Sereal-Decoder
view release on metacpan or search on metacpan
zstd/decompress/zstd_decompress_block.c view on Meta::CPAN
/*-*******************************************************
* Dependencies
*********************************************************/
#include "../common/zstd_deps.h" /* ZSTD_memcpy, ZSTD_memmove, ZSTD_memset */
#include "../common/compiler.h" /* prefetch */
#include "../common/cpu.h" /* bmi2 */
#include "../common/mem.h" /* low level memory routines */
#define FSE_STATIC_LINKING_ONLY
#include "../common/fse.h"
#define HUF_STATIC_LINKING_ONLY
#include "../common/huf.h"
#include "../common/zstd_internal.h"
#include "zstd_decompress_internal.h" /* ZSTD_DCtx */
#include "zstd_ddict.h" /* ZSTD_DDictDictContent */
#include "zstd_decompress_block.h"
/*_*******************************************************
* Macros
**********************************************************/
/* These two optional macros force the use one way or another of the two
* ZSTD_decompressSequences implementations. You can't force in both directions
* at the same time.
*/
#if defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
#error "Cannot force the use of the short and the long ZSTD_decompressSequences variants!"
#endif
/*_*******************************************************
* Memory operations
**********************************************************/
static void ZSTD_copy4(void* dst, const void* src) { ZSTD_memcpy(dst, src, 4); }
/*-*************************************************************
* Block decoding
***************************************************************/
/*! ZSTD_getcBlockSize() :
* Provides the size of compressed block from block header `src` */
size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
blockProperties_t* bpPtr)
{
RETURN_ERROR_IF(srcSize < ZSTD_blockHeaderSize, srcSize_wrong, "");
{ U32 const cBlockHeader = MEM_readLE24(src);
U32 const cSize = cBlockHeader >> 3;
bpPtr->lastBlock = cBlockHeader & 1;
bpPtr->blockType = (blockType_e)((cBlockHeader >> 1) & 3);
bpPtr->origSize = cSize; /* only useful for RLE */
if (bpPtr->blockType == bt_rle) return 1;
RETURN_ERROR_IF(bpPtr->blockType == bt_reserved, corruption_detected, "");
return cSize;
}
}
/* Allocate buffer for literals, either overlapping current dst, or split between dst and litExtraBuffer, or stored entirely within litExtraBuffer */
static void ZSTD_allocateLiteralsBuffer(ZSTD_DCtx* dctx, void* const dst, const size_t dstCapacity, const size_t litSize,
const streaming_operation streaming, const size_t expectedWriteSize, const unsigned splitImmediately)
{
if (streaming == not_streaming && dstCapacity > ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH + litSize + WILDCOPY_OVERLENGTH)
{
/* room for litbuffer to fit without read faulting */
dctx->litBuffer = (BYTE*)dst + ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH;
dctx->litBufferEnd = dctx->litBuffer + litSize;
dctx->litBufferLocation = ZSTD_in_dst;
}
else if (litSize > ZSTD_LITBUFFEREXTRASIZE)
{
/* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */
if (splitImmediately) {
/* won't fit in litExtraBuffer, so it will be split between end of dst and extra buffer */
dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
dctx->litBufferEnd = dctx->litBuffer + litSize - ZSTD_LITBUFFEREXTRASIZE;
}
else {
/* initially this will be stored entirely in dst during huffman decoding, it will partially shifted to litExtraBuffer after */
dctx->litBuffer = (BYTE*)dst + expectedWriteSize - litSize;
dctx->litBufferEnd = (BYTE*)dst + expectedWriteSize;
}
dctx->litBufferLocation = ZSTD_split;
}
else
{
/* fits entirely within litExtraBuffer, so no split is necessary */
dctx->litBuffer = dctx->litExtraBuffer;
dctx->litBufferEnd = dctx->litBuffer + litSize;
dctx->litBufferLocation = ZSTD_not_in_dst;
}
}
/* Hidden declaration for fullbench */
size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
const void* src, size_t srcSize,
void* dst, size_t dstCapacity, const streaming_operation streaming);
/*! ZSTD_decodeLiteralsBlock() :
* Where it is possible to do so without being stomped by the output during decompression, the literals block will be stored
* in the dstBuffer. If there is room to do so, it will be stored in full in the excess dst space after where the current
* block will be output. Otherwise it will be stored at the end of the current dst blockspace, with a small portion being
* stored in dctx->litExtraBuffer to help keep it "ahead" of the current output write.
*
* @return : nb of bytes read from src (< srcSize )
* note : symbol not declared but exposed for fullbench */
size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
const void* src, size_t srcSize, /* note : srcSize < BLOCKSIZE */
void* dst, size_t dstCapacity, const streaming_operation streaming)
{
DEBUGLOG(5, "ZSTD_decodeLiteralsBlock");
RETURN_ERROR_IF(srcSize < MIN_CBLOCK_SIZE, corruption_detected, "");
{ const BYTE* const istart = (const BYTE*) src;
symbolEncodingType_e const litEncType = (symbolEncodingType_e)(istart[0] & 3);
switch(litEncType)
{
case set_repeat:
DEBUGLOG(5, "set_repeat flag : re-using stats from previous compressed literals block");
RETURN_ERROR_IF(dctx->litEntropy==0, dictionary_corrupted, "");
ZSTD_FALLTHROUGH;
case set_compressed:
RETURN_ERROR_IF(srcSize < 5, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 3; here we need up to 5 for case 3");
{ size_t lhSize, litSize, litCSize;
U32 singleStream=0;
U32 const lhlCode = (istart[0] >> 2) & 3;
U32 const lhc = MEM_readLE32(istart);
size_t hufSuccess;
size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity);
switch(lhlCode)
{
case 0: case 1: default: /* note : default is impossible, since lhlCode into [0..3] */
/* 2 - 2 - 10 - 10 */
singleStream = !lhlCode;
lhSize = 3;
litSize = (lhc >> 4) & 0x3FF;
litCSize = (lhc >> 14) & 0x3FF;
break;
case 2:
/* 2 - 2 - 14 - 14 */
lhSize = 4;
litSize = (lhc >> 4) & 0x3FFF;
litCSize = lhc >> 18;
break;
case 3:
/* 2 - 2 - 18 - 18 */
lhSize = 5;
litSize = (lhc >> 4) & 0x3FFFF;
litCSize = (lhc >> 22) + ((size_t)istart[4] << 10);
break;
}
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, "");
RETURN_ERROR_IF(litCSize + lhSize > srcSize, corruption_detected, "");
RETURN_ERROR_IF(expectedWriteSize < litSize , dstSize_tooSmall, "");
ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 0);
/* prefetch huffman table if cold */
if (dctx->ddictIsCold && (litSize > 768 /* heuristic */)) {
PREFETCH_AREA(dctx->HUFptr, sizeof(dctx->entropy.hufTable));
}
if (litEncType==set_repeat) {
if (singleStream) {
hufSuccess = HUF_decompress1X_usingDTable_bmi2(
dctx->litBuffer, litSize, istart+lhSize, litCSize,
dctx->HUFptr, ZSTD_DCtx_get_bmi2(dctx));
} else {
hufSuccess = HUF_decompress4X_usingDTable_bmi2(
dctx->litBuffer, litSize, istart+lhSize, litCSize,
dctx->HUFptr, ZSTD_DCtx_get_bmi2(dctx));
}
} else {
if (singleStream) {
#if defined(HUF_FORCE_DECOMPRESS_X2)
hufSuccess = HUF_decompress1X_DCtx_wksp(
dctx->entropy.hufTable, dctx->litBuffer, litSize,
istart+lhSize, litCSize, dctx->workspace,
sizeof(dctx->workspace));
#else
hufSuccess = HUF_decompress1X1_DCtx_wksp_bmi2(
dctx->entropy.hufTable, dctx->litBuffer, litSize,
istart+lhSize, litCSize, dctx->workspace,
sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx));
#endif
} else {
hufSuccess = HUF_decompress4X_hufOnly_wksp_bmi2(
dctx->entropy.hufTable, dctx->litBuffer, litSize,
istart+lhSize, litCSize, dctx->workspace,
sizeof(dctx->workspace), ZSTD_DCtx_get_bmi2(dctx));
}
}
if (dctx->litBufferLocation == ZSTD_split)
{
ZSTD_memcpy(dctx->litExtraBuffer, dctx->litBufferEnd - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE);
ZSTD_memmove(dctx->litBuffer + ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH, dctx->litBuffer, litSize - ZSTD_LITBUFFEREXTRASIZE);
dctx->litBuffer += ZSTD_LITBUFFEREXTRASIZE - WILDCOPY_OVERLENGTH;
dctx->litBufferEnd -= WILDCOPY_OVERLENGTH;
}
RETURN_ERROR_IF(HUF_isError(hufSuccess), corruption_detected, "");
dctx->litPtr = dctx->litBuffer;
dctx->litSize = litSize;
dctx->litEntropy = 1;
if (litEncType==set_compressed) dctx->HUFptr = dctx->entropy.hufTable;
return litCSize + lhSize;
}
case set_basic:
{ size_t litSize, lhSize;
U32 const lhlCode = ((istart[0]) >> 2) & 3;
size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity);
switch(lhlCode)
{
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
lhSize = 1;
litSize = istart[0] >> 3;
break;
case 1:
lhSize = 2;
litSize = MEM_readLE16(istart) >> 4;
break;
case 3:
lhSize = 3;
litSize = MEM_readLE24(istart) >> 4;
break;
}
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
if (lhSize+litSize+WILDCOPY_OVERLENGTH > srcSize) { /* risk reading beyond src buffer with wildcopy */
RETURN_ERROR_IF(litSize+lhSize > srcSize, corruption_detected, "");
if (dctx->litBufferLocation == ZSTD_split)
{
ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize - ZSTD_LITBUFFEREXTRASIZE);
ZSTD_memcpy(dctx->litExtraBuffer, istart + lhSize + litSize - ZSTD_LITBUFFEREXTRASIZE, ZSTD_LITBUFFEREXTRASIZE);
}
else
{
ZSTD_memcpy(dctx->litBuffer, istart + lhSize, litSize);
}
dctx->litPtr = dctx->litBuffer;
dctx->litSize = litSize;
return lhSize+litSize;
}
/* direct reference into compressed stream */
dctx->litPtr = istart+lhSize;
dctx->litSize = litSize;
dctx->litBufferEnd = dctx->litPtr + litSize;
dctx->litBufferLocation = ZSTD_not_in_dst;
return lhSize+litSize;
}
case set_rle:
{ U32 const lhlCode = ((istart[0]) >> 2) & 3;
size_t litSize, lhSize;
size_t expectedWriteSize = MIN(ZSTD_BLOCKSIZE_MAX, dstCapacity);
switch(lhlCode)
{
case 0: case 2: default: /* note : default is impossible, since lhlCode into [0..3] */
lhSize = 1;
litSize = istart[0] >> 3;
break;
case 1:
lhSize = 2;
litSize = MEM_readLE16(istart) >> 4;
break;
case 3:
lhSize = 3;
litSize = MEM_readLE24(istart) >> 4;
RETURN_ERROR_IF(srcSize<4, corruption_detected, "srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4");
break;
}
RETURN_ERROR_IF(litSize > 0 && dst == NULL, dstSize_tooSmall, "NULL not handled");
RETURN_ERROR_IF(litSize > ZSTD_BLOCKSIZE_MAX, corruption_detected, "");
RETURN_ERROR_IF(expectedWriteSize < litSize, dstSize_tooSmall, "");
ZSTD_allocateLiteralsBuffer(dctx, dst, dstCapacity, litSize, streaming, expectedWriteSize, 1);
if (dctx->litBufferLocation == ZSTD_split)
{
ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize - ZSTD_LITBUFFEREXTRASIZE);
ZSTD_memset(dctx->litExtraBuffer, istart[lhSize], ZSTD_LITBUFFEREXTRASIZE);
}
else
{
ZSTD_memset(dctx->litBuffer, istart[lhSize], litSize);
}
dctx->litPtr = dctx->litBuffer;
dctx->litSize = litSize;
return lhSize+1;
}
default:
RETURN_ERROR(corruption_detected, "impossible");
}
}
}
/* Default FSE distribution tables.
* These are pre-calculated FSE decoding tables using default distributions as defined in specification :
* https://github.com/facebook/zstd/blob/release/doc/zstd_compression_format.md#default-distributions
* They were generated programmatically with following method :
* - start from default distributions, present in /lib/common/zstd_internal.h
* - generate tables normally, using ZSTD_buildFSETable()
* - printout the content of tables
* - pretify output, report below, test with fuzzer to ensure it's correct */
/* Default FSE distribution table for Literal Lengths */
static const ZSTD_seqSymbol LL_defaultDTable[(1<<LL_DEFAULTNORMLOG)+1] = {
{ 1, 1, 1, LL_DEFAULTNORMLOG}, /* header : fastMode, tableLog */
/* nextState, nbAddBits, nbBits, baseVal */
{ 0, 0, 4, 0}, { 16, 0, 4, 0},
{ 32, 0, 5, 1}, { 0, 0, 5, 3},
{ 0, 0, 5, 4}, { 0, 0, 5, 6},
{ 0, 0, 5, 7}, { 0, 0, 5, 9},
{ 0, 0, 5, 10}, { 0, 0, 5, 12},
{ 0, 0, 6, 14}, { 0, 1, 5, 16},
{ 0, 1, 5, 20}, { 0, 1, 5, 22},
{ 0, 2, 5, 28}, { 0, 3, 5, 32},
{ 0, 4, 5, 48}, { 32, 6, 5, 64},
{ 0, 7, 5, 128}, { 0, 8, 6, 256},
{ 0, 10, 6, 1024}, { 0, 12, 6, 4096},
{ 32, 0, 4, 0}, { 0, 0, 4, 1},
{ 0, 0, 5, 2}, { 32, 0, 5, 4},
{ 0, 0, 5, 5}, { 32, 0, 5, 7},
{ 0, 0, 5, 8}, { 32, 0, 5, 10},
{ 0, 0, 5, 11}, { 0, 0, 6, 13},
{ 32, 1, 5, 16}, { 0, 1, 5, 18},
{ 32, 1, 5, 22}, { 0, 2, 5, 24},
{ 32, 3, 5, 32}, { 0, 3, 5, 40},
{ 0, 6, 4, 64}, { 16, 6, 4, 64},
{ 32, 7, 5, 128}, { 0, 9, 6, 512},
{ 0, 11, 6, 2048}, { 48, 0, 4, 0},
{ 16, 0, 4, 1}, { 32, 0, 5, 2},
{ 32, 0, 5, 3}, { 32, 0, 5, 5},
{ 32, 0, 5, 6}, { 32, 0, 5, 8},
{ 32, 0, 5, 9}, { 32, 0, 5, 11},
{ 32, 0, 5, 12}, { 0, 0, 6, 15},
{ 32, 1, 5, 18}, { 32, 1, 5, 20},
zstd/decompress/zstd_decompress_block.c view on Meta::CPAN
}
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG */
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
/* ZSTD_decompressSequencesLong() :
* decompression function triggered when a minimum share of offsets is considered "long",
* aka out of cache.
* note : "long" definition seems overloaded here, sometimes meaning "wider than bitstream register", and sometimes meaning "farther than memory cache distance".
* This function will try to mitigate main memory latency through the use of prefetching */
static size_t
ZSTD_decompressSequencesLong(ZSTD_DCtx* dctx,
void* dst, size_t maxDstSize,
const void* seqStart, size_t seqSize, int nbSeq,
const ZSTD_longOffset_e isLongOffset,
const int frame)
{
DEBUGLOG(5, "ZSTD_decompressSequencesLong");
#if DYNAMIC_BMI2
if (ZSTD_DCtx_get_bmi2(dctx)) {
return ZSTD_decompressSequencesLong_bmi2(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
}
#endif
return ZSTD_decompressSequencesLong_default(dctx, dst, maxDstSize, seqStart, seqSize, nbSeq, isLongOffset, frame);
}
#endif /* ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT */
#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
!defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
/* ZSTD_getLongOffsetsShare() :
* condition : offTable must be valid
* @return : "share" of long offsets (arbitrarily defined as > (1<<23))
* compared to maximum possible of (1<<OffFSELog) */
static unsigned
ZSTD_getLongOffsetsShare(const ZSTD_seqSymbol* offTable)
{
const void* ptr = offTable;
U32 const tableLog = ((const ZSTD_seqSymbol_header*)ptr)[0].tableLog;
const ZSTD_seqSymbol* table = offTable + 1;
U32 const max = 1 << tableLog;
U32 u, total = 0;
DEBUGLOG(5, "ZSTD_getLongOffsetsShare: (tableLog=%u)", tableLog);
assert(max <= (1 << OffFSELog)); /* max not too large */
for (u=0; u<max; u++) {
if (table[u].nbAdditionalBits > 22) total += 1;
}
assert(tableLog <= OffFSELog);
total <<= (OffFSELog - tableLog); /* scale to OffFSELog */
return total;
}
#endif
size_t
ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize, const int frame, const streaming_operation streaming)
{ /* blockType == blockCompressed */
const BYTE* ip = (const BYTE*)src;
/* isLongOffset must be true if there are long offsets.
* Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN.
* We don't expect that to be the case in 64-bit mode.
* In block mode, window size is not known, so we have to be conservative.
* (note: but it could be evaluated from current-lowLimit)
*/
ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || (dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN))));
DEBUGLOG(5, "ZSTD_decompressBlock_internal (size : %u)", (U32)srcSize);
RETURN_ERROR_IF(srcSize >= ZSTD_BLOCKSIZE_MAX, srcSize_wrong, "");
/* Decode literals section */
{ size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize, dst, dstCapacity, streaming);
DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize);
if (ZSTD_isError(litCSize)) return litCSize;
ip += litCSize;
srcSize -= litCSize;
}
/* Build Decoding Tables */
{
/* These macros control at build-time which decompressor implementation
* we use. If neither is defined, we do some inspection and dispatch at
* runtime.
*/
#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
!defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
int usePrefetchDecoder = dctx->ddictIsCold;
#endif
int nbSeq;
size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, srcSize);
if (ZSTD_isError(seqHSize)) return seqHSize;
ip += seqHSize;
srcSize -= seqHSize;
RETURN_ERROR_IF(dst == NULL && nbSeq > 0, dstSize_tooSmall, "NULL not handled");
#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
!defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
if ( !usePrefetchDecoder
&& (!frame || (dctx->fParams.windowSize > (1<<24)))
&& (nbSeq>ADVANCED_SEQS) ) { /* could probably use a larger nbSeq limit */
U32 const shareLongOffsets = ZSTD_getLongOffsetsShare(dctx->OFTptr);
U32 const minShare = MEM_64bits() ? 7 : 20; /* heuristic values, correspond to 2.73% and 7.81% */
usePrefetchDecoder = (shareLongOffsets >= minShare);
}
#endif
dctx->ddictIsCold = 0;
#if !defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT) && \
!defined(ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG)
if (usePrefetchDecoder)
#endif
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT
return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame);
#endif
#ifndef ZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG
/* else */
if (dctx->litBufferLocation == ZSTD_split)
return ZSTD_decompressSequencesSplitLitBuffer(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame);
else
return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, nbSeq, isLongOffset, frame);
#endif
}
}
void ZSTD_checkContinuity(ZSTD_DCtx* dctx, const void* dst, size_t dstSize)
{
if (dst != dctx->previousDstEnd && dstSize > 0) { /* not contiguous */
dctx->dictEnd = dctx->previousDstEnd;
dctx->virtualStart = (const char*)dst - ((const char*)(dctx->previousDstEnd) - (const char*)(dctx->prefixStart));
dctx->prefixStart = dst;
dctx->previousDstEnd = dst;
}
}
size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
void* dst, size_t dstCapacity,
const void* src, size_t srcSize)
{
size_t dSize;
ZSTD_checkContinuity(dctx, dst, dstCapacity);
dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0, not_streaming);
dctx->previousDstEnd = (char*)dst + dSize;
return dSize;
}
( run in 1.416 second using v1.01-cache-2.11-cpan-39bf76dae61 )