ddidderr/zstd-rs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #3019 from facebook/huf_traces

Browse Source 
More traces to improved debugging of literals compression
This commit is contained in:
Yann Collet
2022-01-26 18:02:05 -08:00
committed by GitHub
parent a0acf9aa49 8b46895588
commit 7543085013
15 changed files with 214 additions and 106 deletions
Download Patch File Download Diff File Expand all files Collapse all files
doc/zstd_manual.html
+3 -1
View File
@@ -1164,7 +1164,9 @@ size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
<pre><b>ZSTDLIB_STATIC_API size_t ZSTD_compressSequences(ZSTD_CCtx* const cctx, void* dst, size_t dstSize,
const ZSTD_Sequence* inSeqs, size_t inSeqsSize,
const void* src, size_t srcSize);
</b><p> Compress an array of ZSTD_Sequence, generated from the original source buffer, into dst.
</b><p> Compress an array of ZSTD_Sequence, associted with @src buffer, into dst.
@src contains the entire input (not just the literals).
If @srcSize > sum(sequence.length), the remaining bytes are considered all literals
If a dictionary is included, then the cctx should reference the dict. (see: ZSTD_CCtx_refCDict(), ZSTD_CCtx_loadDictionary(), etc.)
The entire source is compressed into a single frame.
lib/common/zstd_internal.h
+4 -4
View File
@@ -95,7 +95,6 @@ typedef enum { bt_raw, bt_rle, bt_compressed, bt_reserved } blockType_e;
#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */
#define HufLog 12
typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingType_e;
#define LONGNBSEQ 0x7F00
@@ -103,6 +102,7 @@ typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingTy
#define MINMATCH 3
#define Litbits 8
#define LitHufLog 11
#define MaxLit ((1<<Litbits) - 1)
#define MaxML 52
#define MaxLL 35
@@ -378,7 +378,7 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val) /* compress, dictBuilder, decodeCorpus
}
# endif
# elif defined(__GNUC__) && (__GNUC__ >= 3) /* GCC Intrinsic */
return __builtin_clz (val) ^ 31;
return (U32)__builtin_clz (val) ^ 31;
# elif defined(__ICCARM__) /* IAR Intrinsic */
return 31 - __CLZ(val);
# else /* Software version */
@@ -417,7 +417,7 @@ MEM_STATIC unsigned ZSTD_countTrailingZeros(size_t val)
}
# endif
# elif defined(__GNUC__) && (__GNUC__ >= 4)
return __builtin_ctzll((U64)val);
return (unsigned)__builtin_ctzll((U64)val);
# else
static const int DeBruijnBytePos[64] = { 0, 1, 2, 7, 3, 13, 8, 19,
4, 25, 14, 28, 9, 34, 20, 56,
@@ -440,7 +440,7 @@ MEM_STATIC unsigned ZSTD_countTrailingZeros(size_t val)
__assume(0);
}
# elif defined(__GNUC__) && (__GNUC__ >= 3)
return __builtin_ctz((U32)val);
return (unsigned)__builtin_ctz((U32)val);
# else
static const int DeBruijnBytePos[32] = { 0, 1, 28, 2, 29, 14, 24, 3,
30, 22, 20, 15, 25, 17, 4, 8,
lib/compress/huf_compress.c
+126 -50
View File
@@ -42,13 +42,67 @@
/* **************************************************************
* Utils
* Required declarations
****************************************************************/
unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
typedef struct nodeElt_s {
U32 count;
U16 parent;
BYTE byte;
BYTE nbBits;
} nodeElt;
/* **************************************************************
* Debug Traces
****************************************************************/
#if DEBUGLEVEL >= 2
static size_t showU32(const U32* arr, size_t size)
{
return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
size_t u;
for (u=0; u<size; u++) {
RAWLOG(6, " %u", arr[u]); (void)arr;
}
RAWLOG(6, " \n");
return size;
}
static size_t HUF_getNbBits(HUF_CElt elt);
static size_t showCTableBits(const HUF_CElt* ctable, size_t size)
{
size_t u;
for (u=0; u<size; u++) {
RAWLOG(6, " %zu", HUF_getNbBits(ctable[u])); (void)ctable;
}
RAWLOG(6, " \n");
return size;
}
static size_t showHNodeSymbols(const nodeElt* hnode, size_t size)
{
size_t u;
for (u=0; u<size; u++) {
RAWLOG(6, " %u", hnode[u].byte); (void)hnode;
}
RAWLOG(6, " \n");
return size;
}
static size_t showHNodeBits(const nodeElt* hnode, size_t size)
{
size_t u;
for (u=0; u<size; u++) {
RAWLOG(6, " %u", hnode[u].nbBits); (void)hnode;
}
RAWLOG(6, " \n");
return size;
}
#endif
/* *******************************************************
* HUF : Huffman block compression
@@ -89,7 +143,10 @@ typedef struct {
S16 norm[HUF_TABLELOG_MAX+1];
} HUF_CompressWeightsWksp;
static size_t HUF_compressWeights(void* dst, size_t dstSize, const void* weightTable, size_t wtSize, void* workspace, size_t workspaceSize)
static size_t
HUF_compressWeights(void* dst, size_t dstSize,
const void* weightTable, size_t wtSize,
void* workspace, size_t workspaceSize)
{
BYTE* const ostart = (BYTE*) dst;
BYTE* op = ostart;
@@ -140,7 +197,7 @@ static size_t HUF_getNbBitsFast(HUF_CElt elt)
static size_t HUF_getValue(HUF_CElt elt)
{
return elt & ~0xFF;
return elt & ~(size_t)0xFF;
}
static size_t HUF_getValueFast(HUF_CElt elt)
@@ -272,68 +329,64 @@ size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void
U32 HUF_getNbBitsFromCTable(HUF_CElt const* CTable, U32 symbolValue)
{
const HUF_CElt* ct = CTable + 1;
const HUF_CElt* const ct = CTable + 1;
assert(symbolValue <= HUF_SYMBOLVALUE_MAX);
return (U32)HUF_getNbBits(ct[symbolValue]);
}
typedef struct nodeElt_s {
U32 count;
U16 parent;
BYTE byte;
BYTE nbBits;
} nodeElt;
/**
* HUF_setMaxHeight():
* Enforces maxNbBits on the Huffman tree described in huffNode.
* Try to enforce @targetNbBits on the Huffman tree described in @huffNode.
*
* It sets all nodes with nbBits > maxNbBits to be maxNbBits. Then it adjusts
* the tree to so that it is a valid canonical Huffman tree.
* It attempts to convert all nodes with nbBits > @targetNbBits
* to employ @targetNbBits instead. Then it adjusts the tree
* so that it remains a valid canonical Huffman tree.
*
* @pre The sum of the ranks of each symbol == 2^largestBits,
* where largestBits == huffNode[lastNonNull].nbBits.
* @post The sum of the ranks of each symbol == 2^largestBits,
* where largestBits is the return value <= maxNbBits.
* where largestBits is the return value (expected <= targetNbBits).
*
* @param huffNode The Huffman tree modified in place to enforce maxNbBits.
* @param huffNode The Huffman tree modified in place to enforce targetNbBits.
* It's presumed sorted, from most frequent to rarest symbol.
* @param lastNonNull The symbol with the lowest count in the Huffman tree.
* @param maxNbBits The maximum allowed number of bits, which the Huffman tree
* @param targetNbBits The allowed number of bits, which the Huffman tree
* may not respect. After this function the Huffman tree will
* respect maxNbBits.
* @return The maximum number of bits of the Huffman tree after adjustment,
* necessarily no more than maxNbBits.
* respect targetNbBits.
* @return The maximum number of bits of the Huffman tree after adjustment.
*/
static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 targetNbBits)
{
const U32 largestBits = huffNode[lastNonNull].nbBits;
/* early exit : no elt > maxNbBits, so the tree is already valid. */
if (largestBits <= maxNbBits) return largestBits;
/* early exit : no elt > targetNbBits, so the tree is already valid. */
if (largestBits <= targetNbBits) return largestBits;
DEBUGLOG(5, "HUF_setMaxHeight (targetNbBits = %u)", targetNbBits);
/* there are several too large elements (at least >= 2) */
{ int totalCost = 0;
const U32 baseCost = 1 << (largestBits - maxNbBits);
const U32 baseCost = 1 << (largestBits - targetNbBits);
int n = (int)lastNonNull;
/* Adjust any ranks > maxNbBits to maxNbBits.
/* Adjust any ranks > targetNbBits to targetNbBits.
* Compute totalCost, which is how far the sum of the ranks is
* we are over 2^largestBits after adjust the offending ranks.
*/
while (huffNode[n].nbBits > maxNbBits) {
while (huffNode[n].nbBits > targetNbBits) {
totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
huffNode[n].nbBits = (BYTE)maxNbBits;
huffNode[n].nbBits = (BYTE)targetNbBits;
n--;
}
/* n stops at huffNode[n].nbBits <= maxNbBits */
assert(huffNode[n].nbBits <= maxNbBits);
/* n end at index of smallest symbol using < maxNbBits */
while (huffNode[n].nbBits == maxNbBits) --n;
/* n stops at huffNode[n].nbBits <= targetNbBits */
assert(huffNode[n].nbBits <= targetNbBits);
/* n end at index of smallest symbol using < targetNbBits */
while (huffNode[n].nbBits == targetNbBits) --n;
/* renorm totalCost from 2^largestBits to 2^maxNbBits
/* renorm totalCost from 2^largestBits to 2^targetNbBits
* note : totalCost is necessarily a multiple of baseCost */
assert((totalCost & (baseCost - 1)) == 0);
totalCost >>= (largestBits - maxNbBits);
totalCost >>= (largestBits - targetNbBits);
assert(totalCost > 0);
/* repay normalized cost */
@@ -342,12 +395,12 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
/* Get pos of last (smallest = lowest cum. count) symbol per rank */
ZSTD_memset(rankLast, 0xF0, sizeof(rankLast));
{ U32 currentNbBits = maxNbBits;
{ U32 currentNbBits = targetNbBits;
int pos;
for (pos=n ; pos >= 0; pos--) {
if (huffNode[pos].nbBits >= currentNbBits) continue;
currentNbBits = huffNode[pos].nbBits; /* < maxNbBits */
rankLast[maxNbBits-currentNbBits] = (U32)pos;
currentNbBits = huffNode[pos].nbBits; /* < targetNbBits */
rankLast[targetNbBits-currentNbBits] = (U32)pos;
} }
while (totalCost > 0) {
@@ -394,7 +447,7 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
rankLast[nBitsToDecrease] = noSymbol;
else {
rankLast[nBitsToDecrease]--;
if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits-nBitsToDecrease)
if (huffNode[rankLast[nBitsToDecrease]].nbBits != targetNbBits-nBitsToDecrease)
rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */
}
} /* while (totalCost > 0) */
@@ -406,11 +459,11 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
* TODO.
*/
while (totalCost < 0) { /* Sometimes, cost correction overshoot */
/* special case : no rank 1 symbol (using maxNbBits-1);
* let's create one from largest rank 0 (using maxNbBits).
/* special case : no rank 1 symbol (using targetNbBits-1);
* let's create one from largest rank 0 (using targetNbBits).
*/
if (rankLast[1] == noSymbol) {
while (huffNode[n].nbBits == maxNbBits) n--;
while (huffNode[n].nbBits == targetNbBits) n--;
huffNode[n+1].nbBits--;
assert(n >= 0);
rankLast[1] = (U32)(n+1);
@@ -424,7 +477,7 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
} /* repay normalized cost */
} /* there are several too large elements (at least >= 2) */
return maxNbBits;
return targetNbBits;
}
typedef struct {
@@ -583,7 +636,7 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy
/* Sort each bucket. */
for (n = RANK_POSITION_DISTINCT_COUNT_CUTOFF; n < RANK_POSITION_TABLE_SIZE - 1; ++n) {
U32 const bucketSize = rankPosition[n].curr-rankPosition[n].base;
int const bucketSize = rankPosition[n].curr - rankPosition[n].base;
U32 const bucketStartIdx = rankPosition[n].base;
if (bucketSize > 1) {
assert(bucketStartIdx < maxSymbolValue1);
@@ -594,6 +647,7 @@ static void HUF_sort(nodeElt huffNode[], const unsigned count[], U32 const maxSy
assert(HUF_isSorted(huffNode, maxSymbolValue1));
}
/** 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 sizeof(HUF_buildCTable_wksp_tables).
@@ -614,6 +668,7 @@ static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue)
int lowS, lowN;
int nodeNb = STARTNODE;
int n, nodeRoot;
DEBUGLOG(5, "HUF_buildTree (alphabet size = %u)", maxSymbolValue + 1);
/* init for parents */
nonNullRank = (int)maxSymbolValue;
while(huffNode[nonNullRank].count == 0) nonNullRank--;
@@ -640,6 +695,8 @@ static int HUF_buildTree(nodeElt* huffNode, U32 maxSymbolValue)
for (n=0; n<=nonNullRank; n++)
huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
DEBUGLOG(6, "Initial distribution of bits completed (%zu sorted symbols)", showHNodeBits(huffNode, maxSymbolValue+1));
return nonNullRank;
}
@@ -677,28 +734,34 @@ static void HUF_buildCTableFromTree(HUF_CElt* CTable, nodeElt const* huffNode, i
CTable[0] = maxNbBits;
}
size_t HUF_buildCTable_wksp (HUF_CElt* CTable, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
size_t
HUF_buildCTable_wksp(HUF_CElt* CTable, const unsigned* count, U32 maxSymbolValue, U32 maxNbBits,
void* workSpace, size_t wkspSize)
{
HUF_buildCTable_wksp_tables* const wksp_tables = (HUF_buildCTable_wksp_tables*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(U32));
HUF_buildCTable_wksp_tables* const wksp_tables =
(HUF_buildCTable_wksp_tables*)HUF_alignUpWorkspace(workSpace, &wkspSize, ZSTD_ALIGNOF(U32));
nodeElt* const huffNode0 = wksp_tables->huffNodeTbl;
nodeElt* const huffNode = huffNode0+1;
int nonNullRank;
DEBUGLOG(5, "HUF_buildCTable_wksp (alphabet size = %u)", maxSymbolValue+1);
/* safety checks */
if (wkspSize < sizeof(HUF_buildCTable_wksp_tables))
return ERROR(workSpace_tooSmall);
return ERROR(workSpace_tooSmall);
if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
return ERROR(maxSymbolValue_tooLarge);
return ERROR(maxSymbolValue_tooLarge);
ZSTD_memset(huffNode0, 0, sizeof(huffNodeTable));
/* sort, decreasing order */
HUF_sort(huffNode, count, maxSymbolValue, wksp_tables->rankPosition);
DEBUGLOG(6, "sorted symbols completed (%zu symbols)", showHNodeSymbols(huffNode, maxSymbolValue+1));
/* build tree */
nonNullRank = HUF_buildTree(huffNode, maxSymbolValue);
/* enforce maxTableLog */
/* determine and enforce maxTableLog */
maxNbBits = HUF_setMaxHeight(huffNode, (U32)nonNullRank, maxNbBits);
if (maxNbBits > HUF_TABLELOG_MAX) return ERROR(GENERIC); /* check fit into table */
@@ -887,7 +950,7 @@ static size_t HUF_closeCStream(HUF_CStream_t* bitC)
{
size_t const nbBits = bitC->bitPos[0] & 0xFF;
if (bitC->ptr >= bitC->endPtr) return 0; /* overflow detected */
return (bitC->ptr - bitC->startPtr) + (nbBits > 0);
return (size_t)(bitC->ptr - bitC->startPtr) + (nbBits > 0);
}
}
@@ -1158,6 +1221,12 @@ static size_t HUF_compressCTable_internal(
return (size_t)(op-ostart);
}
unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
{
return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
}
typedef struct {
unsigned count[HUF_SYMBOLVALUE_MAX + 1];
HUF_CElt CTable[HUF_CTABLE_SIZE_ST(HUF_SYMBOLVALUE_MAX)];
@@ -1188,6 +1257,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
BYTE* const oend = ostart + dstSize;
BYTE* op = ostart;
DEBUGLOG(5, "HUF_compress_internal (srcSize=%zu)", srcSize);
HUF_STATIC_ASSERT(sizeof(*table) + HUF_WORKSPACE_MAX_ALIGNMENT <= HUF_WORKSPACE_SIZE);
/* checks & inits */
@@ -1211,6 +1281,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
DEBUG_STATIC_ASSERT(SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO >= 2);
if (suspectUncompressible && srcSize >= (SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE * SUSPECT_INCOMPRESSIBLE_SAMPLE_RATIO)) {
size_t largestTotal = 0;
DEBUGLOG(5, "input suspected incompressible : sampling to check");
{ unsigned maxSymbolValueBegin = maxSymbolValue;
CHECK_V_F(largestBegin, HIST_count_simple (table->count, &maxSymbolValueBegin, (const BYTE*)src, SUSPECT_INCOMPRESSIBLE_SAMPLE_SIZE) );
largestTotal += largestBegin;
@@ -1227,6 +1298,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; } /* single symbol, rle */
if (largest <= (srcSize >> 7)+4) return 0; /* heuristic : probably not compressible enough */
}
DEBUGLOG(6, "histogram detail completed (%zu symbols)", showU32(table->count, maxSymbolValue+1));
/* Check validity of previous table */
if ( repeat
@@ -1248,6 +1320,7 @@ HUF_compress_internal (void* dst, size_t dstSize,
&table->wksps.buildCTable_wksp, sizeof(table->wksps.buildCTable_wksp));
CHECK_F(maxBits);
huffLog = (U32)maxBits;
DEBUGLOG(6, "bit distribution completed (%zu symbols)", showCTableBits(table->CTable + 1, maxSymbolValue+1));
}
/* Zero unused symbols in CTable, so we can check it for validity */
{
@@ -1300,6 +1373,7 @@ size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat,
int bmi2, unsigned suspectUncompressible)
{
DEBUGLOG(5, "HUF_compress1X_repeat (srcSize = %zu)", srcSize);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_singleStream,
workSpace, wkspSize, hufTable,
@@ -1314,6 +1388,7 @@ size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
unsigned maxSymbolValue, unsigned huffLog,
void* workSpace, size_t wkspSize)
{
DEBUGLOG(5, "HUF_compress4X_wksp (srcSize = %zu)", srcSize);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_fourStreams,
workSpace, wkspSize,
@@ -1330,6 +1405,7 @@ size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
void* workSpace, size_t wkspSize,
HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2, unsigned suspectUncompressible)
{
DEBUGLOG(5, "HUF_compress4X_repeat (srcSize = %zu)", srcSize);
return HUF_compress_internal(dst, dstSize, src, srcSize,
maxSymbolValue, huffLog, HUF_fourStreams,
workSpace, wkspSize,
lib/compress/zstd_compress.c
+21 -20
View File
@@ -746,12 +746,12 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
case ZSTD_c_minMatch :
if (value!=0) /* 0 => use default */
BOUNDCHECK(ZSTD_c_minMatch, value);
CCtxParams->cParams.minMatch = value;
CCtxParams->cParams.minMatch = (U32)value;
return CCtxParams->cParams.minMatch;
case ZSTD_c_targetLength :
BOUNDCHECK(ZSTD_c_targetLength, value);
CCtxParams->cParams.targetLength = value;
CCtxParams->cParams.targetLength = (U32)value;
return CCtxParams->cParams.targetLength;
case ZSTD_c_strategy :
@@ -764,12 +764,12 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
/* Content size written in frame header _when known_ (default:1) */
DEBUGLOG(4, "set content size flag = %u", (value!=0));
CCtxParams->fParams.contentSizeFlag = value != 0;
return CCtxParams->fParams.contentSizeFlag;
return (size_t)CCtxParams->fParams.contentSizeFlag;
case ZSTD_c_checksumFlag :
/* A 32-bits content checksum will be calculated and written at end of frame (default:0) */
CCtxParams->fParams.checksumFlag = value != 0;
return CCtxParams->fParams.checksumFlag;
return (size_t)CCtxParams->fParams.checksumFlag;
case ZSTD_c_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
DEBUGLOG(4, "set dictIDFlag = %u", (value!=0));
@@ -778,7 +778,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
case ZSTD_c_forceMaxWindow :
CCtxParams->forceWindow = (value != 0);
return CCtxParams->forceWindow;
return (size_t)CCtxParams->forceWindow;
case ZSTD_c_forceAttachDict : {
const ZSTD_dictAttachPref_e pref = (ZSTD_dictAttachPref_e)value;
@@ -840,7 +840,7 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
case ZSTD_c_enableDedicatedDictSearch :
CCtxParams->enableDedicatedDictSearch = (value!=0);
return CCtxParams->enableDedicatedDictSearch;
return (size_t)CCtxParams->enableDedicatedDictSearch;
case ZSTD_c_enableLongDistanceMatching :
CCtxParams->ldmParams.enableLdm = (ZSTD_paramSwitch_e)value;
@@ -849,38 +849,38 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
case ZSTD_c_ldmHashLog :
if (value!=0) /* 0 ==> auto */
BOUNDCHECK(ZSTD_c_ldmHashLog, value);
CCtxParams->ldmParams.hashLog = value;
CCtxParams->ldmParams.hashLog = (U32)value;
return CCtxParams->ldmParams.hashLog;
case ZSTD_c_ldmMinMatch :
if (value!=0) /* 0 ==> default */
BOUNDCHECK(ZSTD_c_ldmMinMatch, value);
CCtxParams->ldmParams.minMatchLength = value;
CCtxParams->ldmParams.minMatchLength = (U32)value;
return CCtxParams->ldmParams.minMatchLength;
case ZSTD_c_ldmBucketSizeLog :
if (value!=0) /* 0 ==> default */
BOUNDCHECK(ZSTD_c_ldmBucketSizeLog, value);
CCtxParams->ldmParams.bucketSizeLog = value;
CCtxParams->ldmParams.bucketSizeLog = (U32)value;
return CCtxParams->ldmParams.bucketSizeLog;
case ZSTD_c_ldmHashRateLog :
if (value!=0) /* 0 ==> default */
BOUNDCHECK(ZSTD_c_ldmHashRateLog, value);
CCtxParams->ldmParams.hashRateLog = value;
CCtxParams->ldmParams.hashRateLog = (U32)value;
return CCtxParams->ldmParams.hashRateLog;
case ZSTD_c_targetCBlockSize :
if (value!=0) /* 0 ==> default */
BOUNDCHECK(ZSTD_c_targetCBlockSize, value);
CCtxParams->targetCBlockSize = value;
CCtxParams->targetCBlockSize = (U32)value;
return CCtxParams->targetCBlockSize;
case ZSTD_c_srcSizeHint :
if (value!=0) /* 0 ==> default */
BOUNDCHECK(ZSTD_c_srcSizeHint, value);
CCtxParams->srcSizeHint = value;
return CCtxParams->srcSizeHint;
return (size_t)CCtxParams->srcSizeHint;
case ZSTD_c_stableInBuffer:
BOUNDCHECK(ZSTD_c_stableInBuffer, value);
@@ -3063,9 +3063,9 @@ static size_t ZSTD_buildBlockEntropyStats_literals(void* const src, size_t srcSi
unsigned* const countWksp = (unsigned*)workspace;
const size_t countWkspSize = (HUF_SYMBOLVALUE_MAX + 1) * sizeof(unsigned);
BYTE* const nodeWksp = countWkspStart + countWkspSize;
const size_t nodeWkspSize = wkspEnd-nodeWksp;
const size_t nodeWkspSize = (size_t)(wkspEnd - nodeWksp);
unsigned maxSymbolValue = HUF_SYMBOLVALUE_MAX;
unsigned huffLog = HUF_TABLELOG_DEFAULT;
unsigned huffLog = LitHufLog;
HUF_repeat repeat = prevHuf->repeatMode;
DEBUGLOG(5, "ZSTD_buildBlockEntropyStats_literals (srcSize=%zu)", srcSize);
@@ -5850,12 +5850,13 @@ ZSTD_copySequencesToSeqStoreExplicitBlockDelim(ZSTD_CCtx* cctx,
return 0;
}
/* Returns the number of bytes to move the current read position back by. Only non-zero
* if we ended up splitting a sequence. Otherwise, it may return a ZSTD error if something
* went wrong.
/* Returns the number of bytes to move the current read position back by.
* Only non-zero if we ended up splitting a sequence.
* Otherwise, it may return a ZSTD error if something went wrong.
*
* This function will attempt to scan through blockSize bytes represented by the sequences
* in inSeqs, storing any (partial) sequences.
* This function will attempt to scan through blockSize bytes
* represented by the sequences in @inSeqs,
* storing any (partial) sequences.
*
* Occasionally, we may want to change the actual number of bytes we consumed from inSeqs to
* avoid splitting a match, or to avoid splitting a match such that it would produce a match
@@ -5883,7 +5884,7 @@ ZSTD_copySequencesToSeqStoreNoBlockDelim(ZSTD_CCtx* cctx, ZSTD_sequencePosition*
} else {
dictSize = 0;
}
DEBUGLOG(5, "ZSTD_copySequencesToSeqStore: idx: %u PIS: %u blockSize: %zu", idx, startPosInSequence, blockSize);
DEBUGLOG(5, "ZSTD_copySequencesToSeqStoreNoBlockDelim: idx: %u PIS: %u blockSize: %zu", idx, startPosInSequence, blockSize);
DEBUGLOG(5, "Start seq: idx: %u (of: %u ml: %u ll: %u)", idx, inSeqs[idx].offset, inSeqs[idx].matchLength, inSeqs[idx].litLength);
ZSTD_memcpy(updatedRepcodes.rep, cctx->blockState.prevCBlock->rep, sizeof(repcodes_t));
while (endPosInSequence && idx < inSeqsSize && !finalMatchSplit) {
lib/compress/zstd_compress_internal.h
+4 -4
View File
@@ -713,7 +713,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
}
# endif
# elif defined(__GNUC__) && (__GNUC__ >= 4)
return (__builtin_ctzll((U64)val) >> 3);
return (unsigned)(__builtin_ctzll((U64)val) >> 3);
# else
static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
0, 3, 1, 3, 1, 4, 2, 7,
@@ -736,7 +736,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
__assume(0);
}
# elif defined(__GNUC__) && (__GNUC__ >= 3)
return (__builtin_ctz((U32)val) >> 3);
return (unsigned)(__builtin_ctz((U32)val) >> 3);
# else
static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
3, 2, 2, 1, 3, 2, 0, 1,
@@ -761,7 +761,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
}
# endif
# elif defined(__GNUC__) && (__GNUC__ >= 4)
return (__builtin_clzll(val) >> 3);
return (unsigned)(__builtin_clzll(val) >> 3);
# else
unsigned r;
const unsigned n32 = sizeof(size_t)*4; /* calculate this way due to compiler complaining in 32-bits mode */
@@ -781,7 +781,7 @@ static unsigned ZSTD_NbCommonBytes (size_t val)
__assume(0);
}
# elif defined(__GNUC__) && (__GNUC__ >= 3)
return (__builtin_clz((U32)val) >> 3);
return (unsigned)(__builtin_clz((U32)val) >> 3);
# else
unsigned r;
if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
lib/compress/zstd_compress_literals.c
+38 -12
View File
@@ -13,6 +13,29 @@
***************************************/
#include "zstd_compress_literals.h"
/* **************************************************************
* Debug Traces
****************************************************************/
#if DEBUGLEVEL >= 2
static size_t showHexa(const void* src, size_t srcSize)
{
const BYTE* const ip = (const BYTE*)src;
size_t u;
for (u=0; u<srcSize; u++) {
RAWLOG(6, " %02X", ip[u]); (void)ip;
}
RAWLOG(6, " \n");
return srcSize;
}
#endif
/* **************************************************************
* Literals compression - special cases
****************************************************************/
size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize)
{
BYTE* const ostart = (BYTE*)dst;
@@ -36,7 +59,7 @@ size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src,
}
ZSTD_memcpy(ostart + flSize, src, srcSize);
DEBUGLOG(5, "Raw literals: %u -> %u", (U32)srcSize, (U32)(srcSize + flSize));
DEBUGLOG(5, "Raw (uncompressed) literals: %u -> %u", (U32)srcSize, (U32)(srcSize + flSize));
return srcSize + flSize;
}
@@ -86,6 +109,8 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
DEBUGLOG(5,"ZSTD_compressLiterals (disableLiteralCompression=%i srcSize=%u)",
disableLiteralCompression, (U32)srcSize);
DEBUGLOG(6, "Completed literals listing (%zu bytes)", showHexa(src, srcSize));
/* Prepare nextEntropy assuming reusing the existing table */
ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf));
@@ -100,17 +125,18 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
RETURN_ERROR_IF(dstCapacity < lhSize+1, dstSize_tooSmall, "not enough space for compression");
{ HUF_repeat repeat = prevHuf->repeatMode;
int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
int const preferRepeat = (strategy < ZSTD_lazy) ? srcSize <= 1024 : 0;
typedef size_t (*huf_compress_f)(void*, size_t, const void*, size_t, unsigned, unsigned, void*, size_t, HUF_CElt*, HUF_repeat*, int, int, unsigned);
huf_compress_f huf_compress;
if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
cLitSize = singleStream ?
HUF_compress1X_repeat(
ostart+lhSize, dstCapacity-lhSize, src, srcSize,
HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
(HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2, suspectUncompressible) :
HUF_compress4X_repeat(
ostart+lhSize, dstCapacity-lhSize, src, srcSize,
HUF_SYMBOLVALUE_MAX, HUF_TABLELOG_DEFAULT, entropyWorkspace, entropyWorkspaceSize,
(HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2, suspectUncompressible);
huf_compress = singleStream ? HUF_compress1X_repeat : HUF_compress4X_repeat;
cLitSize = huf_compress(ostart+lhSize, dstCapacity-lhSize,
src, srcSize,
HUF_SYMBOLVALUE_MAX, LitHufLog,
entropyWorkspace, entropyWorkspaceSize,
(HUF_CElt*)nextHuf->CTable,
&repeat, preferRepeat,
bmi2, suspectUncompressible);
if (repeat != HUF_repeat_none) {
/* reused the existing table */
DEBUGLOG(5, "Reusing previous huffman table");
@@ -136,7 +162,7 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
switch(lhSize)
{
case 3: /* 2 - 2 - 10 - 10 */
{ U32 const lhc = hType + ((!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
{ U32 const lhc = hType + ((U32)(!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
MEM_writeLE24(ostart, lhc);
break;
}
lib/compress/zstd_cwksp.h
+1 -1
View File
@@ -499,7 +499,7 @@ MEM_STATIC void ZSTD_cwksp_clean_tables(ZSTD_cwksp* ws) {
assert(ws->tableValidEnd >= ws->objectEnd);
assert(ws->tableValidEnd <= ws->allocStart);
if (ws->tableValidEnd < ws->tableEnd) {
ZSTD_memset(ws->tableValidEnd, 0, (BYTE*)ws->tableEnd - (BYTE*)ws->tableValidEnd);
ZSTD_memset(ws->tableValidEnd, 0, (size_t)((BYTE*)ws->tableEnd - (BYTE*)ws->tableValidEnd));
}
ZSTD_cwksp_mark_tables_clean(ws);
}
lib/decompress/zstd_ddict.c
+1 -1
View File
@@ -134,7 +134,7 @@ static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict,
ZSTD_memcpy(internalBuffer, dict, dictSize);
}
ddict->dictSize = dictSize;
ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */
ddict->entropy.hufTable[0] = (HUF_DTable)((ZSTD_HUFFDTABLE_CAPACITY_LOG)*0x1000001); /* cover both little and big endian */
/* parse dictionary content */
FORWARD_IF_ERROR( ZSTD_loadEntropy_intoDDict(ddict, dictContentType) , "");
lib/decompress/zstd_decompress.c
+1 -1
View File
@@ -1453,7 +1453,7 @@ size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
dctx->prefixStart = NULL;
dctx->virtualStart = NULL;
dctx->dictEnd = NULL;
dctx->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001); /* cover both little and big endian */
dctx->entropy.hufTable[0] = (HUF_DTable)((ZSTD_HUFFDTABLE_CAPACITY_LOG)*0x1000001); /* cover both little and big endian */
dctx->litEntropy = dctx->fseEntropy = 0;
dctx->dictID = 0;
dctx->bType = bt_reserved;
lib/decompress/zstd_decompress_block.c
+1 -1
View File
@@ -1993,7 +1993,7 @@ ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
/* Decode literals section */
{ size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, streaming);
DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize);
DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : cSize=%u, nbLiterals=%zu", (U32)litCSize, dctx->litSize);
if (ZSTD_isError(litCSize)) return litCSize;
ip += litCSize;
srcSize -= litCSize;
lib/decompress/zstd_decompress_internal.h
+2 -1
View File
@@ -75,12 +75,13 @@ static UNUSED_ATTR const U32 ML_base[MaxML+1] = {
#define ZSTD_BUILD_FSE_TABLE_WKSP_SIZE (sizeof(S16) * (MaxSeq + 1) + (1u << MaxFSELog) + sizeof(U64))
#define ZSTD_BUILD_FSE_TABLE_WKSP_SIZE_U32 ((ZSTD_BUILD_FSE_TABLE_WKSP_SIZE + sizeof(U32) - 1) / sizeof(U32))
#define ZSTD_HUFFDTABLE_CAPACITY_LOG 12
typedef struct {
ZSTD_seqSymbol LLTable[SEQSYMBOL_TABLE_SIZE(LLFSELog)]; /* Note : Space reserved for FSE Tables */
ZSTD_seqSymbol OFTable[SEQSYMBOL_TABLE_SIZE(OffFSELog)]; /* is also used as temporary workspace while building hufTable during DDict creation */
ZSTD_seqSymbol MLTable[SEQSYMBOL_TABLE_SIZE(MLFSELog)]; /* and therefore must be at least HUF_DECOMPRESS_WORKSPACE_SIZE large */
HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)]; /* can accommodate HUF_decompress4X */
HUF_DTable hufTable[HUF_DTABLE_SIZE(ZSTD_HUFFDTABLE_CAPACITY_LOG)]; /* can accommodate HUF_decompress4X */
U32 rep[ZSTD_REP_NUM];
U32 workspace[ZSTD_BUILD_FSE_TABLE_WKSP_SIZE_U32];
} ZSTD_entropyDTables_t;
lib/legacy/zstd_v05.c
+3 -3
View File
@@ -485,7 +485,7 @@ static const size_t ZSTDv05_frameHeaderSize_min = 5;
#define FSEv05_ENCODING_DYNAMIC 3
#define HufLog 12
#define ZSTD_HUFFDTABLE_CAPACITY_LOG 12
#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */
@@ -2645,7 +2645,7 @@ struct ZSTDv05_DCtx_s
FSEv05_DTable LLTable[FSEv05_DTABLE_SIZE_U32(LLFSEv05Log)];
FSEv05_DTable OffTable[FSEv05_DTABLE_SIZE_U32(OffFSEv05Log)];
FSEv05_DTable MLTable[FSEv05_DTABLE_SIZE_U32(MLFSEv05Log)];
unsigned hufTableX4[HUFv05_DTABLE_SIZE(HufLog)];
unsigned hufTableX4[HUFv05_DTABLE_SIZE(ZSTD_HUFFDTABLE_CAPACITY_LOG)];
const void* previousDstEnd;
const void* base;
const void* vBase;
@@ -2673,7 +2673,7 @@ size_t ZSTDv05_decompressBegin(ZSTDv05_DCtx* dctx)
dctx->base = NULL;
dctx->vBase = NULL;
dctx->dictEnd = NULL;
dctx->hufTableX4[0] = HufLog;
dctx->hufTableX4[0] = ZSTD_HUFFDTABLE_CAPACITY_LOG;
dctx->flagStaticTables = 0;
return 0;
}
lib/legacy/zstd_v06.c
+3 -3
View File
@@ -479,7 +479,7 @@ typedef enum { bt_compressed, bt_raw, bt_rle, bt_end } blockType_t;
#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */
#define HufLog 12
#define ZSTD_HUFFDTABLE_CAPACITY_LOG 12
#define IS_HUF 0
#define IS_PCH 1
@@ -2806,7 +2806,7 @@ struct ZSTDv06_DCtx_s
FSEv06_DTable LLTable[FSEv06_DTABLE_SIZE_U32(LLFSELog)];
FSEv06_DTable OffTable[FSEv06_DTABLE_SIZE_U32(OffFSELog)];
FSEv06_DTable MLTable[FSEv06_DTABLE_SIZE_U32(MLFSELog)];
unsigned hufTableX4[HUFv06_DTABLE_SIZE(HufLog)];
unsigned hufTableX4[HUFv06_DTABLE_SIZE(ZSTD_HUFFDTABLE_CAPACITY_LOG)];
const void* previousDstEnd;
const void* base;
const void* vBase;
@@ -2834,7 +2834,7 @@ size_t ZSTDv06_decompressBegin(ZSTDv06_DCtx* dctx)
dctx->base = NULL;
dctx->vBase = NULL;
dctx->dictEnd = NULL;
dctx->hufTableX4[0] = HufLog;
dctx->hufTableX4[0] = ZSTD_HUFFDTABLE_CAPACITY_LOG;
dctx->flagRepeatTable = 0;
return 0;
}
lib/legacy/zstd_v07.c
+3 -3
View File
@@ -2717,7 +2717,7 @@ typedef enum { bt_compressed, bt_raw, bt_rle, bt_end } blockType_t;
#define MIN_SEQUENCES_SIZE 1 /* nbSeq==0 */
#define MIN_CBLOCK_SIZE (1 /*litCSize*/ + 1 /* RLE or RAW */ + MIN_SEQUENCES_SIZE /* nbSeq==0 */) /* for a non-null block */
#define HufLog 12
#define ZSTD_HUFFDTABLE_CAPACITY_LOG 12
typedef enum { lbt_huffman, lbt_repeat, lbt_raw, lbt_rle } litBlockType_t;
#define LONGNBSEQ 0x7F00
@@ -2931,7 +2931,7 @@ struct ZSTDv07_DCtx_s
FSEv07_DTable LLTable[FSEv07_DTABLE_SIZE_U32(LLFSELog)];
FSEv07_DTable OffTable[FSEv07_DTABLE_SIZE_U32(OffFSELog)];
FSEv07_DTable MLTable[FSEv07_DTABLE_SIZE_U32(MLFSELog)];
HUFv07_DTable hufTable[HUFv07_DTABLE_SIZE(HufLog)]; /* can accommodate HUFv07_decompress4X */
HUFv07_DTable hufTable[HUFv07_DTABLE_SIZE(ZSTD_HUFFDTABLE_CAPACITY_LOG)]; /* can accommodate HUFv07_decompress4X */
const void* previousDstEnd;
const void* base;
const void* vBase;
@@ -2967,7 +2967,7 @@ size_t ZSTDv07_decompressBegin(ZSTDv07_DCtx* dctx)
dctx->base = NULL;
dctx->vBase = NULL;
dctx->dictEnd = NULL;
dctx->hufTable[0] = (HUFv07_DTable)((HufLog)*0x1000001);
dctx->hufTable[0] = (HUFv07_DTable)((ZSTD_HUFFDTABLE_CAPACITY_LOG)*0x1000001);
dctx->litEntropy = dctx->fseEntropy = 0;
dctx->dictID = 0;
{ int i; for (i=0; i<ZSTDv07_REP_NUM; i++) dctx->rep[i] = repStartValue[i]; }
lib/zstd.h
+3 -1
View File
@@ -1409,7 +1409,9 @@ ZSTD_generateSequences( ZSTD_CCtx* zc,
ZSTDLIB_STATIC_API size_t ZSTD_mergeBlockDelimiters(ZSTD_Sequence* sequences, size_t seqsSize);
/*! ZSTD_compressSequences() :
* Compress an array of ZSTD_Sequence, generated from the original source buffer, into dst.
* Compress an array of ZSTD_Sequence, associted with @src buffer, into dst.
* @src contains the entire input (not just the literals).
* If @srcSize > sum(sequence.length), the remaining bytes are considered all literals
* If a dictionary is included, then the cctx should reference the dict. (see: ZSTD_CCtx_refCDict(), ZSTD_CCtx_loadDictionary(), etc.)
* The entire source is compressed into a single frame.
*