zhangengru/tracy
1
0
Fork 0
mirror of https://github.com/wolfpld/tracy.git synced 2025-03-20 07:40:02 +08:00
Code Issues Projects Wiki Activity

Bump zstd to 1.5.1.

Browse Source
This commit is contained in:
Bartosz Taudul
2021-12-21 12:21:29 +01:00
parent 6fa3491bb7
commit 8c45ed33fd
52 changed files with 10926 additions and 3110 deletions
Download Patch File Download Diff File Expand all files Collapse all files
zstd/dictBuilder/cover.c
+9 -2
View File
@@ -40,6 +40,13 @@
/*-*************************************
* Constants
***************************************/
/**
* There are 32bit indexes used to ref samples, so limit samples size to 4GB
* on 64bit builds.
* For 32bit builds we choose 1 GB.
* Most 32bit platforms have 2GB user-mode addressable space and we allocate a large
* contiguous buffer, so 1GB is already a high limit.
*/
#define COVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((unsigned)-1) : ((unsigned)1 GB))
#define COVER_DEFAULT_SPLITPOINT 1.0
@@ -47,7 +54,7 @@
* Console display
***************************************/
#ifndef LOCALDISPLAYLEVEL
static int g_displayLevel = 2;
static int g_displayLevel = 0;
#endif
#undef DISPLAY
#define DISPLAY(...) \
@@ -735,7 +742,7 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
COVER_map_t activeDmers;
parameters.splitPoint = 1.0;
/* Initialize global data */
g_displayLevel = parameters.zParams.notificationLevel;
g_displayLevel = (int)parameters.zParams.notificationLevel;
/* Checks */
if (!COVER_checkParameters(parameters, dictBufferCapacity)) {
DISPLAYLEVEL(1, "Cover parameters incorrect\n");
zstd/dictBuilder/fastcover.c
+11 -4
View File
@@ -32,6 +32,13 @@
/*-*************************************
* Constants
***************************************/
/**
* There are 32bit indexes used to ref samples, so limit samples size to 4GB
* on 64bit builds.
* For 32bit builds we choose 1 GB.
* Most 32bit platforms have 2GB user-mode addressable space and we allocate a large
* contiguous buffer, so 1GB is already a high limit.
*/
#define FASTCOVER_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((unsigned)-1) : ((unsigned)1 GB))
#define FASTCOVER_MAX_F 31
#define FASTCOVER_MAX_ACCEL 10
@@ -44,7 +51,7 @@
* Console display
***************************************/
#ifndef LOCALDISPLAYLEVEL
static int g_displayLevel = 2;
static int g_displayLevel = 0;
#endif
#undef DISPLAY
#define DISPLAY(...) \
@@ -549,7 +556,7 @@ ZDICT_trainFromBuffer_fastCover(void* dictBuffer, size_t dictBufferCapacity,
ZDICT_cover_params_t coverParams;
FASTCOVER_accel_t accelParams;
/* Initialize global data */
g_displayLevel = parameters.zParams.notificationLevel;
g_displayLevel = (int)parameters.zParams.notificationLevel;
/* Assign splitPoint and f if not provided */
parameters.splitPoint = 1.0;
parameters.f = parameters.f == 0 ? DEFAULT_F : parameters.f;
@@ -632,7 +639,7 @@ ZDICT_optimizeTrainFromBuffer_fastCover(
const unsigned accel = parameters->accel == 0 ? DEFAULT_ACCEL : parameters->accel;
const unsigned shrinkDict = 0;
/* Local variables */
const int displayLevel = parameters->zParams.notificationLevel;
const int displayLevel = (int)parameters->zParams.notificationLevel;
unsigned iteration = 1;
unsigned d;
unsigned k;
@@ -716,7 +723,7 @@ ZDICT_optimizeTrainFromBuffer_fastCover(
data->parameters.splitPoint = splitPoint;
data->parameters.steps = kSteps;
data->parameters.shrinkDict = shrinkDict;
data->parameters.zParams.notificationLevel = g_displayLevel;
data->parameters.zParams.notificationLevel = (unsigned)g_displayLevel;
/* Check the parameters */
if (!FASTCOVER_checkParameters(data->parameters, dictBufferCapacity,
data->ctx->f, accel)) {
zstd/dictBuilder/zdict.c
+99 -28
View File
@@ -135,22 +135,32 @@ static unsigned ZDICT_NbCommonBytes (size_t val)
if (MEM_isLittleEndian()) {
if (MEM_64bits()) {
# if defined(_MSC_VER) && defined(_WIN64)
unsigned long r = 0;
_BitScanForward64( &r, (U64)val );
return (unsigned)(r>>3);
if (val != 0) {
unsigned long r;
_BitScanForward64(&r, (U64)val);
return (unsigned)(r >> 3);
} else {
/* Should not reach this code path */
__assume(0);
}
# elif defined(__GNUC__) && (__GNUC__ >= 3)
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, 0, 2, 3, 6, 1, 5, 3, 5, 1, 3, 4, 4, 2, 5, 6, 7, 7, 0, 1, 2, 3, 3, 4, 6, 2, 6, 5, 5, 3, 4, 5, 6, 7, 1, 2, 4, 6, 4, 4, 5, 7, 2, 6, 5, 7, 6, 7, 7 };
return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
# endif
} else { /* 32 bits */
# if defined(_MSC_VER)
unsigned long r=0;
_BitScanForward( &r, (U32)val );
return (unsigned)(r>>3);
if (val != 0) {
unsigned long r;
_BitScanForward(&r, (U32)val);
return (unsigned)(r >> 3);
} else {
/* Should not reach this code path */
__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, 3, 3, 1, 2, 2, 2, 2, 0, 3, 1, 2, 0, 1, 0, 1, 1 };
return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
@@ -159,11 +169,16 @@ static unsigned ZDICT_NbCommonBytes (size_t val)
} else { /* Big Endian CPU */
if (MEM_64bits()) {
# if defined(_MSC_VER) && defined(_WIN64)
unsigned long r = 0;
_BitScanReverse64( &r, val );
return (unsigned)(r>>3);
if (val != 0) {
unsigned long r;
_BitScanReverse64(&r, val);
return (unsigned)(r >> 3);
} else {
/* Should not reach this code path */
__assume(0);
}
# elif defined(__GNUC__) && (__GNUC__ >= 3)
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 */
@@ -174,11 +189,16 @@ static unsigned ZDICT_NbCommonBytes (size_t val)
# endif
} else { /* 32 bits */
# if defined(_MSC_VER)
unsigned long r = 0;
_BitScanReverse( &r, (unsigned long)val );
return (unsigned)(r>>3);
if (val != 0) {
unsigned long r;
_BitScanReverse(&r, (unsigned long)val);
return (unsigned)(r >> 3);
} else {
/* Should not reach this code path */
__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; }
@@ -235,7 +255,7 @@ static dictItem ZDICT_analyzePos(
U32 savings[LLIMIT] = {0};
const BYTE* b = (const BYTE*)buffer;
size_t maxLength = LLIMIT;
size_t pos = suffix[start];
size_t pos = (size_t)suffix[start];
U32 end = start;
dictItem solution;
@@ -369,7 +389,7 @@ static dictItem ZDICT_analyzePos(
savings[i] = savings[i-1] + (lengthList[i] * (i-3));
DISPLAYLEVEL(4, "Selected dict at position %u, of length %u : saves %u (ratio: %.2f) \n",
(unsigned)pos, (unsigned)maxLength, (unsigned)savings[maxLength], (double)savings[maxLength] / maxLength);
(unsigned)pos, (unsigned)maxLength, (unsigned)savings[maxLength], (double)savings[maxLength] / (double)maxLength);
solution.pos = (U32)pos;
solution.length = (U32)maxLength;
@@ -379,7 +399,7 @@ static dictItem ZDICT_analyzePos(
{ U32 id;
for (id=start; id<end; id++) {
U32 p, pEnd, length;
U32 const testedPos = suffix[id];
U32 const testedPos = (U32)suffix[id];
if (testedPos == pos)
length = solution.length;
else {
@@ -442,7 +462,7 @@ static U32 ZDICT_tryMerge(dictItem* table, dictItem elt, U32 eltNbToSkip, const
if ((table[u].pos + table[u].length >= elt.pos) && (table[u].pos < elt.pos)) { /* overlap, existing < new */
/* append */
int const addedLength = (int)eltEnd - (table[u].pos + table[u].length);
int const addedLength = (int)eltEnd - (int)(table[u].pos + table[u].length);
table[u].savings += elt.length / 8; /* rough approx bonus */
if (addedLength > 0) { /* otherwise, elt fully included into existing */
table[u].length += addedLength;
@@ -766,6 +786,13 @@ static size_t ZDICT_analyzeEntropy(void* dstBuffer, size_t maxDstSize,
pos += fileSizes[u];
}
if (notificationLevel >= 4) {
/* writeStats */
DISPLAYLEVEL(4, "Offset Code Frequencies : \n");
for (u=0; u<=offcodeMax; u++) {
DISPLAYLEVEL(4, "%2u :%7u \n", u, offcodeCount[u]);
} }
/* analyze, build stats, starting with literals */
{ size_t maxNbBits = HUF_buildCTable (hufTable, countLit, 255, huffLog);
if (HUF_isError(maxNbBits)) {
@@ -872,7 +899,7 @@ static size_t ZDICT_analyzeEntropy(void* dstBuffer, size_t maxDstSize,
MEM_writeLE32(dstPtr+8, bestRepOffset[2].offset);
#else
/* at this stage, we don't use the result of "most common first offset",
as the impact of statistics is not properly evaluated */
* as the impact of statistics is not properly evaluated */
MEM_writeLE32(dstPtr+0, repStartValue[0]);
MEM_writeLE32(dstPtr+4, repStartValue[1]);
MEM_writeLE32(dstPtr+8, repStartValue[2]);
@@ -888,6 +915,17 @@ _cleanup:
}
/**
* @returns the maximum repcode value
*/
static U32 ZDICT_maxRep(U32 const reps[ZSTD_REP_NUM])
{
U32 maxRep = reps[0];
int r;
for (r = 1; r < ZSTD_REP_NUM; ++r)
maxRep = MAX(maxRep, reps[r]);
return maxRep;
}
size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
const void* customDictContent, size_t dictContentSize,
@@ -899,11 +937,13 @@ size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
BYTE header[HBUFFSIZE];
int const compressionLevel = (params.compressionLevel == 0) ? ZSTD_CLEVEL_DEFAULT : params.compressionLevel;
U32 const notificationLevel = params.notificationLevel;
/* The final dictionary content must be at least as large as the largest repcode */
size_t const minContentSize = (size_t)ZDICT_maxRep(repStartValue);
size_t paddingSize;
/* check conditions */
DEBUGLOG(4, "ZDICT_finalizeDictionary");
if (dictBufferCapacity < dictContentSize) return ERROR(dstSize_tooSmall);
if (dictContentSize < ZDICT_CONTENTSIZE_MIN) return ERROR(srcSize_wrong);
if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) return ERROR(dstSize_tooSmall);
/* dictionary header */
@@ -927,12 +967,43 @@ size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
hSize += eSize;
}
/* copy elements in final buffer ; note : src and dst buffer can overlap */
if (hSize + dictContentSize > dictBufferCapacity) dictContentSize = dictBufferCapacity - hSize;
{ size_t const dictSize = hSize + dictContentSize;
char* dictEnd = (char*)dictBuffer + dictSize;
memmove(dictEnd - dictContentSize, customDictContent, dictContentSize);
memcpy(dictBuffer, header, hSize);
/* Shrink the content size if it doesn't fit in the buffer */
if (hSize + dictContentSize > dictBufferCapacity) {
dictContentSize = dictBufferCapacity - hSize;
}
/* Pad the dictionary content with zeros if it is too small */
if (dictContentSize < minContentSize) {
RETURN_ERROR_IF(hSize + minContentSize > dictBufferCapacity, dstSize_tooSmall,
"dictBufferCapacity too small to fit max repcode");
paddingSize = minContentSize - dictContentSize;
} else {
paddingSize = 0;
}
{
size_t const dictSize = hSize + paddingSize + dictContentSize;
/* The dictionary consists of the header, optional padding, and the content.
* The padding comes before the content because the "best" position in the
* dictionary is the last byte.
*/
BYTE* const outDictHeader = (BYTE*)dictBuffer;
BYTE* const outDictPadding = outDictHeader + hSize;
BYTE* const outDictContent = outDictPadding + paddingSize;
assert(dictSize <= dictBufferCapacity);
assert(outDictContent + dictContentSize == (BYTE*)dictBuffer + dictSize);
/* First copy the customDictContent into its final location.
* `customDictContent` and `dictBuffer` may overlap, so we must
* do this before any other writes into the output buffer.
* Then copy the header & padding into the output buffer.
*/
memmove(outDictContent, customDictContent, dictContentSize);
memcpy(outDictHeader, header, hSize);
memset(outDictPadding, 0, paddingSize);
return dictSize;
}
}