/* first search (depth 0) */
{ size_t offsetFound = 99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound);
if (ml2 > matchLength)
matchLength = ml2, start = ip, offset=offsetFound;
}
if (matchLength < 4) {
ip += ((ip-anchor) >> g_searchStrength) + 1; /* jump faster over incompressible sections */
ip += ((ip-anchor) >> kSearchStrength) + 1; /* jump faster over incompressible sections */
continue;
}
@ -661,7 +708,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
/* search match, depth 1 */
{ size_t offset2=99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
if ((ml2 >= 4) && (gain2 > gain1)) {
@ -691,7 +738,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
/* search match, depth 2 */
{ size_t offset2=99999999;
size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1)); /* raw approx */
int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
if ((ml2 >= 4) && (gain2 > gain1)) {
@ -713,7 +760,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
/* store sequence */
_storeSequence:
{ size_t const litLength = start - anchor;
ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
anchor = ip = start + matchLength;
}
@ -728,7 +775,7 @@ _storeSequence:
const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap offset history */
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
ip += matchLength;
anchor = ip;
continue; /* faster when present ... (?) */
@ -737,29 +784,41 @@ _storeSequence:
} }
/* Save reps for next block */
seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
rep[0] = offset_1;
rep[1] = offset_2;
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_greedy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 0);
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0);
}
size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_lazy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 1);
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1);
}
size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_lazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 2);
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2);
}
size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btlazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 1, 2);
return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2);
}

43
thirdparty/zstd/compress/zstd_lazy.h vendored
View file

@ -15,22 +15,39 @@
extern "C" {
#endif
#include "mem.h" /* U32 */
#include "zstd.h" /* ZSTD_CCtx, size_t */
#include "zstd_compress_internal.h"
U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls);
void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
U32 ZSTD_insertAndFindFirstIndex(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip);
size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
void ZSTD_preserveUnsortedMark (U32* const table, U32 const size, U32 const reducerValue); /*! used in ZSTD_reduceIndex(). pre-emptively increase value of ZSTD_DUBT_UNSORTED_MARK */
size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_btlazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_greedy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_lazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btlazy2_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
#if defined (__cplusplus)
}

684
thirdparty/zstd/compress/zstd_ldm.c vendored
View file

@ -17,36 +17,45 @@
#define LDM_HASH_RLOG 7
#define LDM_HASH_CHAR_OFFSET 10
size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm)
void ZSTD_ldm_adjustParameters(ldmParams_t* params,
ZSTD_compressionParameters const* cParams)
{
U32 const windowLog = cParams->windowLog;
ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
params->enableLdm = enableLdm>0;
params->hashLog = 0;
params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
params->minMatchLength = LDM_MIN_MATCH_LENGTH;
params->hashEveryLog = ZSTD_LDM_HASHEVERYLOG_NOTSET;
return 0;
}
void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog)
{
DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
if (cParams->strategy >= ZSTD_btopt) {
/* Get out of the way of the optimal parser */
U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
params->minMatchLength = minMatch;
}
if (params->hashLog == 0) {
params->hashLog = MAX(ZSTD_HASHLOG_MIN, windowLog - LDM_HASH_RLOG);
assert(params->hashLog <= ZSTD_HASHLOG_MAX);
}
if (params->hashEveryLog == ZSTD_LDM_HASHEVERYLOG_NOTSET) {
if (params->hashEveryLog == 0) {
params->hashEveryLog =
windowLog < params->hashLog ? 0 : windowLog - params->hashLog;
}
params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
}
size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog) {
size_t const ldmHSize = ((size_t)1) << hashLog;
size_t const ldmBucketSizeLog = MIN(bucketSizeLog, hashLog);
size_t ZSTD_ldm_getTableSize(ldmParams_t params)
{
size_t const ldmHSize = ((size_t)1) << params.hashLog;
size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog);
size_t const ldmBucketSize =
((size_t)1) << (hashLog - ldmBucketSizeLog);
return ldmBucketSize + (ldmHSize * (sizeof(ldmEntry_t)));
((size_t)1) << (params.hashLog - ldmBucketSizeLog);
size_t const totalSize = ldmBucketSize + ldmHSize * sizeof(ldmEntry_t);
return params.enableLdm ? totalSize : 0;
}
size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize)
{
return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
}
/** ZSTD_ldm_getSmallHash() :
@ -167,6 +176,7 @@ static U64 ZSTD_ldm_ipow(U64 base, U64 exp)
}
U64 ZSTD_ldm_getHashPower(U32 minMatchLength) {
DEBUGLOG(4, "ZSTD_ldm_getHashPower: mml=%u", minMatchLength);
assert(minMatchLength >= ZSTD_LDM_MINMATCH_MIN);
return ZSTD_ldm_ipow(prime8bytes, minMatchLength - 1);
}
@ -205,21 +215,22 @@ static size_t ZSTD_ldm_countBackwardsMatch(
*
* The tables for the other strategies are filled within their
* block compressors. */
static size_t ZSTD_ldm_fillFastTables(ZSTD_CCtx* zc, const void* end)
static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t* ms,
ZSTD_compressionParameters const* cParams,
void const* end)
{
const BYTE* const iend = (const BYTE*)end;
const U32 mls = zc->appliedParams.cParams.searchLength;
switch(zc->appliedParams.cParams.strategy)
switch(cParams->strategy)
{
case ZSTD_fast:
ZSTD_fillHashTable(zc, iend, mls);
zc->nextToUpdate = (U32)(iend - zc->base);
ZSTD_fillHashTable(ms, cParams, iend);
ms->nextToUpdate = (U32)(iend - ms->window.base);
break;
case ZSTD_dfast:
ZSTD_fillDoubleHashTable(zc, iend, mls);
zc->nextToUpdate = (U32)(iend - zc->base);
ZSTD_fillDoubleHashTable(ms, cParams, iend);
ms->nextToUpdate = (U32)(iend - ms->window.base);
break;
case ZSTD_greedy:
@ -268,69 +279,62 @@ static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
* Sets cctx->nextToUpdate to a position corresponding closer to anchor
* if it is far way
* (after a long match, only update tables a limited amount). */
static void ZSTD_ldm_limitTableUpdate(ZSTD_CCtx* cctx, const BYTE* anchor)
static void ZSTD_ldm_limitTableUpdate(ZSTD_matchState_t* ms, const BYTE* anchor)
{
U32 const current = (U32)(anchor - cctx->base);
if (current > cctx->nextToUpdate + 1024) {
cctx->nextToUpdate =
current - MIN(512, current - cctx->nextToUpdate - 1024);
U32 const current = (U32)(anchor - ms->window.base);
if (current > ms->nextToUpdate + 1024) {
ms->nextToUpdate =
current - MIN(512, current - ms->nextToUpdate - 1024);
}
}
typedef size_t (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
/* defined in zstd_compress.c */
ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict);
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
const void* src, size_t srcSize)
static size_t ZSTD_ldm_generateSequences_internal(
ldmState_t* ldmState, rawSeqStore_t* rawSeqStore,
ldmParams_t const* params, void const* src, size_t srcSize)
{
ldmState_t* const ldmState = &(cctx->ldmState);
const ldmParams_t ldmParams = cctx->appliedParams.ldmParams;
const U64 hashPower = ldmState->hashPower;
const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
seqStore_t* const seqStorePtr = &(cctx->seqStore);
const BYTE* const base = cctx->base;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = cctx->dictLimit;
const BYTE* const lowest = base + lowestIndex;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
const ZSTD_blockCompressor blockCompressor =
ZSTD_selectBlockCompressor(cctx->appliedParams.cParams.strategy, 0);
U32* const repToConfirm = seqStorePtr->repToConfirm;
U32 savedRep[ZSTD_REP_NUM];
/* LDM parameters */
int const extDict = ZSTD_window_hasExtDict(ldmState->window);
U32 const minMatchLength = params->minMatchLength;
U64 const hashPower = ldmState->hashPower;
U32 const hBits = params->hashLog - params->bucketSizeLog;
U32 const ldmBucketSize = 1U << params->bucketSizeLog;
U32 const hashEveryLog = params->hashEveryLog;
U32 const ldmTagMask = (1U << params->hashEveryLog) - 1;
/* Prefix and extDict parameters */
U32 const dictLimit = ldmState->window.dictLimit;
U32 const lowestIndex = extDict ? ldmState->window.lowLimit : dictLimit;
BYTE const* const base = ldmState->window.base;
BYTE const* const dictBase = extDict ? ldmState->window.dictBase : NULL;
BYTE const* const dictStart = extDict ? dictBase + lowestIndex : NULL;
BYTE const* const dictEnd = extDict ? dictBase + dictLimit : NULL;
BYTE const* const lowPrefixPtr = base + dictLimit;
/* Input bounds */
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
BYTE const* const ilimit = iend - MAX(minMatchLength, HASH_READ_SIZE);
/* Input positions */
BYTE const* anchor = istart;
BYTE const* ip = istart;
/* Rolling hash */
BYTE const* lastHashed = NULL;
U64 rollingHash = 0;
const BYTE* lastHashed = NULL;
size_t i, lastLiterals;
/* Save seqStorePtr->rep and copy repToConfirm */
for (i = 0; i < ZSTD_REP_NUM; i++)
savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
/* Main Search Loop */
while (ip < ilimit) { /* < instead of <=, because repcode check at (ip+1) */
while (ip <= ilimit) {
size_t mLength;
U32 const current = (U32)(ip - base);
size_t forwardMatchLength = 0, backwardMatchLength = 0;
ldmEntry_t* bestEntry = NULL;
if (ip != istart) {
rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
lastHashed[ldmParams.minMatchLength],
lastHashed[minMatchLength],
hashPower);
} else {
rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
rollingHash = ZSTD_ldm_getRollingHash(ip, minMatchLength);
}
lastHashed = ip;
/* Do not insert and do not look for a match */
if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
ldmTagMask) {
if (ZSTD_ldm_getTag(rollingHash, hBits, hashEveryLog) != ldmTagMask) {
ip++;
continue;
}
@ -340,27 +344,49 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
ldmEntry_t* const bucket =
ZSTD_ldm_getBucket(ldmState,
ZSTD_ldm_getSmallHash(rollingHash, hBits),
ldmParams);
*params);
ldmEntry_t* cur;
size_t bestMatchLength = 0;
U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
const BYTE* const pMatch = cur->offset + base;
size_t curForwardMatchLength, curBackwardMatchLength,
curTotalMatchLength;
if (cur->checksum != checksum || cur->offset <= lowestIndex) {
continue;
}
if (extDict) {
BYTE const* const curMatchBase =
cur->offset < dictLimit ? dictBase : base;
BYTE const* const pMatch = curMatchBase + cur->offset;
BYTE const* const matchEnd =
cur->offset < dictLimit ? dictEnd : iend;
BYTE const* const lowMatchPtr =
cur->offset < dictLimit ? dictStart : lowPrefixPtr;
curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
if (curForwardMatchLength < ldmParams.minMatchLength) {
continue;
curForwardMatchLength = ZSTD_count_2segments(
ip, pMatch, iend,
matchEnd, lowPrefixPtr);
if (curForwardMatchLength < minMatchLength) {
continue;
}
curBackwardMatchLength =
ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
lowMatchPtr);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
} else { /* !extDict */
BYTE const* const pMatch = base + cur->offset;
curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
if (curForwardMatchLength < minMatchLength) {
continue;
}
curBackwardMatchLength =
ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
lowPrefixPtr);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
}
curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
ip, anchor, pMatch, lowest);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
if (curTotalMatchLength > bestMatchLength) {
bestMatchLength = curTotalMatchLength;
@ -375,7 +401,7 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
if (bestEntry == NULL) {
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash,
hBits, current,
ldmParams);
*params);
ip++;
continue;
}
@ -384,324 +410,244 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
mLength = forwardMatchLength + backwardMatchLength;
ip -= backwardMatchLength;
/* Call the block compressor on the remaining literals */
{
/* Store the sequence:
* ip = current - backwardMatchLength
* The match is at (bestEntry->offset - backwardMatchLength)
*/
U32 const matchIndex = bestEntry->offset;
const BYTE* const match = base + matchIndex - backwardMatchLength;
U32 const offset = (U32)(ip - match);
U32 const offset = current - matchIndex;
rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
/* Overwrite rep codes */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = repToConfirm[i];
/* Fill tables for block compressor */
ZSTD_ldm_limitTableUpdate(cctx, anchor);
ZSTD_ldm_fillFastTables(cctx, anchor);
/* Call block compressor and get remaining literals */
lastLiterals = blockCompressor(cctx, anchor, ip - anchor);
cctx->nextToUpdate = (U32)(ip - base);
/* Update repToConfirm with the new offset */
for (i = ZSTD_REP_NUM - 1; i > 0; i--)
repToConfirm[i] = repToConfirm[i-1];
repToConfirm[0] = offset;
/* Store the sequence with the leftover literals */
ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
offset + ZSTD_REP_MOVE, mLength - MINMATCH);
/* Out of sequence storage */
if (rawSeqStore->size == rawSeqStore->capacity)
return ERROR(dstSize_tooSmall);
seq->litLength = (U32)(ip - anchor);
seq->matchLength = (U32)mLength;
seq->offset = offset;
rawSeqStore->size++;
}
/* Insert the current entry into the hash table */
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
(U32)(lastHashed - base),
ldmParams);
*params);
assert(ip + backwardMatchLength == lastHashed);
/* Fill the hash table from lastHashed+1 to ip+mLength*/
/* Heuristic: don't need to fill the entire table at end of block */
if (ip + mLength < ilimit) {
if (ip + mLength <= ilimit) {
rollingHash = ZSTD_ldm_fillLdmHashTable(
ldmState, rollingHash, lastHashed,
ip + mLength, base, hBits, ldmParams);
ip + mLength, base, hBits, *params);
lastHashed = ip + mLength - 1;
}
ip += mLength;
anchor = ip;
/* Check immediate repcode */
while ( (ip < ilimit)
&& ( (repToConfirm[1] > 0) && (repToConfirm[1] <= (U32)(ip-lowest))
&& (MEM_read32(ip) == MEM_read32(ip - repToConfirm[1])) )) {
}
return iend - anchor;
}
size_t const rLength = ZSTD_count(ip+4, ip+4-repToConfirm[1],
iend) + 4;
/* Swap repToConfirm[1] <=> repToConfirm[0] */
{
U32 const tmpOff = repToConfirm[1];
repToConfirm[1] = repToConfirm[0];
repToConfirm[0] = tmpOff;
}
/*! ZSTD_ldm_reduceTable() :
* reduce table indexes by `reducerValue` */
static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
U32 const reducerValue)
{
U32 u;
for (u = 0; u < size; u++) {
if (table[u].offset < reducerValue) table[u].offset = 0;
else table[u].offset -= reducerValue;
}
}
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
size_t ZSTD_ldm_generateSequences(
ldmState_t* ldmState, rawSeqStore_t* sequences,
ldmParams_t const* params, void const* src, size_t srcSize)
{
U32 const maxDist = 1U << params->windowLog;
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
size_t const kMaxChunkSize = 1 << 20;
size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
size_t chunk;
size_t leftoverSize = 0;
/* Fill the hash table from lastHashed+1 to ip+rLength*/
if (ip + rLength < ilimit) {
rollingHash = ZSTD_ldm_fillLdmHashTable(
ldmState, rollingHash, lastHashed,
ip + rLength, base, hBits, ldmParams);
lastHashed = ip + rLength - 1;
}
ip += rLength;
anchor = ip;
assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
/* Check that ZSTD_window_update() has been called for this chunk prior
* to passing it to this function.
*/
assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize);
/* The input could be very large (in zstdmt), so it must be broken up into
* chunks to enforce the maximmum distance and handle overflow correction.
*/
assert(sequences->pos <= sequences->size);
assert(sequences->size <= sequences->capacity);
for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
size_t const remaining = (size_t)(iend - chunkStart);
BYTE const *const chunkEnd =
(remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
size_t const chunkSize = chunkEnd - chunkStart;
size_t newLeftoverSize;
size_t const prevSize = sequences->size;
assert(chunkStart < iend);
/* 1. Perform overflow correction if necessary. */
if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
U32 const ldmHSize = 1U << params->hashLog;
U32 const correction = ZSTD_window_correctOverflow(
&ldmState->window, /* cycleLog */ 0, maxDist, src);
ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
}
}
/* Overwrite rep */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = repToConfirm[i];
ZSTD_ldm_limitTableUpdate(cctx, anchor);
ZSTD_ldm_fillFastTables(cctx, anchor);
lastLiterals = blockCompressor(cctx, anchor, iend - anchor);
cctx->nextToUpdate = (U32)(iend - base);
/* Restore seqStorePtr->rep */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = savedRep[i];
/* Return the last literals size */
return lastLiterals;
}
size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
{
return ZSTD_compressBlock_ldm_generic(ctx, src, srcSize);
}
static size_t ZSTD_compressBlock_ldm_extDict_generic(
ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
{
ldmState_t* const ldmState = &(ctx->ldmState);
const ldmParams_t ldmParams = ctx->appliedParams.ldmParams;
const U64 hashPower = ldmState->hashPower;
const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
seqStore_t* const seqStorePtr = &(ctx->seqStore);
const BYTE* const base = ctx->base;
const BYTE* const dictBase = ctx->dictBase;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const U32 lowestIndex = ctx->lowLimit;
const BYTE* const dictStart = dictBase + lowestIndex;
const U32 dictLimit = ctx->dictLimit;
const BYTE* const lowPrefixPtr = base + dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
const ZSTD_blockCompressor blockCompressor =
ZSTD_selectBlockCompressor(ctx->appliedParams.cParams.strategy, 1);
U32* const repToConfirm = seqStorePtr->repToConfirm;
U32 savedRep[ZSTD_REP_NUM];
U64 rollingHash = 0;
const BYTE* lastHashed = NULL;
size_t i, lastLiterals;
/* Save seqStorePtr->rep and copy repToConfirm */
for (i = 0; i < ZSTD_REP_NUM; i++) {
savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
}
/* Search Loop */
while (ip < ilimit) { /* < instead of <=, because (ip+1) */
size_t mLength;
const U32 current = (U32)(ip-base);
size_t forwardMatchLength = 0, backwardMatchLength = 0;
ldmEntry_t* bestEntry = NULL;
if (ip != istart) {
rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
lastHashed[ldmParams.minMatchLength],
hashPower);
/* 2. We enforce the maximum offset allowed.
*
* kMaxChunkSize should be small enough that we don't lose too much of
* the window through early invalidation.
* TODO: * Test the chunk size.
* * Try invalidation after the sequence generation and test the
* the offset against maxDist directly.
*/
ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, NULL);
/* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
newLeftoverSize = ZSTD_ldm_generateSequences_internal(
ldmState, sequences, params, chunkStart, chunkSize);
if (ZSTD_isError(newLeftoverSize))
return newLeftoverSize;
/* 4. We add the leftover literals from previous iterations to the first
* newly generated sequence, or add the `newLeftoverSize` if none are
* generated.
*/
/* Prepend the leftover literals from the last call */
if (prevSize < sequences->size) {
sequences->seq[prevSize].litLength += (U32)leftoverSize;
leftoverSize = newLeftoverSize;
} else {
rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
}
lastHashed = ip;
if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
ldmTagMask) {
/* Don't insert and don't look for a match */
ip++;
continue;
}
/* Get the best entry and compute the match lengths */
{
ldmEntry_t* const bucket =
ZSTD_ldm_getBucket(ldmState,
ZSTD_ldm_getSmallHash(rollingHash, hBits),
ldmParams);
ldmEntry_t* cur;
size_t bestMatchLength = 0;
U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
const BYTE* const curMatchBase =
cur->offset < dictLimit ? dictBase : base;
const BYTE* const pMatch = curMatchBase + cur->offset;
const BYTE* const matchEnd =
cur->offset < dictLimit ? dictEnd : iend;
const BYTE* const lowMatchPtr =
cur->offset < dictLimit ? dictStart : lowPrefixPtr;
size_t curForwardMatchLength, curBackwardMatchLength,
curTotalMatchLength;
if (cur->checksum != checksum || cur->offset <= lowestIndex) {
continue;
}
curForwardMatchLength = ZSTD_count_2segments(
ip, pMatch, iend,
matchEnd, lowPrefixPtr);
if (curForwardMatchLength < ldmParams.minMatchLength) {
continue;
}
curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
ip, anchor, pMatch, lowMatchPtr);
curTotalMatchLength = curForwardMatchLength +
curBackwardMatchLength;
if (curTotalMatchLength > bestMatchLength) {
bestMatchLength = curTotalMatchLength;
forwardMatchLength = curForwardMatchLength;
backwardMatchLength = curBackwardMatchLength;
bestEntry = cur;
}
}
}
/* No match found -- continue searching */
if (bestEntry == NULL) {
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
(U32)(lastHashed - base),
ldmParams);
ip++;
continue;
}
/* Match found */
mLength = forwardMatchLength + backwardMatchLength;
ip -= backwardMatchLength;
/* Call the block compressor on the remaining literals */
{
/* ip = current - backwardMatchLength
* The match is at (bestEntry->offset - backwardMatchLength) */
U32 const matchIndex = bestEntry->offset;
U32 const offset = current - matchIndex;
/* Overwrite rep codes */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = repToConfirm[i];
/* Fill the hash table for the block compressor */
ZSTD_ldm_limitTableUpdate(ctx, anchor);
ZSTD_ldm_fillFastTables(ctx, anchor);
/* Call block compressor and get remaining literals */
lastLiterals = blockCompressor(ctx, anchor, ip - anchor);
ctx->nextToUpdate = (U32)(ip - base);
/* Update repToConfirm with the new offset */
for (i = ZSTD_REP_NUM - 1; i > 0; i--)
repToConfirm[i] = repToConfirm[i-1];
repToConfirm[0] = offset;
/* Store the sequence with the leftover literals */
ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
offset + ZSTD_REP_MOVE, mLength - MINMATCH);
}
/* Insert the current entry into the hash table */
ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
(U32)(lastHashed - base),
ldmParams);
/* Fill the hash table from lastHashed+1 to ip+mLength */
assert(ip + backwardMatchLength == lastHashed);
if (ip + mLength < ilimit) {
rollingHash = ZSTD_ldm_fillLdmHashTable(
ldmState, rollingHash, lastHashed,
ip + mLength, base, hBits,
ldmParams);
lastHashed = ip + mLength - 1;
}
ip += mLength;
anchor = ip;
/* check immediate repcode */
while (ip < ilimit) {
U32 const current2 = (U32)(ip-base);
U32 const repIndex2 = current2 - repToConfirm[1];
const BYTE* repMatch2 = repIndex2 < dictLimit ?
dictBase + repIndex2 : base + repIndex2;
if ( (((U32)((dictLimit-1) - repIndex2) >= 3) &
(repIndex2 > lowestIndex)) /* intentional overflow */
&& (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
const BYTE* const repEnd2 = repIndex2 < dictLimit ?
dictEnd : iend;
size_t const repLength2 =
ZSTD_count_2segments(ip+4, repMatch2+4, iend,
repEnd2, lowPrefixPtr) + 4;
U32 tmpOffset = repToConfirm[1];
repToConfirm[1] = repToConfirm[0];
repToConfirm[0] = tmpOffset;
ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
/* Fill the hash table from lastHashed+1 to ip+repLength2*/
if (ip + repLength2 < ilimit) {
rollingHash = ZSTD_ldm_fillLdmHashTable(
ldmState, rollingHash, lastHashed,
ip + repLength2, base, hBits,
ldmParams);
lastHashed = ip + repLength2 - 1;
}
ip += repLength2;
anchor = ip;
continue;
}
break;
assert(newLeftoverSize == chunkSize);
leftoverSize += chunkSize;
}
}
/* Overwrite rep */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = repToConfirm[i];
ZSTD_ldm_limitTableUpdate(ctx, anchor);
ZSTD_ldm_fillFastTables(ctx, anchor);
/* Call the block compressor one last time on the last literals */
lastLiterals = blockCompressor(ctx, anchor, iend - anchor);
ctx->nextToUpdate = (U32)(iend - base);
/* Restore seqStorePtr->rep */
for (i = 0; i < ZSTD_REP_NUM; i++)
seqStorePtr->rep[i] = savedRep[i];
/* Return the last literals size */
return lastLiterals;
return 0;
}
size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
const void* src, size_t srcSize)
void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize, U32 const minMatch) {
while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
if (srcSize <= seq->litLength) {
/* Skip past srcSize literals */
seq->litLength -= (U32)srcSize;
return;
}
srcSize -= seq->litLength;
seq->litLength = 0;
if (srcSize < seq->matchLength) {
/* Skip past the first srcSize of the match */
seq->matchLength -= (U32)srcSize;
if (seq->matchLength < minMatch) {
/* The match is too short, omit it */
if (rawSeqStore->pos + 1 < rawSeqStore->size) {
seq[1].litLength += seq[0].matchLength;
}
rawSeqStore->pos++;
}
return;
}
srcSize -= seq->matchLength;
seq->matchLength = 0;
rawSeqStore->pos++;
}
}
/**
* If the sequence length is longer than remaining then the sequence is split
* between this block and the next.
*
* Returns the current sequence to handle, or if the rest of the block should
* be literals, it returns a sequence with offset == 0.
*/
static rawSeq maybeSplitSequence(rawSeqStore_t* rawSeqStore,
U32 const remaining, U32 const minMatch)
{
return ZSTD_compressBlock_ldm_extDict_generic(ctx, src, srcSize);
rawSeq sequence = rawSeqStore->seq[rawSeqStore->pos];
assert(sequence.offset > 0);
/* Likely: No partial sequence */
if (remaining >= sequence.litLength + sequence.matchLength) {
rawSeqStore->pos++;
return sequence;
}
/* Cut the sequence short (offset == 0 ==> rest is literals). */
if (remaining <= sequence.litLength) {
sequence.offset = 0;
} else if (remaining < sequence.litLength + sequence.matchLength) {
sequence.matchLength = remaining - sequence.litLength;
if (sequence.matchLength < minMatch) {
sequence.offset = 0;
}
}
/* Skip past `remaining` bytes for the future sequences. */
ZSTD_ldm_skipSequences(rawSeqStore, remaining, minMatch);
return sequence;
}
size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
int const extDict)
{
unsigned const minMatch = cParams->searchLength;
ZSTD_blockCompressor const blockCompressor =
ZSTD_selectBlockCompressor(cParams->strategy, extDict);
BYTE const* const base = ms->window.base;
/* Input bounds */
BYTE const* const istart = (BYTE const*)src;
BYTE const* const iend = istart + srcSize;
/* Input positions */
BYTE const* ip = istart;
assert(rawSeqStore->pos <= rawSeqStore->size);
assert(rawSeqStore->size <= rawSeqStore->capacity);
/* Loop through each sequence and apply the block compressor to the lits */
while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
/* maybeSplitSequence updates rawSeqStore->pos */
rawSeq const sequence = maybeSplitSequence(rawSeqStore,
(U32)(iend - ip), minMatch);
int i;
/* End signal */
if (sequence.offset == 0)
break;
assert(sequence.offset <= (1U << cParams->windowLog));
assert(ip + sequence.litLength + sequence.matchLength <= iend);
/* Fill tables for block compressor */
ZSTD_ldm_limitTableUpdate(ms, ip);
ZSTD_ldm_fillFastTables(ms, cParams, ip);
/* Run the block compressor */
{
size_t const newLitLength =
blockCompressor(ms, seqStore, rep, cParams, ip,
sequence.litLength);
ip += sequence.litLength;
ms->nextToUpdate = (U32)(ip - base);
/* Update the repcodes */
for (i = ZSTD_REP_NUM - 1; i > 0; i--)
rep[i] = rep[i-1];
rep[0] = sequence.offset;
/* Store the sequence */
ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength,
sequence.offset + ZSTD_REP_MOVE,
sequence.matchLength - MINMATCH);
ip += sequence.matchLength;
}
}
/* Fill the tables for the block compressor */
ZSTD_ldm_limitTableUpdate(ms, ip);
ZSTD_ldm_fillFastTables(ms, cParams, ip);
/* Compress the last literals */
{
size_t const lastLiterals = blockCompressor(ms, seqStore, rep, cParams,
ip, iend - ip);
ms->nextToUpdate = (U32)(iend - base);
return lastLiterals;
}
}

87
thirdparty/zstd/compress/zstd_ldm.h vendored
View file

@ -22,32 +22,71 @@ extern "C" {
***************************************/
#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_DEFAULTMAX
#define ZSTD_LDM_HASHEVERYLOG_NOTSET 9999
/** ZSTD_compressBlock_ldm_generic() :
/**
* ZSTD_ldm_generateSequences():
*
* This is a block compressor intended for long distance matching.
* Generates the sequences using the long distance match finder.
* Generates long range matching sequences in `sequences`, which parse a prefix
* of the source. `sequences` must be large enough to store every sequence,
* which can be checked with `ZSTD_ldm_getMaxNbSeq()`.
* @returns 0 or an error code.
*
* The function searches for matches of length at least
* ldmParams.minMatchLength using a hash table in cctx->ldmState.
* Matches can be at a distance of up to cParams.windowLog.
*
* Upon finding a match, the unmatched literals are compressed using a
* ZSTD_blockCompressor (depending on the strategy in the compression
* parameters), which stores the matched sequences. The "long distance"
* match is then stored with the remaining literals from the
* ZSTD_blockCompressor. */
size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* cctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
const void* src, size_t srcSize);
* NOTE: The user must have called ZSTD_window_update() for all of the input
* they have, even if they pass it to ZSTD_ldm_generateSequences() in chunks.
* NOTE: This function returns an error if it runs out of space to store
* sequences.
*/
size_t ZSTD_ldm_generateSequences(
ldmState_t* ldms, rawSeqStore_t* sequences,
ldmParams_t const* params, void const* src, size_t srcSize);
/**
* ZSTD_ldm_blockCompress():
*
* Compresses a block using the predefined sequences, along with a secondary
* block compressor. The literals section of every sequence is passed to the
* secondary block compressor, and those sequences are interspersed with the
* predefined sequences. Returns the length of the last literals.
* Updates `rawSeqStore.pos` to indicate how many sequences have been consumed.
* `rawSeqStore.seq` may also be updated to split the last sequence between two
* blocks.
* @return The length of the last literals.
*
* NOTE: The source must be at most the maximum block size, but the predefined
* sequences can be any size, and may be longer than the block. In the case that
* they are longer than the block, the last sequences may need to be split into
* two. We handle that case correctly, and update `rawSeqStore` appropriately.
* NOTE: This function does not return any errors.
*/
size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams,
void const* src, size_t srcSize,
int const extDict);
/**
* ZSTD_ldm_skipSequences():
*
* Skip past `srcSize` bytes worth of sequences in `rawSeqStore`.
* Avoids emitting matches less than `minMatch` bytes.
* Must be called for data with is not passed to ZSTD_ldm_blockCompress().
*/
void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize,
U32 const minMatch);
/** ZSTD_ldm_initializeParameters() :
* Initialize the long distance matching parameters to their default values. */
size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm);
/** ZSTD_ldm_getTableSize() :
* Estimate the space needed for long distance matching tables. */
size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog);
* Estimate the space needed for long distance matching tables or 0 if LDM is
* disabled.
*/
size_t ZSTD_ldm_getTableSize(ldmParams_t params);
/** ZSTD_ldm_getSeqSpace() :
* Return an upper bound on the number of sequences that can be produced by
* the long distance matcher, or 0 if LDM is disabled.
*/
size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
/** ZSTD_ldm_getTableSize() :
* Return prime8bytes^(minMatchLength-1) */
@ -58,8 +97,12 @@ U64 ZSTD_ldm_getHashPower(U32 minMatchLength);
* windowLog and params->hashLog.
*
* Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
* params->hashLog if it is not). */
void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog);
* params->hashLog if it is not).
*
* Ensures that the minMatchLength >= targetLength during optimal parsing.
*/
void ZSTD_ldm_adjustParameters(ldmParams_t* params,
ZSTD_compressionParameters const* cParams);
#if defined (__cplusplus)
}

250
thirdparty/zstd/compress/zstd_opt.c vendored
View file

@ -10,7 +10,6 @@
#include "zstd_compress_internal.h"
#include "zstd_opt.h"
#include "zstd_lazy.h" /* ZSTD_updateTree, ZSTD_updateTree_extDict */
#define ZSTD_LITFREQ_ADD 2 /* scaling factor for litFreq, so that frequencies adapt faster to new stats. Also used for matchSum (?) */
@ -244,14 +243,15 @@ MEM_STATIC U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
/* Update hashTable3 up to ip (excluded)
Assumption : always within prefix (i.e. not within extDict) */
static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* const cctx, const BYTE* const ip)
static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_matchState_t* ms, const BYTE* const ip)
{
U32* const hashTable3 = cctx->hashTable3;
U32 const hashLog3 = cctx->hashLog3;
const BYTE* const base = cctx->base;
U32 idx = cctx->nextToUpdate3;
U32 const target = cctx->nextToUpdate3 = (U32)(ip - base);
U32* const hashTable3 = ms->hashTable3;
U32 const hashLog3 = ms->hashLog3;
const BYTE* const base = ms->window.base;
U32 idx = ms->nextToUpdate3;
U32 const target = ms->nextToUpdate3 = (U32)(ip - base);
size_t const hash3 = ZSTD_hash3Ptr(ip, hashLog3);
assert(hashLog3 > 0);
while(idx < target) {
hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
@ -265,36 +265,173 @@ static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* const cctx, const BYTE*
/*-*************************************
* Binary Tree search
***************************************/
/** ZSTD_insertBt1() : add one or multiple positions to tree.
* ip : assumed <= iend-8 .
* @return : nb of positions added */
static U32 ZSTD_insertBt1(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iend,
U32 const mls, U32 const extDict)
{
U32* const hashTable = ms->hashTable;
U32 const hashLog = cParams->hashLog;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask = (1 << btLog) - 1;
U32 matchIndex = hashTable[h];
size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const base = ms->window.base;
const BYTE* const dictBase = ms->window.dictBase;
const U32 dictLimit = ms->window.dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
const BYTE* match;
const U32 current = (U32)(ip-base);
const U32 btLow = btMask >= current ? 0 : current - btMask;
U32* smallerPtr = bt + 2*(current&btMask);
U32* largerPtr = smallerPtr + 1;
U32 dummy32; /* to be nullified at the end */
U32 const windowLow = ms->window.lowLimit;
U32 matchEndIdx = current+8+1;
size_t bestLength = 8;
U32 nbCompares = 1U << cParams->searchLog;
#ifdef ZSTD_C_PREDICT
U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
predictedSmall += (predictedSmall>0);
predictedLarge += (predictedLarge>0);
#endif /* ZSTD_C_PREDICT */
DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
assert(ip <= iend-8); /* required for h calculation */
hashTable[h] = current; /* Update Hash Table */
while (nbCompares-- && (matchIndex > windowLow)) {
U32* const nextPtr = bt + 2*(matchIndex & btMask);
size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger); /* guaranteed minimum nb of common bytes */
assert(matchIndex < current);
#ifdef ZSTD_C_PREDICT /* note : can create issues when hlog small <= 11 */
const U32* predictPtr = bt + 2*((matchIndex-1) & btMask); /* written this way, as bt is a roll buffer */
if (matchIndex == predictedSmall) {
/* no need to check length, result known */
*smallerPtr = matchIndex;
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop the search */
smallerPtr = nextPtr+1; /* new "smaller" => larger of match */
matchIndex = nextPtr[1]; /* new matchIndex larger than previous (closer to current) */
predictedSmall = predictPtr[1] + (predictPtr[1]>0);
continue;
}
if (matchIndex == predictedLarge) {
*largerPtr = matchIndex;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop the search */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
predictedLarge = predictPtr[0] + (predictPtr[0]>0);
continue;
}
#endif
if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
assert(matchIndex+matchLength >= dictLimit); /* might be wrong if extDict is incorrectly set to 0 */
match = base + matchIndex;
matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
} else {
match = dictBase + matchIndex;
matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
if (matchIndex+matchLength >= dictLimit)
match = base + matchIndex; /* to prepare for next usage of match[matchLength] */
}
if (matchLength > bestLength) {
bestLength = matchLength;
if (matchLength > matchEndIdx - matchIndex)
matchEndIdx = matchIndex + (U32)matchLength;
}
if (ip+matchLength == iend) { /* equal : no way to know if inf or sup */
break; /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
}
if (match[matchLength] < ip[matchLength]) { /* necessarily within buffer */
/* match is smaller than current */
*smallerPtr = matchIndex; /* update smaller idx */
commonLengthSmaller = matchLength; /* all smaller will now have at least this guaranteed common length */
if (matchIndex <= btLow) { smallerPtr=&dummy32; break; } /* beyond tree size, stop searching */
smallerPtr = nextPtr+1; /* new "candidate" => larger than match, which was smaller than target */
matchIndex = nextPtr[1]; /* new matchIndex, larger than previous and closer to current */
} else {
/* match is larger than current */
*largerPtr = matchIndex;
commonLengthLarger = matchLength;
if (matchIndex <= btLow) { largerPtr=&dummy32; break; } /* beyond tree size, stop searching */
largerPtr = nextPtr;
matchIndex = nextPtr[0];
} }
*smallerPtr = *largerPtr = 0;
if (bestLength > 384) return MIN(192, (U32)(bestLength - 384)); /* speed optimization */
assert(matchEndIdx > current + 8);
return matchEndIdx - (current + 8);
}
FORCE_INLINE_TEMPLATE
void ZSTD_updateTree_internal(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iend,
const U32 mls, const U32 extDict)
{
const BYTE* const base = ms->window.base;
U32 const target = (U32)(ip - base);
U32 idx = ms->nextToUpdate;
DEBUGLOG(7, "ZSTD_updateTree_internal, from %u to %u (extDict:%u)",
idx, target, extDict);
while(idx < target)
idx += ZSTD_insertBt1(ms, cParams, base+idx, iend, mls, extDict);
ms->nextToUpdate = target;
}
void ZSTD_updateTree(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* iend)
{
ZSTD_updateTree_internal(ms, cParams, ip, iend, cParams->searchLength, 0 /*extDict*/);
}
FORCE_INLINE_TEMPLATE
U32 ZSTD_insertBtAndGetAllMatches (
ZSTD_CCtx* zc,
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* const ip, const BYTE* const iLimit, int const extDict,
U32 nbCompares, U32 const mls, U32 const sufficient_len,
U32 rep[ZSTD_REP_NUM], U32 const ll0,
ZSTD_match_t* matches, const U32 lengthToBeat)
ZSTD_match_t* matches, const U32 lengthToBeat, U32 const mls /* template */)
{
const BYTE* const base = zc->base;
U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
const BYTE* const base = ms->window.base;
U32 const current = (U32)(ip-base);
U32 const hashLog = zc->appliedParams.cParams.hashLog;
U32 const hashLog = cParams->hashLog;
U32 const minMatch = (mls==3) ? 3 : 4;
U32* const hashTable = zc->hashTable;
U32* const hashTable = ms->hashTable;
size_t const h = ZSTD_hashPtr(ip, hashLog, mls);
U32 matchIndex = hashTable[h];
U32* const bt = zc->chainTable;
U32 const btLog = zc->appliedParams.cParams.chainLog - 1;
U32* const bt = ms->chainTable;
U32 const btLog = cParams->chainLog - 1;
U32 const btMask= (1U << btLog) - 1;
size_t commonLengthSmaller=0, commonLengthLarger=0;
const BYTE* const dictBase = zc->dictBase;
U32 const dictLimit = zc->dictLimit;
const BYTE* const dictBase = ms->window.dictBase;
U32 const dictLimit = ms->window.dictLimit;
const BYTE* const dictEnd = dictBase + dictLimit;
const BYTE* const prefixStart = base + dictLimit;
U32 const btLow = btMask >= current ? 0 : current - btMask;
U32 const windowLow = zc->lowLimit;
U32 const windowLow = ms->window.lowLimit;
U32* smallerPtr = bt + 2*(current&btMask);
U32* largerPtr = bt + 2*(current&btMask) + 1;
U32 matchEndIdx = current+8+1; /* farthest referenced position of any match => detects repetitive patterns */
U32 dummy32; /* to be nullified at the end */
U32 mnum = 0;
U32 nbCompares = 1U << cParams->searchLog;
size_t bestLength = lengthToBeat-1;
DEBUGLOG(7, "ZSTD_insertBtAndGetAllMatches");
@ -335,7 +472,7 @@ U32 ZSTD_insertBtAndGetAllMatches (
/* HC3 match finder */
if ((mls == 3) /*static*/ && (bestLength < mls)) {
U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3(ms, ip);
if ((matchIndex3 > windowLow)
& (current - matchIndex3 < (1<<18)) /*heuristic : longer distance likely too expensive*/ ) {
size_t mlen;
@ -359,7 +496,7 @@ U32 ZSTD_insertBtAndGetAllMatches (
mnum = 1;
if ( (mlen > sufficient_len) |
(ip+mlen == iLimit) ) { /* best possible length */
zc->nextToUpdate = current+1; /* skip insertion */
ms->nextToUpdate = current+1; /* skip insertion */
return 1;
} } } }
@ -416,30 +553,29 @@ U32 ZSTD_insertBtAndGetAllMatches (
*smallerPtr = *largerPtr = 0;
assert(matchEndIdx > current+8);
zc->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
ms->nextToUpdate = matchEndIdx - 8; /* skip repetitive patterns */
return mnum;
}
FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
ZSTD_CCtx* zc, /* Index table will be updated */
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* const iHighLimit, int const extDict,
U32 const maxNbAttempts, U32 const matchLengthSearch, U32 const sufficient_len,
U32 rep[ZSTD_REP_NUM], U32 const ll0,
ZSTD_match_t* matches, U32 const lengthToBeat)
{
U32 const matchLengthSearch = cParams->searchLength;
DEBUGLOG(7, "ZSTD_BtGetAllMatches");
if (ip < zc->base + zc->nextToUpdate) return 0; /* skipped area */
if (extDict) ZSTD_updateTree_extDict(zc, ip, iHighLimit, maxNbAttempts, matchLengthSearch);
else ZSTD_updateTree(zc, ip, iHighLimit, maxNbAttempts, matchLengthSearch);
if (ip < ms->window.base + ms->nextToUpdate) return 0; /* skipped area */
ZSTD_updateTree_internal(ms, cParams, ip, iHighLimit, matchLengthSearch, extDict);
switch(matchLengthSearch)
{
case 3 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 3, sufficient_len, rep, ll0, matches, lengthToBeat);
case 3 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 3);
default :
case 4 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 4, sufficient_len, rep, ll0, matches, lengthToBeat);
case 5 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 5, sufficient_len, rep, ll0, matches, lengthToBeat);
case 4 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 4);
case 5 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 5);
case 7 :
case 6 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 6, sufficient_len, rep, ll0, matches, lengthToBeat);
case 6 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 6);
}
}
@ -527,36 +663,33 @@ static int ZSTD_literalsContribution_cached(
}
FORCE_INLINE_TEMPLATE
size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
size_t ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,seqStore_t* seqStore,
U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams,
const void* src, size_t srcSize,
const int optLevel, const int extDict)
{
seqStore_t* const seqStorePtr = &(ctx->seqStore);
optState_t* const optStatePtr = &(ctx->optState);
optState_t* const optStatePtr = &ms->opt;
const BYTE* const istart = (const BYTE*)src;
const BYTE* ip = istart;
const BYTE* anchor = istart;
const BYTE* const iend = istart + srcSize;
const BYTE* const ilimit = iend - 8;
const BYTE* const base = ctx->base;
const BYTE* const prefixStart = base + ctx->dictLimit;
const BYTE* const base = ms->window.base;
const BYTE* const prefixStart = base + ms->window.dictLimit;
U32 const maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
U32 const sufficient_len = MIN(ctx->appliedParams.cParams.targetLength, ZSTD_OPT_NUM -1);
U32 const mls = ctx->appliedParams.cParams.searchLength;
U32 const minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
U32 const minMatch = (cParams->searchLength == 3) ? 3 : 4;
ZSTD_optimal_t* const opt = optStatePtr->priceTable;
ZSTD_match_t* const matches = optStatePtr->matchTable;
cachedLiteralPrice_t cachedLitPrice;
U32 rep[ZSTD_REP_NUM];
/* init */
DEBUGLOG(5, "ZSTD_compressBlock_opt_generic");
ctx->nextToUpdate3 = ctx->nextToUpdate;
ms->nextToUpdate3 = ms->nextToUpdate;
ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
ip += (ip==prefixStart);
{ int i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
memset(&cachedLitPrice, 0, sizeof(cachedLitPrice));
/* Match Loop */
@ -567,7 +700,7 @@ size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
/* find first match */
{ U32 const litlen = (U32)(ip - anchor);
U32 const ll0 = !litlen;
U32 const nbMatches = ZSTD_BtGetAllMatches(ctx, ip, iend, extDict, maxSearches, mls, sufficient_len, rep, ll0, matches, minMatch);
U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, ip, iend, extDict, rep, ll0, matches, minMatch);
if (!nbMatches) { ip++; continue; }
/* initialize opt[0] */
@ -653,7 +786,7 @@ size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
U32 const litlen = (opt[cur].mlen == 1) ? opt[cur].litlen : 0;
U32 const previousPrice = (cur > litlen) ? opt[cur-litlen].price : 0;
U32 const basePrice = previousPrice + ZSTD_fullLiteralsCost(inr-litlen, litlen, optStatePtr);
U32 const nbMatches = ZSTD_BtGetAllMatches(ctx, inr, iend, extDict, maxSearches, mls, sufficient_len, opt[cur].rep, ll0, matches, minMatch);
U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, inr, iend, extDict, opt[cur].rep, ll0, matches, minMatch);
U32 matchNb;
if (!nbMatches) continue;
@ -749,37 +882,42 @@ _shortestPath: /* cur, last_pos, best_mlen, best_off have to be set */
}
ZSTD_updateStats(optStatePtr, llen, anchor, offset, mlen);
ZSTD_storeSeq(seqStorePtr, llen, anchor, offset, mlen-MINMATCH);
ZSTD_storeSeq(seqStore, llen, anchor, offset, mlen-MINMATCH);
anchor = ip;
} }
ZSTD_setLog2Prices(optStatePtr);
} /* while (ip < ilimit) */
/* Save reps for next block */
{ int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
/* Return the last literals size */
return iend - anchor;
}
size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btopt(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
DEBUGLOG(5, "ZSTD_compressBlock_btopt");
return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0 /*optLevel*/, 0 /*extDict*/);
return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 0 /*optLevel*/, 0 /*extDict*/);
}
size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btultra(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 2 /*optLevel*/, 0 /*extDict*/);
return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 0 /*extDict*/);
}
size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btopt_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0 /*optLevel*/, 1 /*extDict*/);
return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 0 /*optLevel*/, 1 /*extDict*/);
}
size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
size_t ZSTD_compressBlock_btultra_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
{
return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 2 /*optLevel*/, 1 /*extDict*/);
return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 1 /*extDict*/);
}

22
thirdparty/zstd/compress/zstd_opt.h vendored
View file

@ -15,13 +15,25 @@
extern "C" {
#endif
#include "zstd.h" /* ZSTD_CCtx, size_t */
#include "zstd_compress_internal.h"
size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
void ZSTD_updateTree(
ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
const BYTE* ip, const BYTE* iend); /* used in ZSTD_loadDictionaryContent() */
size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
size_t ZSTD_compressBlock_btopt(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btopt_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
size_t ZSTD_compressBlock_btultra_extDict(
ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
#if defined (__cplusplus)
}

1578
thirdparty/zstd/compress/zstdmt_compress.c vendored
View file

File diff suppressed because it is too large Load diff

48
thirdparty/zstd/compress/zstdmt_compress.h vendored
View file

@ -30,15 +30,15 @@
/* === Memory management === */
typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads);
ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads,
ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers);
ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers,
ZSTD_customMem cMem);
ZSTDLIB_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
ZSTDLIB_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
/* === Simple buffer-to-butter one-pass function === */
/* === Simple one-pass compression function === */
ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
void* dst, size_t dstCapacity,
@ -50,7 +50,7 @@ ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
/* === Streaming functions === */
ZSTDLIB_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize); /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it may change in the future, to mean "empty" */
ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize); /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it will change in the future to mean "empty" */
ZSTDLIB_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
@ -68,7 +68,7 @@ ZSTDLIB_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize,
const ZSTD_CDict* cdict,
ZSTD_parameters const params,
ZSTD_parameters params,
unsigned overlapLog);
ZSTDLIB_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
@ -85,7 +85,7 @@ ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
* List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
typedef enum {
ZSTDMT_p_jobSize, /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
ZSTDMT_p_overlapSectionLog /* Each job may reload a part of previous job to enhance compressionr ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window */
ZSTDMT_p_overlapSectionLog /* Each job may reload a part of previous job to enhance compressionr ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
} ZSTDMT_parameter;
/* ZSTDMT_setMTCtxParameter() :
@ -97,30 +97,46 @@ ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter
/*! ZSTDMT_compressStream_generic() :
* Combines ZSTDMT_compressStream() with ZSTDMT_flushStream() or ZSTDMT_endStream()
* Combines ZSTDMT_compressStream() with optional ZSTDMT_flushStream() or ZSTDMT_endStream()
* depending on flush directive.
* @return : minimum amount of data still to be flushed
* 0 if fully flushed
* or an error code */
* or an error code
* note : needs to be init using any ZSTD_initCStream*() variant */
ZSTDLIB_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
ZSTD_outBuffer* output,
ZSTD_inBuffer* input,
ZSTD_EndDirective endOp);
/* === Private definitions; never ever use directly === */
/* ========================================================
* === Private interface, for use by ZSTD_compress.c ===
* === Not exposed in libzstd. Never invoke directly ===
* ======================================================== */
size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, unsigned value);
/* ZSTDMT_CCtxParam_setNbThreads()
* Set nbThreads, and clamp it correctly,
* also reset jobSize and overlapLog */
size_t ZSTDMT_CCtxParam_setNbThreads(ZSTD_CCtx_params* params, unsigned nbThreads);
/* ZSTDMT_CCtxParam_setNbWorkers()
* Set nbWorkers, and clamp it.
* Also reset jobSize and overlapLog */
size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorkers);
/* ZSTDMT_getNbThreads():
/*! ZSTDMT_updateCParams_whileCompressing() :
* Updates only a selected set of compression parameters, to remain compatible with current frame.
* New parameters will be applied to next compression job. */
void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_params* cctxParams);
/* ZSTDMT_getNbWorkers():
* @return nb threads currently active in mtctx.
* mtctx must be valid */
size_t ZSTDMT_getNbThreads(const ZSTDMT_CCtx* mtctx);
unsigned ZSTDMT_getNbWorkers(const ZSTDMT_CCtx* mtctx);
/* ZSTDMT_getFrameProgression():
* tells how much data has been consumed (input) and produced (output) for current frame.
* able to count progression inside worker threads.
*/
ZSTD_frameProgression ZSTDMT_getFrameProgression(ZSTDMT_CCtx* mtctx);
/*! ZSTDMT_initCStream_internal() :
* Private use only. Init streaming operation.
@ -128,7 +144,7 @@ size_t ZSTDMT_getNbThreads(const ZSTDMT_CCtx* mtctx);
* must receive dict, or cdict, or none, but not both.
* @return : 0, or an error code */
size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
const void* dict, size_t dictSize, ZSTD_dictMode_e dictMode,
const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
const ZSTD_CDict* cdict,
ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);