linux/lib/zstd/huf_compress.c

773 lines
26 KiB
C
Raw Normal View History

lib: Add zstd modules Add zstd compression and decompression kernel modules. zstd offers a wide varity of compression speed and quality trade-offs. It can compress at speeds approaching lz4, and quality approaching lzma. zstd decompressions at speeds more than twice as fast as zlib, and decompression speed remains roughly the same across all compression levels. The code was ported from the upstream zstd source repository. The `linux/zstd.h` header was modified to match linux kernel style. The cross-platform and allocation code was stripped out. Instead zstd requires the caller to pass a preallocated workspace. The source files were clang-formatted [1] to match the Linux Kernel style as much as possible. Otherwise, the code was unmodified. We would like to avoid as much further manual modification to the source code as possible, so it will be easier to keep the kernel zstd up to date. I benchmarked zstd compression as a special character device. I ran zstd and zlib compression at several levels, as well as performing no compression, which measure the time spent copying the data to kernel space. Data is passed to the compresser 4096 B at a time. The benchmark file is located in the upstream zstd source repository under `contrib/linux-kernel/zstd_compress_test.c` [2]. I ran the benchmarks on a Ubuntu 14.04 VM with 2 cores and 4 GiB of RAM. The VM is running on a MacBook Pro with a 3.1 GHz Intel Core i7 processor, 16 GB of RAM, and a SSD. I benchmarked using `silesia.tar` [3], which is 211,988,480 B large. Run the following commands for the benchmark: sudo modprobe zstd_compress_test sudo mknod zstd_compress_test c 245 0 sudo cp silesia.tar zstd_compress_test The time is reported by the time of the userland `cp`. The MB/s is computed with 1,536,217,008 B / time(buffer size, hash) which includes the time to copy from userland. The Adjusted MB/s is computed with 1,536,217,088 B / (time(buffer size, hash) - time(buffer size, none)). The memory reported is the amount of memory the compressor requests. | Method | Size (B) | Time (s) | Ratio | MB/s | Adj MB/s | Mem (MB) | |----------|----------|----------|-------|---------|----------|----------| | none | 11988480 | 0.100 | 1 | 2119.88 | - | - | | zstd -1 | 73645762 | 1.044 | 2.878 | 203.05 | 224.56 | 1.23 | | zstd -3 | 66988878 | 1.761 | 3.165 | 120.38 | 127.63 | 2.47 | | zstd -5 | 65001259 | 2.563 | 3.261 | 82.71 | 86.07 | 2.86 | | zstd -10 | 60165346 | 13.242 | 3.523 | 16.01 | 16.13 | 13.22 | | zstd -15 | 58009756 | 47.601 | 3.654 | 4.45 | 4.46 | 21.61 | | zstd -19 | 54014593 | 102.835 | 3.925 | 2.06 | 2.06 | 60.15 | | zlib -1 | 77260026 | 2.895 | 2.744 | 73.23 | 75.85 | 0.27 | | zlib -3 | 72972206 | 4.116 | 2.905 | 51.50 | 52.79 | 0.27 | | zlib -6 | 68190360 | 9.633 | 3.109 | 22.01 | 22.24 | 0.27 | | zlib -9 | 67613382 | 22.554 | 3.135 | 9.40 | 9.44 | 0.27 | I benchmarked zstd decompression using the same method on the same machine. The benchmark file is located in the upstream zstd repo under `contrib/linux-kernel/zstd_decompress_test.c` [4]. The memory reported is the amount of memory required to decompress data compressed with the given compression level. If you know the maximum size of your input, you can reduce the memory usage of decompression irrespective of the compression level. | Method | Time (s) | MB/s | Adjusted MB/s | Memory (MB) | |----------|----------|---------|---------------|-------------| | none | 0.025 | 8479.54 | - | - | | zstd -1 | 0.358 | 592.15 | 636.60 | 0.84 | | zstd -3 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -5 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -10 | 0.374 | 566.81 | 607.42 | 2.51 | | zstd -15 | 0.379 | 559.34 | 598.84 | 4.61 | | zstd -19 | 0.412 | 514.54 | 547.77 | 8.80 | | zlib -1 | 0.940 | 225.52 | 231.68 | 0.04 | | zlib -3 | 0.883 | 240.08 | 247.07 | 0.04 | | zlib -6 | 0.844 | 251.17 | 258.84 | 0.04 | | zlib -9 | 0.837 | 253.27 | 287.64 | 0.04 | Tested in userland using the test-suite in the zstd repo under `contrib/linux-kernel/test/UserlandTest.cpp` [5] by mocking the kernel functions. Fuzz tested using libfuzzer [6] with the fuzz harnesses under `contrib/linux-kernel/test/{RoundTripCrash.c,DecompressCrash.c}` [7] [8] with ASAN, UBSAN, and MSAN. Additionaly, it was tested while testing the BtrFS and SquashFS patches coming next. [1] https://clang.llvm.org/docs/ClangFormat.html [2] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_compress_test.c [3] http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia [4] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_decompress_test.c [5] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/UserlandTest.cpp [6] http://llvm.org/docs/LibFuzzer.html [7] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/RoundTripCrash.c [8] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/DecompressCrash.c zstd source repository: https://github.com/facebook/zstd Signed-off-by: Nick Terrell <terrelln@fb.com> Signed-off-by: Chris Mason <clm@fb.com>
2017-08-10 05:35:53 +03:00
/*
* Huffman encoder, part of New Generation Entropy library
* Copyright (C) 2013-2016, Yann Collet.
*
* BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* This program is free software; you can redistribute it and/or modify it under
* the terms of the GNU General Public License version 2 as published by the
* Free Software Foundation. This program is dual-licensed; you may select
* either version 2 of the GNU General Public License ("GPL") or BSD license
* ("BSD").
*
* You can contact the author at :
* - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
*/
/* **************************************************************
* Includes
****************************************************************/
#include "bitstream.h"
#include "fse.h" /* header compression */
#include "huf.h"
#include <linux/kernel.h>
#include <linux/string.h> /* memcpy, memset */
/* **************************************************************
* Error Management
****************************************************************/
#define HUF_STATIC_ASSERT(c) \
{ \
enum { HUF_static_assert = 1 / (int)(!!(c)) }; \
} /* use only *after* variable declarations */
#define CHECK_V_F(e, f) \
size_t const e = f; \
if (ERR_isError(e)) \
return f
#define CHECK_F(f) \
{ \
CHECK_V_F(_var_err__, f); \
}
/* **************************************************************
* Utils
****************************************************************/
unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue)
{
return FSE_optimalTableLog_internal(maxTableLog, srcSize, maxSymbolValue, 1);
}
/* *******************************************************
* HUF : Huffman block compression
*********************************************************/
/* HUF_compressWeights() :
* Same as FSE_compress(), but dedicated to huff0's weights compression.
* The use case needs much less stack memory.
* Note : all elements within weightTable are supposed to be <= HUF_TABLELOG_MAX.
*/
#define MAX_FSE_TABLELOG_FOR_HUFF_HEADER 6
size_t HUF_compressWeights_wksp(void *dst, size_t dstSize, const void *weightTable, size_t wtSize, void *workspace, size_t workspaceSize)
{
BYTE *const ostart = (BYTE *)dst;
BYTE *op = ostart;
BYTE *const oend = ostart + dstSize;
U32 maxSymbolValue = HUF_TABLELOG_MAX;
U32 tableLog = MAX_FSE_TABLELOG_FOR_HUFF_HEADER;
FSE_CTable *CTable;
U32 *count;
S16 *norm;
size_t spaceUsed32 = 0;
HUF_STATIC_ASSERT(sizeof(FSE_CTable) == sizeof(U32));
CTable = (FSE_CTable *)((U32 *)workspace + spaceUsed32);
spaceUsed32 += FSE_CTABLE_SIZE_U32(MAX_FSE_TABLELOG_FOR_HUFF_HEADER, HUF_TABLELOG_MAX);
count = (U32 *)workspace + spaceUsed32;
spaceUsed32 += HUF_TABLELOG_MAX + 1;
norm = (S16 *)((U32 *)workspace + spaceUsed32);
spaceUsed32 += ALIGN(sizeof(S16) * (HUF_TABLELOG_MAX + 1), sizeof(U32)) >> 2;
if ((spaceUsed32 << 2) > workspaceSize)
return ERROR(tableLog_tooLarge);
workspace = (U32 *)workspace + spaceUsed32;
workspaceSize -= (spaceUsed32 << 2);
/* init conditions */
if (wtSize <= 1)
return 0; /* Not compressible */
/* Scan input and build symbol stats */
{
CHECK_V_F(maxCount, FSE_count_simple(count, &maxSymbolValue, weightTable, wtSize));
if (maxCount == wtSize)
return 1; /* only a single symbol in src : rle */
if (maxCount == 1)
return 0; /* each symbol present maximum once => not compressible */
}
tableLog = FSE_optimalTableLog(tableLog, wtSize, maxSymbolValue);
CHECK_F(FSE_normalizeCount(norm, tableLog, count, wtSize, maxSymbolValue));
/* Write table description header */
{
CHECK_V_F(hSize, FSE_writeNCount(op, oend - op, norm, maxSymbolValue, tableLog));
op += hSize;
}
/* Compress */
CHECK_F(FSE_buildCTable_wksp(CTable, norm, maxSymbolValue, tableLog, workspace, workspaceSize));
{
CHECK_V_F(cSize, FSE_compress_usingCTable(op, oend - op, weightTable, wtSize, CTable));
if (cSize == 0)
return 0; /* not enough space for compressed data */
op += cSize;
}
return op - ostart;
}
struct HUF_CElt_s {
U16 val;
BYTE nbBits;
}; /* typedef'd to HUF_CElt within "huf.h" */
/*! HUF_writeCTable_wksp() :
`CTable` : Huffman tree to save, using huf representation.
@return : size of saved CTable */
size_t HUF_writeCTable_wksp(void *dst, size_t maxDstSize, const HUF_CElt *CTable, U32 maxSymbolValue, U32 huffLog, void *workspace, size_t workspaceSize)
{
BYTE *op = (BYTE *)dst;
U32 n;
BYTE *bitsToWeight;
BYTE *huffWeight;
size_t spaceUsed32 = 0;
bitsToWeight = (BYTE *)((U32 *)workspace + spaceUsed32);
spaceUsed32 += ALIGN(HUF_TABLELOG_MAX + 1, sizeof(U32)) >> 2;
huffWeight = (BYTE *)((U32 *)workspace + spaceUsed32);
spaceUsed32 += ALIGN(HUF_SYMBOLVALUE_MAX, sizeof(U32)) >> 2;
if ((spaceUsed32 << 2) > workspaceSize)
return ERROR(tableLog_tooLarge);
workspace = (U32 *)workspace + spaceUsed32;
workspaceSize -= (spaceUsed32 << 2);
/* check conditions */
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
return ERROR(maxSymbolValue_tooLarge);
/* convert to weight */
bitsToWeight[0] = 0;
for (n = 1; n < huffLog + 1; n++)
bitsToWeight[n] = (BYTE)(huffLog + 1 - n);
for (n = 0; n < maxSymbolValue; n++)
huffWeight[n] = bitsToWeight[CTable[n].nbBits];
/* attempt weights compression by FSE */
{
CHECK_V_F(hSize, HUF_compressWeights_wksp(op + 1, maxDstSize - 1, huffWeight, maxSymbolValue, workspace, workspaceSize));
if ((hSize > 1) & (hSize < maxSymbolValue / 2)) { /* FSE compressed */
op[0] = (BYTE)hSize;
return hSize + 1;
}
}
/* write raw values as 4-bits (max : 15) */
if (maxSymbolValue > (256 - 128))
return ERROR(GENERIC); /* should not happen : likely means source cannot be compressed */
if (((maxSymbolValue + 1) / 2) + 1 > maxDstSize)
return ERROR(dstSize_tooSmall); /* not enough space within dst buffer */
op[0] = (BYTE)(128 /*special case*/ + (maxSymbolValue - 1));
huffWeight[maxSymbolValue] = 0; /* to be sure it doesn't cause msan issue in final combination */
for (n = 0; n < maxSymbolValue; n += 2)
op[(n / 2) + 1] = (BYTE)((huffWeight[n] << 4) + huffWeight[n + 1]);
return ((maxSymbolValue + 1) / 2) + 1;
}
size_t HUF_readCTable_wksp(HUF_CElt *CTable, U32 maxSymbolValue, const void *src, size_t srcSize, void *workspace, size_t workspaceSize)
{
U32 *rankVal;
BYTE *huffWeight;
U32 tableLog = 0;
U32 nbSymbols = 0;
size_t readSize;
size_t spaceUsed32 = 0;
rankVal = (U32 *)workspace + spaceUsed32;
spaceUsed32 += HUF_TABLELOG_ABSOLUTEMAX + 1;
huffWeight = (BYTE *)((U32 *)workspace + spaceUsed32);
spaceUsed32 += ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
if ((spaceUsed32 << 2) > workspaceSize)
return ERROR(tableLog_tooLarge);
workspace = (U32 *)workspace + spaceUsed32;
workspaceSize -= (spaceUsed32 << 2);
/* get symbol weights */
readSize = HUF_readStats_wksp(huffWeight, HUF_SYMBOLVALUE_MAX + 1, rankVal, &nbSymbols, &tableLog, src, srcSize, workspace, workspaceSize);
if (ERR_isError(readSize))
return readSize;
/* check result */
if (tableLog > HUF_TABLELOG_MAX)
return ERROR(tableLog_tooLarge);
if (nbSymbols > maxSymbolValue + 1)
return ERROR(maxSymbolValue_tooSmall);
/* Prepare base value per rank */
{
U32 n, nextRankStart = 0;
for (n = 1; n <= tableLog; n++) {
U32 curr = nextRankStart;
nextRankStart += (rankVal[n] << (n - 1));
rankVal[n] = curr;
}
}
/* fill nbBits */
{
U32 n;
for (n = 0; n < nbSymbols; n++) {
const U32 w = huffWeight[n];
CTable[n].nbBits = (BYTE)(tableLog + 1 - w);
}
}
/* fill val */
{
U16 nbPerRank[HUF_TABLELOG_MAX + 2] = {0}; /* support w=0=>n=tableLog+1 */
U16 valPerRank[HUF_TABLELOG_MAX + 2] = {0};
{
U32 n;
for (n = 0; n < nbSymbols; n++)
nbPerRank[CTable[n].nbBits]++;
}
/* determine stating value per rank */
valPerRank[tableLog + 1] = 0; /* for w==0 */
{
U16 min = 0;
U32 n;
for (n = tableLog; n > 0; n--) { /* start at n=tablelog <-> w=1 */
valPerRank[n] = min; /* get starting value within each rank */
min += nbPerRank[n];
min >>= 1;
}
}
/* assign value within rank, symbol order */
{
U32 n;
for (n = 0; n <= maxSymbolValue; n++)
CTable[n].val = valPerRank[CTable[n].nbBits]++;
}
}
return readSize;
}
typedef struct nodeElt_s {
U32 count;
U16 parent;
BYTE byte;
BYTE nbBits;
} nodeElt;
static U32 HUF_setMaxHeight(nodeElt *huffNode, U32 lastNonNull, U32 maxNbBits)
{
const U32 largestBits = huffNode[lastNonNull].nbBits;
if (largestBits <= maxNbBits)
return largestBits; /* early exit : no elt > maxNbBits */
/* there are several too large elements (at least >= 2) */
{
int totalCost = 0;
const U32 baseCost = 1 << (largestBits - maxNbBits);
U32 n = lastNonNull;
while (huffNode[n].nbBits > maxNbBits) {
totalCost += baseCost - (1 << (largestBits - huffNode[n].nbBits));
huffNode[n].nbBits = (BYTE)maxNbBits;
n--;
} /* n stops at huffNode[n].nbBits <= maxNbBits */
while (huffNode[n].nbBits == maxNbBits)
n--; /* n end at index of smallest symbol using < maxNbBits */
/* renorm totalCost */
totalCost >>= (largestBits - maxNbBits); /* note : totalCost is necessarily a multiple of baseCost */
/* repay normalized cost */
{
U32 const noSymbol = 0xF0F0F0F0;
U32 rankLast[HUF_TABLELOG_MAX + 2];
int pos;
/* Get pos of last (smallest) symbol per rank */
memset(rankLast, 0xF0, sizeof(rankLast));
{
U32 currNbBits = maxNbBits;
for (pos = n; pos >= 0; pos--) {
if (huffNode[pos].nbBits >= currNbBits)
continue;
currNbBits = huffNode[pos].nbBits; /* < maxNbBits */
rankLast[maxNbBits - currNbBits] = pos;
}
}
while (totalCost > 0) {
U32 nBitsToDecrease = BIT_highbit32(totalCost) + 1;
for (; nBitsToDecrease > 1; nBitsToDecrease--) {
U32 highPos = rankLast[nBitsToDecrease];
U32 lowPos = rankLast[nBitsToDecrease - 1];
if (highPos == noSymbol)
continue;
if (lowPos == noSymbol)
break;
{
U32 const highTotal = huffNode[highPos].count;
U32 const lowTotal = 2 * huffNode[lowPos].count;
if (highTotal <= lowTotal)
break;
}
}
/* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
/* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
while ((nBitsToDecrease <= HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
nBitsToDecrease++;
totalCost -= 1 << (nBitsToDecrease - 1);
if (rankLast[nBitsToDecrease - 1] == noSymbol)
rankLast[nBitsToDecrease - 1] = rankLast[nBitsToDecrease]; /* this rank is no longer empty */
huffNode[rankLast[nBitsToDecrease]].nbBits++;
if (rankLast[nBitsToDecrease] == 0) /* special case, reached largest symbol */
rankLast[nBitsToDecrease] = noSymbol;
else {
rankLast[nBitsToDecrease]--;
if (huffNode[rankLast[nBitsToDecrease]].nbBits != maxNbBits - nBitsToDecrease)
rankLast[nBitsToDecrease] = noSymbol; /* this rank is now empty */
}
} /* while (totalCost > 0) */
while (totalCost < 0) { /* Sometimes, cost correction overshoot */
if (rankLast[1] == noSymbol) { /* special case : no rank 1 symbol (using maxNbBits-1); let's create one from largest rank 0
(using maxNbBits) */
while (huffNode[n].nbBits == maxNbBits)
n--;
huffNode[n + 1].nbBits--;
rankLast[1] = n + 1;
totalCost++;
continue;
}
huffNode[rankLast[1] + 1].nbBits--;
rankLast[1]++;
totalCost++;
}
}
} /* there are several too large elements (at least >= 2) */
return maxNbBits;
}
typedef struct {
U32 base;
U32 curr;
} rankPos;
static void HUF_sort(nodeElt *huffNode, const U32 *count, U32 maxSymbolValue)
{
rankPos rank[32];
U32 n;
memset(rank, 0, sizeof(rank));
for (n = 0; n <= maxSymbolValue; n++) {
U32 r = BIT_highbit32(count[n] + 1);
rank[r].base++;
}
for (n = 30; n > 0; n--)
rank[n - 1].base += rank[n].base;
for (n = 0; n < 32; n++)
rank[n].curr = rank[n].base;
for (n = 0; n <= maxSymbolValue; n++) {
U32 const c = count[n];
U32 const r = BIT_highbit32(c + 1) + 1;
U32 pos = rank[r].curr++;
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;
}
}
/** 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.
*/
#define STARTNODE (HUF_SYMBOLVALUE_MAX + 1)
typedef nodeElt huffNodeTable[2 * HUF_SYMBOLVALUE_MAX + 1 + 1];
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;
nodeElt *const huffNode = huffNode0 + 1;
U32 n, nonNullRank;
int lowS, lowN;
U16 nodeNb = STARTNODE;
U32 nodeRoot;
/* safety checks */
if (wkspSize < sizeof(huffNodeTable))
return ERROR(GENERIC); /* workSpace is not large enough */
if (maxNbBits == 0)
maxNbBits = HUF_TABLELOG_DEFAULT;
if (maxSymbolValue > HUF_SYMBOLVALUE_MAX)
return ERROR(GENERIC);
memset(huffNode0, 0, sizeof(huffNodeTable));
/* sort, decreasing order */
HUF_sort(huffNode, count, maxSymbolValue);
/* init for parents */
nonNullRank = maxSymbolValue;
while (huffNode[nonNullRank].count == 0)
nonNullRank--;
lowS = nonNullRank;
nodeRoot = nodeNb + lowS - 1;
lowN = nodeNb;
huffNode[nodeNb].count = huffNode[lowS].count + huffNode[lowS - 1].count;
huffNode[lowS].parent = huffNode[lowS - 1].parent = nodeNb;
nodeNb++;
lowS -= 2;
for (n = nodeNb; n <= nodeRoot; n++)
huffNode[n].count = (U32)(1U << 30);
huffNode0[0].count = (U32)(1U << 31); /* fake entry, strong barrier */
/* create parents */
while (nodeNb <= nodeRoot) {
U32 n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
U32 n2 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
huffNode[nodeNb].count = huffNode[n1].count + huffNode[n2].count;
huffNode[n1].parent = huffNode[n2].parent = nodeNb;
nodeNb++;
}
/* distribute weights (unlimited tree height) */
huffNode[nodeRoot].nbBits = 0;
for (n = nodeRoot - 1; n >= STARTNODE; n--)
huffNode[n].nbBits = huffNode[huffNode[n].parent].nbBits + 1;
for (n = 0; n <= nonNullRank; n++)
huffNode[n].nbBits = huffNode[huffNode[n].parent].nbBits + 1;
/* enforce maxTableLog */
maxNbBits = HUF_setMaxHeight(huffNode, nonNullRank, maxNbBits);
/* fill result into tree (val, nbBits) */
{
U16 nbPerRank[HUF_TABLELOG_MAX + 1] = {0};
U16 valPerRank[HUF_TABLELOG_MAX + 1] = {0};
if (maxNbBits > HUF_TABLELOG_MAX)
return ERROR(GENERIC); /* check fit into table */
for (n = 0; n <= nonNullRank; n++)
nbPerRank[huffNode[n].nbBits]++;
/* determine stating value per rank */
{
U16 min = 0;
for (n = maxNbBits; n > 0; n--) {
valPerRank[n] = min; /* get starting value within each rank */
min += nbPerRank[n];
min >>= 1;
}
}
for (n = 0; n <= maxSymbolValue; n++)
tree[huffNode[n].byte].nbBits = huffNode[n].nbBits; /* push nbBits per symbol, symbol order */
for (n = 0; n <= maxSymbolValue; n++)
tree[n].val = valPerRank[tree[n].nbBits]++; /* assign value within rank, symbol order */
}
return maxNbBits;
}
static size_t HUF_estimateCompressedSize(HUF_CElt *CTable, const unsigned *count, unsigned maxSymbolValue)
{
size_t nbBits = 0;
int s;
for (s = 0; s <= (int)maxSymbolValue; ++s) {
nbBits += CTable[s].nbBits * count[s];
}
return nbBits >> 3;
}
static int HUF_validateCTable(const HUF_CElt *CTable, const unsigned *count, unsigned maxSymbolValue)
{
int bad = 0;
int s;
for (s = 0; s <= (int)maxSymbolValue; ++s) {
bad |= (count[s] != 0) & (CTable[s].nbBits == 0);
}
return !bad;
}
static 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) \
if (sizeof((stream)->bitContainer) * 8 < HUF_TABLELOG_MAX * 2 + 7) \
HUF_FLUSHBITS(stream)
#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)
{
const BYTE *ip = (const BYTE *)src;
BYTE *const ostart = (BYTE *)dst;
BYTE *const oend = ostart + dstSize;
BYTE *op = ostart;
size_t n;
BIT_CStream_t bitC;
/* init */
if (dstSize < 8)
return 0; /* not enough space to compress */
{
size_t const initErr = BIT_initCStream(&bitC, op, oend - op);
if (HUF_isError(initErr))
return 0;
}
n = srcSize & ~3; /* join to mod 4 */
switch (srcSize & 3) {
case 3: HUF_encodeSymbol(&bitC, ip[n + 2], CTable); HUF_FLUSHBITS_2(&bitC);
lib: zstd: Mark expected switch fall-throughs In preparation to enabling -Wimplicit-fallthrough, mark switch cases where we are expecting to fall through. This patch fixes the following warnings: lib/zstd/bitstream.h:261:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:262:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:263:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:264:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:265:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/compress.c:3183:16: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:1770:18: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2376:15: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2404:15: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2435:16: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/huf_compress.c: In function ‘HUF_compress1X_usingCTable’: lib/zstd/huf_compress.c:535:5: warning: this statement may fall through [-Wimplicit-fallthrough=] if (sizeof((stream)->bitContainer) * 8 < HUF_TABLELOG_MAX * 4 + 7) \ ^ lib/zstd/huf_compress.c:558:54: note: in expansion of macro ‘HUF_FLUSHBITS_2’ case 3: HUF_encodeSymbol(&bitC, ip[n + 2], CTable); HUF_FLUSHBITS_2(&bitC); ^~~~~~~~~~~~~~~ lib/zstd/huf_compress.c:559:2: note: here case 2: HUF_encodeSymbol(&bitC, ip[n + 1], CTable); HUF_FLUSHBITS_1(&bitC); ^~~~ lib/zstd/huf_compress.c:531:5: warning: this statement may fall through [-Wimplicit-fallthrough=] if (sizeof((stream)->bitContainer) * 8 < HUF_TABLELOG_MAX * 2 + 7) \ ^ lib/zstd/huf_compress.c:559:54: note: in expansion of macro ‘HUF_FLUSHBITS_1’ case 2: HUF_encodeSymbol(&bitC, ip[n + 1], CTable); HUF_FLUSHBITS_1(&bitC); ^~~~~~~~~~~~~~~ lib/zstd/huf_compress.c:560:2: note: here case 1: HUF_encodeSymbol(&bitC, ip[n + 0], CTable); HUF_FLUSHBITS(&bitC); ^~~~ AR lib/zstd//built-in.a Warning level 3 was used: -Wimplicit-fallthrough=3 This patch is part of the ongoing efforts to enabling -Wimplicit-fallthrough. Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
2019-01-25 06:37:43 +03:00
/* fall through */
lib: Add zstd modules Add zstd compression and decompression kernel modules. zstd offers a wide varity of compression speed and quality trade-offs. It can compress at speeds approaching lz4, and quality approaching lzma. zstd decompressions at speeds more than twice as fast as zlib, and decompression speed remains roughly the same across all compression levels. The code was ported from the upstream zstd source repository. The `linux/zstd.h` header was modified to match linux kernel style. The cross-platform and allocation code was stripped out. Instead zstd requires the caller to pass a preallocated workspace. The source files were clang-formatted [1] to match the Linux Kernel style as much as possible. Otherwise, the code was unmodified. We would like to avoid as much further manual modification to the source code as possible, so it will be easier to keep the kernel zstd up to date. I benchmarked zstd compression as a special character device. I ran zstd and zlib compression at several levels, as well as performing no compression, which measure the time spent copying the data to kernel space. Data is passed to the compresser 4096 B at a time. The benchmark file is located in the upstream zstd source repository under `contrib/linux-kernel/zstd_compress_test.c` [2]. I ran the benchmarks on a Ubuntu 14.04 VM with 2 cores and 4 GiB of RAM. The VM is running on a MacBook Pro with a 3.1 GHz Intel Core i7 processor, 16 GB of RAM, and a SSD. I benchmarked using `silesia.tar` [3], which is 211,988,480 B large. Run the following commands for the benchmark: sudo modprobe zstd_compress_test sudo mknod zstd_compress_test c 245 0 sudo cp silesia.tar zstd_compress_test The time is reported by the time of the userland `cp`. The MB/s is computed with 1,536,217,008 B / time(buffer size, hash) which includes the time to copy from userland. The Adjusted MB/s is computed with 1,536,217,088 B / (time(buffer size, hash) - time(buffer size, none)). The memory reported is the amount of memory the compressor requests. | Method | Size (B) | Time (s) | Ratio | MB/s | Adj MB/s | Mem (MB) | |----------|----------|----------|-------|---------|----------|----------| | none | 11988480 | 0.100 | 1 | 2119.88 | - | - | | zstd -1 | 73645762 | 1.044 | 2.878 | 203.05 | 224.56 | 1.23 | | zstd -3 | 66988878 | 1.761 | 3.165 | 120.38 | 127.63 | 2.47 | | zstd -5 | 65001259 | 2.563 | 3.261 | 82.71 | 86.07 | 2.86 | | zstd -10 | 60165346 | 13.242 | 3.523 | 16.01 | 16.13 | 13.22 | | zstd -15 | 58009756 | 47.601 | 3.654 | 4.45 | 4.46 | 21.61 | | zstd -19 | 54014593 | 102.835 | 3.925 | 2.06 | 2.06 | 60.15 | | zlib -1 | 77260026 | 2.895 | 2.744 | 73.23 | 75.85 | 0.27 | | zlib -3 | 72972206 | 4.116 | 2.905 | 51.50 | 52.79 | 0.27 | | zlib -6 | 68190360 | 9.633 | 3.109 | 22.01 | 22.24 | 0.27 | | zlib -9 | 67613382 | 22.554 | 3.135 | 9.40 | 9.44 | 0.27 | I benchmarked zstd decompression using the same method on the same machine. The benchmark file is located in the upstream zstd repo under `contrib/linux-kernel/zstd_decompress_test.c` [4]. The memory reported is the amount of memory required to decompress data compressed with the given compression level. If you know the maximum size of your input, you can reduce the memory usage of decompression irrespective of the compression level. | Method | Time (s) | MB/s | Adjusted MB/s | Memory (MB) | |----------|----------|---------|---------------|-------------| | none | 0.025 | 8479.54 | - | - | | zstd -1 | 0.358 | 592.15 | 636.60 | 0.84 | | zstd -3 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -5 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -10 | 0.374 | 566.81 | 607.42 | 2.51 | | zstd -15 | 0.379 | 559.34 | 598.84 | 4.61 | | zstd -19 | 0.412 | 514.54 | 547.77 | 8.80 | | zlib -1 | 0.940 | 225.52 | 231.68 | 0.04 | | zlib -3 | 0.883 | 240.08 | 247.07 | 0.04 | | zlib -6 | 0.844 | 251.17 | 258.84 | 0.04 | | zlib -9 | 0.837 | 253.27 | 287.64 | 0.04 | Tested in userland using the test-suite in the zstd repo under `contrib/linux-kernel/test/UserlandTest.cpp` [5] by mocking the kernel functions. Fuzz tested using libfuzzer [6] with the fuzz harnesses under `contrib/linux-kernel/test/{RoundTripCrash.c,DecompressCrash.c}` [7] [8] with ASAN, UBSAN, and MSAN. Additionaly, it was tested while testing the BtrFS and SquashFS patches coming next. [1] https://clang.llvm.org/docs/ClangFormat.html [2] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_compress_test.c [3] http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia [4] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_decompress_test.c [5] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/UserlandTest.cpp [6] http://llvm.org/docs/LibFuzzer.html [7] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/RoundTripCrash.c [8] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/DecompressCrash.c zstd source repository: https://github.com/facebook/zstd Signed-off-by: Nick Terrell <terrelln@fb.com> Signed-off-by: Chris Mason <clm@fb.com>
2017-08-10 05:35:53 +03:00
case 2: HUF_encodeSymbol(&bitC, ip[n + 1], CTable); HUF_FLUSHBITS_1(&bitC);
lib: zstd: Mark expected switch fall-throughs In preparation to enabling -Wimplicit-fallthrough, mark switch cases where we are expecting to fall through. This patch fixes the following warnings: lib/zstd/bitstream.h:261:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:262:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:263:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:264:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/bitstream.h:265:30: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/compress.c:3183:16: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:1770:18: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2376:15: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2404:15: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/decompress.c:2435:16: warning: this statement may fall through [-Wimplicit-fallthrough=] lib/zstd/huf_compress.c: In function ‘HUF_compress1X_usingCTable’: lib/zstd/huf_compress.c:535:5: warning: this statement may fall through [-Wimplicit-fallthrough=] if (sizeof((stream)->bitContainer) * 8 < HUF_TABLELOG_MAX * 4 + 7) \ ^ lib/zstd/huf_compress.c:558:54: note: in expansion of macro ‘HUF_FLUSHBITS_2’ case 3: HUF_encodeSymbol(&bitC, ip[n + 2], CTable); HUF_FLUSHBITS_2(&bitC); ^~~~~~~~~~~~~~~ lib/zstd/huf_compress.c:559:2: note: here case 2: HUF_encodeSymbol(&bitC, ip[n + 1], CTable); HUF_FLUSHBITS_1(&bitC); ^~~~ lib/zstd/huf_compress.c:531:5: warning: this statement may fall through [-Wimplicit-fallthrough=] if (sizeof((stream)->bitContainer) * 8 < HUF_TABLELOG_MAX * 2 + 7) \ ^ lib/zstd/huf_compress.c:559:54: note: in expansion of macro ‘HUF_FLUSHBITS_1’ case 2: HUF_encodeSymbol(&bitC, ip[n + 1], CTable); HUF_FLUSHBITS_1(&bitC); ^~~~~~~~~~~~~~~ lib/zstd/huf_compress.c:560:2: note: here case 1: HUF_encodeSymbol(&bitC, ip[n + 0], CTable); HUF_FLUSHBITS(&bitC); ^~~~ AR lib/zstd//built-in.a Warning level 3 was used: -Wimplicit-fallthrough=3 This patch is part of the ongoing efforts to enabling -Wimplicit-fallthrough. Reviewed-by: Kees Cook <keescook@chromium.org> Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
2019-01-25 06:37:43 +03:00
/* fall through */
lib: Add zstd modules Add zstd compression and decompression kernel modules. zstd offers a wide varity of compression speed and quality trade-offs. It can compress at speeds approaching lz4, and quality approaching lzma. zstd decompressions at speeds more than twice as fast as zlib, and decompression speed remains roughly the same across all compression levels. The code was ported from the upstream zstd source repository. The `linux/zstd.h` header was modified to match linux kernel style. The cross-platform and allocation code was stripped out. Instead zstd requires the caller to pass a preallocated workspace. The source files were clang-formatted [1] to match the Linux Kernel style as much as possible. Otherwise, the code was unmodified. We would like to avoid as much further manual modification to the source code as possible, so it will be easier to keep the kernel zstd up to date. I benchmarked zstd compression as a special character device. I ran zstd and zlib compression at several levels, as well as performing no compression, which measure the time spent copying the data to kernel space. Data is passed to the compresser 4096 B at a time. The benchmark file is located in the upstream zstd source repository under `contrib/linux-kernel/zstd_compress_test.c` [2]. I ran the benchmarks on a Ubuntu 14.04 VM with 2 cores and 4 GiB of RAM. The VM is running on a MacBook Pro with a 3.1 GHz Intel Core i7 processor, 16 GB of RAM, and a SSD. I benchmarked using `silesia.tar` [3], which is 211,988,480 B large. Run the following commands for the benchmark: sudo modprobe zstd_compress_test sudo mknod zstd_compress_test c 245 0 sudo cp silesia.tar zstd_compress_test The time is reported by the time of the userland `cp`. The MB/s is computed with 1,536,217,008 B / time(buffer size, hash) which includes the time to copy from userland. The Adjusted MB/s is computed with 1,536,217,088 B / (time(buffer size, hash) - time(buffer size, none)). The memory reported is the amount of memory the compressor requests. | Method | Size (B) | Time (s) | Ratio | MB/s | Adj MB/s | Mem (MB) | |----------|----------|----------|-------|---------|----------|----------| | none | 11988480 | 0.100 | 1 | 2119.88 | - | - | | zstd -1 | 73645762 | 1.044 | 2.878 | 203.05 | 224.56 | 1.23 | | zstd -3 | 66988878 | 1.761 | 3.165 | 120.38 | 127.63 | 2.47 | | zstd -5 | 65001259 | 2.563 | 3.261 | 82.71 | 86.07 | 2.86 | | zstd -10 | 60165346 | 13.242 | 3.523 | 16.01 | 16.13 | 13.22 | | zstd -15 | 58009756 | 47.601 | 3.654 | 4.45 | 4.46 | 21.61 | | zstd -19 | 54014593 | 102.835 | 3.925 | 2.06 | 2.06 | 60.15 | | zlib -1 | 77260026 | 2.895 | 2.744 | 73.23 | 75.85 | 0.27 | | zlib -3 | 72972206 | 4.116 | 2.905 | 51.50 | 52.79 | 0.27 | | zlib -6 | 68190360 | 9.633 | 3.109 | 22.01 | 22.24 | 0.27 | | zlib -9 | 67613382 | 22.554 | 3.135 | 9.40 | 9.44 | 0.27 | I benchmarked zstd decompression using the same method on the same machine. The benchmark file is located in the upstream zstd repo under `contrib/linux-kernel/zstd_decompress_test.c` [4]. The memory reported is the amount of memory required to decompress data compressed with the given compression level. If you know the maximum size of your input, you can reduce the memory usage of decompression irrespective of the compression level. | Method | Time (s) | MB/s | Adjusted MB/s | Memory (MB) | |----------|----------|---------|---------------|-------------| | none | 0.025 | 8479.54 | - | - | | zstd -1 | 0.358 | 592.15 | 636.60 | 0.84 | | zstd -3 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -5 | 0.396 | 535.32 | 571.40 | 1.46 | | zstd -10 | 0.374 | 566.81 | 607.42 | 2.51 | | zstd -15 | 0.379 | 559.34 | 598.84 | 4.61 | | zstd -19 | 0.412 | 514.54 | 547.77 | 8.80 | | zlib -1 | 0.940 | 225.52 | 231.68 | 0.04 | | zlib -3 | 0.883 | 240.08 | 247.07 | 0.04 | | zlib -6 | 0.844 | 251.17 | 258.84 | 0.04 | | zlib -9 | 0.837 | 253.27 | 287.64 | 0.04 | Tested in userland using the test-suite in the zstd repo under `contrib/linux-kernel/test/UserlandTest.cpp` [5] by mocking the kernel functions. Fuzz tested using libfuzzer [6] with the fuzz harnesses under `contrib/linux-kernel/test/{RoundTripCrash.c,DecompressCrash.c}` [7] [8] with ASAN, UBSAN, and MSAN. Additionaly, it was tested while testing the BtrFS and SquashFS patches coming next. [1] https://clang.llvm.org/docs/ClangFormat.html [2] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_compress_test.c [3] http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia [4] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/zstd_decompress_test.c [5] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/UserlandTest.cpp [6] http://llvm.org/docs/LibFuzzer.html [7] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/RoundTripCrash.c [8] https://github.com/facebook/zstd/blob/dev/contrib/linux-kernel/test/DecompressCrash.c zstd source repository: https://github.com/facebook/zstd Signed-off-by: Nick Terrell <terrelln@fb.com> Signed-off-by: Chris Mason <clm@fb.com>
2017-08-10 05:35:53 +03:00
case 1: HUF_encodeSymbol(&bitC, ip[n + 0], CTable); HUF_FLUSHBITS(&bitC);
case 0:
default:;
}
for (; n > 0; n -= 4) { /* note : n&3==0 at this stage */
HUF_encodeSymbol(&bitC, ip[n - 1], CTable);
HUF_FLUSHBITS_1(&bitC);
HUF_encodeSymbol(&bitC, ip[n - 2], CTable);
HUF_FLUSHBITS_2(&bitC);
HUF_encodeSymbol(&bitC, ip[n - 3], CTable);
HUF_FLUSHBITS_1(&bitC);
HUF_encodeSymbol(&bitC, ip[n - 4], CTable);
HUF_FLUSHBITS(&bitC);
}
return BIT_closeCStream(&bitC);
}
size_t HUF_compress4X_usingCTable(void *dst, size_t dstSize, const void *src, size_t srcSize, const HUF_CElt *CTable)
{
size_t const segmentSize = (srcSize + 3) / 4; /* first 3 segments */
const BYTE *ip = (const BYTE *)src;
const BYTE *const iend = ip + srcSize;
BYTE *const ostart = (BYTE *)dst;
BYTE *const oend = ostart + dstSize;
BYTE *op = ostart;
if (dstSize < 6 + 1 + 1 + 1 + 8)
return 0; /* minimum space to compress successfully */
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));
if (cSize == 0)
return 0;
ZSTD_writeLE16(ostart, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{
CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend - op, ip, segmentSize, CTable));
if (cSize == 0)
return 0;
ZSTD_writeLE16(ostart + 2, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{
CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend - op, ip, segmentSize, CTable));
if (cSize == 0)
return 0;
ZSTD_writeLE16(ostart + 4, (U16)cSize);
op += cSize;
}
ip += segmentSize;
{
CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend - op, ip, iend - ip, CTable));
if (cSize == 0)
return 0;
op += cSize;
}
return op - ostart;
}
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)
{
size_t const cSize =
singleStream ? HUF_compress1X_usingCTable(op, oend - op, src, srcSize, CTable) : HUF_compress4X_usingCTable(op, oend - op, src, srcSize, CTable);
if (HUF_isError(cSize)) {
return cSize;
}
if (cSize == 0) {
return 0;
} /* uncompressible */
op += cSize;
/* check compressibility */
if ((size_t)(op - ostart) >= srcSize - 1) {
return 0;
}
return op - ostart;
}
/* `workSpace` must a table of at least 1024 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)
{
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 (srcSize > HUF_BLOCKSIZE_MAX)
return ERROR(srcSize_wrong); /* curr block size limit */
if (huffLog > HUF_TABLELOG_MAX)
return ERROR(tableLog_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 */
if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
}
/* Scan input and build symbol stats */
{
CHECK_V_F(largest, FSE_count_wksp(count, &maxSymbolValue, (const BYTE *)src, srcSize, (U32 *)workSpace));
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 */
}
/* Check validity of previous table */
if (repeat && *repeat == HUF_repeat_check && !HUF_validateCTable(oldHufTable, 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);
}
/* Build Huffman Tree */
huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
{
CHECK_V_F(maxBits, HUF_buildCTable_wksp(CTable, count, maxSymbolValue, huffLog, workSpace, wkspSize));
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));
}
/* Write table description header */
{
CHECK_V_F(hSize, HUF_writeCTable_wksp(op, dstSize, CTable, maxSymbolValue, huffLog, workSpace, wkspSize));
/* Check if using the previous table will be 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);
if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
}
}
/* Use the new 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 */
}
return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, CTable);
}
size_t HUF_compress1X_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, 1 /* single stream */, workSpace, wkspSize, NULL, NULL, 0);
}
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)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, hufTable, repeat,
preferRepeat);
}
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);
}
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)
{
return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, hufTable, repeat,
preferRepeat);
}